text
stringlengths 87
880k
| pmid
stringlengths 1
8
| accession_id
stringlengths 9
10
| license
stringclasses 2
values | last_updated
stringlengths 19
19
| retracted
stringclasses 2
values | citation
stringlengths 22
94
| decoded_as
stringclasses 2
values | journal
stringlengths 3
48
| year
int32 1.95k
2.02k
| doi
stringlengths 3
61
| oa_subset
stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-401613139610.1186/1477-7827-3-40ResearchExogenous estradiol enhances apoptosis in regressing post-partum rat corpora lutea possibly mediated by prolactin Goyeneche Alicia A [email protected] Carlos M [email protected] Division of Basic Biomedical Sciences, University of South Dakota School of Medicine, Vermillion, South Dakota 57069, USA2005 30 8 2005 3 40 40 20 6 2005 30 8 2005 Copyright © 2005 Goyeneche and Telleria; licensee BioMed Central Ltd.2005Goyeneche and Telleria; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In pregnant rats, structural luteal regression takes place after parturition and is associated with cell death by apoptosis. We have recently shown that the hormonal environment is responsible for the fate of the corpora lutea (CL). Changing the levels of circulating hormones in post-partum rats, either by injecting androgen, progesterone, or by allowing dams to suckle, was coupled with a delay in the onset of apoptosis in the CL. The objectives of the present investigation were: i) to examine the effect of exogenous estradiol on apoptosis of the rat CL during post-partum luteal regression; and ii) to evaluate the post-partum luteal expression of the estrogen receptor (ER) genes.
Methods
In a first experiment, rats after parturition were separated from their pups and injected daily with vehicle or estradiol benzoate for 4 days. On day 4 post-partum, animals were sacrificed, blood samples were taken to determine serum concentrations of hormones, and the ovaries were isolated to study apoptosis in situ. In a second experiment, non-lactating rats after parturition received vehicle, estradiol benzoate or estradiol benzoate plus bromoergocryptine for 4 days, and their CL were isolated and used to study apoptosis ex vivo. In a third experiment, we obtained CL from rats on day 15 of pregnancy and from non-lactating rats on day 4 post-partum, and studied the expression of the messenger RNAs (mRNAs) encoding the ERalpha and ERbeta genes.
Results
Exogenous administration of estradiol benzoate induced an increase in the number of apoptotic cells within the CL on day 4 post-partum when compared with animals receiving vehicle alone. Animals treated with the estrogen had higher serum prolactin and progesterone concentrations, with no changes in serum androstenedione. Administration of bromoergocryptine blocked the increase in serum prolactin and progesterone concentrations, and DNA fragmentation induced by the estrogen treatment. ERalpha and ERbeta mRNAs were expressed in CL of day 4 post-partum animals at levels similar to those found in CL of day 15 pregnant animals.
Conclusion
We have established that estradiol accelerates apoptosis in the CL during post-partum luteal regression through a mechanism that possibly involves the secretion of pituitary prolactin. We have also shown that the post-partum rat CL express ERalpha and ERbeta mRNAs suggesting that they can be targeted by estrogen.
==== Body
Background
The regression of corpora lutea (CL) is a process that involves two stages. During the first stage (functional regression), production of progesterone is discontinued. In the second stage (structural regression), the CL undergo involution manifested by a decrease in weight and size that is associated with programmed death of the luteal cells [1-6]. In the rat CL, programmed cell death follows a pattern of death by apoptosis characterized by initial condensation of the nuclear chromatin followed or accompanied by nucleosomal fragmentation of DNA and formation of apoptotic bodies, which eventually are eliminated by phagocytosis [7,8].
In the regressing CL of pregnancy, apoptosis is a lengthy process that occurs over the course of many days from the initial decrease in the progesterone producing capacity of the glands, to the decrease in their sizes. As a consequence, the structural changes of the CL undergoing regression are usually studied after parturition [8-10]. The rat ovulates within 24–36 h following parturition [11]. Therefore, when studying luteal regression after parturition, two populations of CL can be analyzed simultaneously, the CL of previous pregnancy and the CL formed after post-partum ovulation [8,12]. We have shown previously that the two populations of CL found within the post-partum ovary have similar rate of apoptosis despite their difference in age [10].
The regression of the CL in the rat ovary after parturition is hormonally regulated. We demonstrated that luteal apoptosis in this species can be accelerated by the administration of either the antigestagen RU486 or prostaglandin F2α [7], both of which induce large declines in the capacity of the CL to produce progesterone. Conversely, we and others have shown that the onset of apoptosis in the post-partum CL can be delayed by administration of androstenedione [9], progesterone [10], or by allowing the dams to suckle [8,12].
During pregnancy in rats, circulating concentration of estradiol increases on day 3, after which remains very low until day 15–16 when it starts to increase progressively towards parturition [13,14]. Moreover, the pregnant rat CL express estrogen receptors (ERs) alpha (ERalpha) and beta (ERbeta) under the regulation of prolactin and placental lactogens [15], and respond to estradiol, which stimulates steroidogenesis [16,17], mediates luteal cell hypertrophy by increasing protein biosynthesis [18], and synergizes the luteotropic effect of prolactin [19]. Whether estradiol regulates luteal function during luteal regression, and whether ERs are expressed in the regressing CL of the post-partum rat, is presently unknown. Therefore, in the present investigation we studied the effect of exogenous estradiol benzoate on apoptosis of regressing CL post-partum and whether these CL express ERalpha and ERbeta mRNAs.
Materials and methods
Animals
Pregnant (day 1 = sperm positive) Sprague-Dawley rats were obtained from Harlan Labs (Indianapolis, IN, USA). They were housed under controlled conditions of light (lights on 05:00–17:00 h) and temperature (21–23°C) with free access to standard rat chow and water. Animals were killed by decapitation and handled in conformance with the Guide for the Care and Use of Laboratory Animals, National Academy of Sciences, USA, 1996. The experimental protocol was approved by the University of South Dakota Animal Care and Use Committee.
Experimental procedure
To determine the effect of estrogen on luteal apoptosis, two groups of post-partum rats were used, each composed of 6 to 8 animals. The pups were removed immediately after parturition, and the rats were injected daily with estradiol benzoate (5 μg/rat s.c.) or vehicle (sunflower seed oil) at 10:00 h. Animals were killed by decapitation at 13:00 h on day 4 post-partum. Trunk blood was obtained to determine hormone concentrations. The ovaries were removed and fixed for 1 h at room temperature in 4% paraformaldehyde, dehydrated in ethanol series, cleared in xylene, and embedded in paraffin for routine hematoxylin and eosin (H&E) staining. Luteal regression was evaluated separately in the two generations of CL (i.e., CL of the previous pregnancy and new CL formed after ovulation post-partum) by studying the number of nuclei undergoing apoptosis. Under the light microscope, the old CL of pregnancy have organized cell distribution and closed capillaries, whereas the newly formed CL have a less organized cell distribution and open capillaries. Two additional groups of post-partum rats, each composed of 6 to 8 animals, were treated and sacrificed as described above, and the CL were isolated from the ovaries under a stereoscopic microscope and weighed. The CL of the previous pregnancy were recognized from the newly formed CL as they were larger and less vascularized.
In a second experiment three groups of rats, each composed of 9 to 10 animals, were treated daily after parturition with vehicle, estradiol benzoate (5 μg/rat s.c.) or estradiol benzoate plus bromoergocryptine (0.5 mg/rat s.c.) at 10:00 h. Animals were killed by decapitation at 13:00 h on day 4 post-partum. Trunk blood was obtained to determine hormone concentrations. The ovaries were removed and the CL of previous pregnancy were isolated under a stereoscopic microscope, pooled and used for ex vivo incubation. The isolation of CL of the previous pregnancy was performed as previously reported [10]. From the analysis of the first experiment we concluded that the incidence of apoptosis in both CL subtypes evaluated in situ was similar. Therefore, for this experiment, only CL of the previous pregnancy were used, since they are larger and easier to isolate from the ovarian stroma than those formed after post-partum ovulation.
In a third experiment, 3 rats per group were sacrificed at 13:00 h on day 15 of pregnancy and on day 4 post-partum. In the latter group of rats, the pups were removed immediately following parturition. The CL were isolated from the ovaries under a stereoscopic microscope, frozen in liquid nitrogen, and stored at -80°C until processed for RNA isolation.
Counting of apoptotic cells
Apoptotic cells were counted in H&E stained tissue sections on the basis of morphological criteria using an optical microscope as previously described [8]. Briefly, only cells with advanced signs of apoptosis (i.e., containing multiple nuclear fragments) were counted. A microscope with a 100 × objective was used and all fields were analyzed in each CL for the presence of fragmented nuclei. All the CL in each section were studied, and an average number of apoptotic nuclei per high power field was obtained. The expression of apoptosis per field rather than per CL more accurately reflects the dynamics of the apoptotic process within each CL in a size-independent manner at any given time after parturition. This morphometric method for the identification of apoptotic cells was previously validated in the CL by in situ 3' end labeling [7].
Incubation of CL
The CL (four to six per well in a 24-well tissue-culture plate) were incubated in serum-free medium (McCoy 5A: Ham F12, 1:1, v/v; Sigma Chemical Co., St. Louis, MO, USA) containing 25 mM Hepes, 200 IU/ml penicillin G, 200 μg/ml streptomycin, and 0.5 μg/ml of amphotericin B at 37°C for various periods of time in an atmosphere of 95% air/5% CO2. After incubation, the CL were immediately frozen in liquid nitrogen and stored at -80°C until DNA isolation.
DNA fragmentation
The internucleosomal cleavage of the DNA was analyzed as follows: the CL of previous pregnancy were isolated on day 4 post-partum and were digested overnight at 50°C in a buffer composed of 100 mM NaCl, 10 mM Tris HCl (pH 8.0), 25 mM EDTA (pH 8.0), 0.5% SDS, and 0.1 mg/ml proteinase K (Life Technologies, Rockville, MD, USA). The genomic DNA was extracted from the digested tissues with phenol/chloroform/isoamyl alcohol (25:24:1, v/v/v), precipitated, and digested for 1 h at 37°C in 1 μg/ml of ribonuclease from bovine pancreas (deoxyribonuclease-free; Roche, Indianapolis, IN, USA). After extraction and precipitation, an equal amount of DNA for each sample (1 μg) was separated by electrophoresis on a 2% agarose gel, impregnated with SYBR Gold nucleic acid gel stain (Molecular Probes, Eugene, OR, USA), examined using an ultraviolet transilluminator, and photographed with the Amersham Typhoon fluorescence imaging system (Amersham Biosciences Corp., Piscataway, NJ, USA). A 100-base pair (bp) DNA ladder (Promega, Madison, WI, USA) was used for determining the size of the DNA fragments. The UN-SCAN-IT gel software (Silk Scientific, Inc., Orem, UT, USA) was used to semiquantitate the fragmented DNA. The densitometry of the DNA fragments that appeared below 2070 bp was recorded. This measurement was normalized against the density of the total genomic DNA of the sample to correct for DNA loading. The density of the total genomic DNA of each sample was calculated as the sum of the density of the DNA fragments below 2070 bp plus the density of the large DNA band found above the 2070 bp marker. In addition, to allow comparisons between different gels, the normalized densitometric values for each of the incubation times were divided by the normalized densitometric values at time zero of incubation. The samples of DNA used for time zero of incubation were obtained from CL that were isolated for ex vivo incubation but were frozen without being incubated.
Hormone assays
Estradiol concentrations were measured by radioimmunoassay (RIA) using a commercially obtained kit (Diagnostics Systems Laboratories, Webster, TX, USA). The sensitivity was 2.2 pg/ml, and the inter- and intra-assay coefficients of variation were 7.5% and 9.3% respectively. Androstenedione was assayed using a RIA previously described [9]. The sensitivity of the assay was 6 pg/tube and the intra- and inter-assay coefficients of variation were 4.4% and 16.7% respectively. Progesterone was assayed by enzyme immunoassay (EIA) using a commercially obtained kit (Cayman Chemical, Ann Arbor, MI, USA). The sensitivity of the assay was 10 pg/ml, and the inter- and intra-assay coefficients of variation were 5% and 6% respectively. Rat prolactin was assayed by EIA using a commercial kit (SPIbio, Massy Cedex, France). The sensitivity of the assay was 0.5 ng/ml and the intra- and inter-assay coefficients of variation were 8.1% and 14% respectively.
RNA isolation and reverse transcription polymerase chain reaction
The CL from rats on day 15 of pregnancy or on day 4 post-partum were obtained and total RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA, USA) per manufacturer's instructions. Reverse transcription (RT) reaction and polymerase chain reaction (PCR) were done using a SuperScript One-Step RT-PCR system with Platinum Taq (Invitrogen). PCR was carried out in a GeneAmp PCR 2700 thermal cycler (Applied Biosystems, Foster City, CA, USA) for 25 cycles using 61°C as the annealing temperature. The conditions were such that the amplification of the products was in the exponential phase, and the assay was linear with respect to the amount of input RNA. Oligonucleotide primer pairs were based on the sequence of the rat ERalpha gene (5'-AATTCTGACAATCGACGCCAG-3') and (5'-GTGCTTCAACATTCTCCCTCCTC-3'), the rat ERbeta gene (5'-AAAGCCAAGAGAAACGGTGGGCAT-3') and (5'-GCCAATCATGTGCACCAGTTCCTT-3'), and the rat ribosomal protein L19 (used as a housekeeping gene) (5'-CTGAAGGTCAAAGGGAATGTG-3') and (5'-GGACAGAGTCTTGATATCTC-3'), as previously described [15,20]. The predicted sizes of the PCR-amplified products were 344 (ERalpha), 204 (ERbeta) and 194 (L19). The PCR products were run on a 2.5% agarose gel, stained with SYBR Gold nucleic acid gel stain (Molecular Probes), and photographed with the Amersham Typhoon imaging system (Amersham). A 100 bp DNA ladder (Promega) was used for determining the size of the PCR products.
Statistics
Comparisons between means of two groups were carried out using Student t-tests. For multiple comparisons, one-way analysis of variance followed by Tukey's or Dunnet's multiple comparison test was used. A difference was considered to be statistically significant at P < 0.05.
Results
Potentiation of apoptosis by estradiol benzoate in the post-partum CL
To study the effect of estradiol on apoptosis during luteal regression, we used rats whose pups were removed immediately after delivery to prevent the initiation of lactation. Under these conditions, the animals undergo extensive luteal regression in both generations of CL (i.e., old CL of previous pregnancy and newly formed CL after post-partum ovulation) by day 3–4 post-partum [10,21]. Figure 1 shows representative images of H&E stained sections from CL of non-lactating animals treated with vehicle or estradiol benzoate and sacrificed on day 4 after parturition. The CL of vehicle-treated controls displayed few apoptotic nuclei per microscope field (Figure 1A), whereas treatment with estradiol benzoate markedly enhanced the number of apoptotic figures (Figure 1B). The images depicted in this figure correspond to an old CL of pregnancy, but similar images were observed in newly formed CL after post-partum ovulation (data not shown). Figures 2A and 2B show the quantification of apoptotic nuclei found within both generations of CL in vehicle-treated rats and in rats receiving estradiol benzoate. The average weight of the CL seen in control animals on day 4 post-partum was not affected by treatment with the estrogen (Figures 2C and 2D).
Figure 1 Micrographs of CL obtained from rats killed on day 4 post-partum and stained with H&E. Panel (A) shows the structure of a an old CL of pregnancy obtained from an animal treated with vehicle, whereas panel (B) displays the structure of an old CL of pregnancy obtained from a rat treated with estradiol benzoate. Similar results were observed in newly formed CL after post-partum ovulation obtained from animals receiving vehicle or estradiol benzoate. The arrows indicate nuclei undergoing apoptosis displaying different features of chromatin condensation. Magnification × 400.
Figure 2 Effect of treatment with estradiol benzoate on CL apoptosis and weight in the rat ovary after parturition. Ovaries were obtained at 13:00 hon day 4 post-partum from animals that had received vehicle (V) [non-lactating animals treated daily with sunflower seed oil at 10:00 h (n = 6–8)], or estradiol benzoate (EB) [non-lactating animals treated daily with estradiol benzoate (5 μg/rat s.c.) at 1000 h (n = 6–8)]. The ovaries were processed for routine H&E staining, and the number of apoptotic figures was counted under a light microscope (A and B). The average weight of the CL was recorded in two other groups of 6 to 8 animals (C and D). Filled bars represent old CL of gestation, whereas open bars represent newly formed CL after ovulation post-partum. *** p < 0.001 compared with vehicle (Student t-test).
Circulating concentrations of 17beta-estradiol, androstenedione, progesterone and prolactin in animals treated with estradiol benzoate after parturition
Animals that were treated with estradiol benzoate, as expected, displayed elevated circulating levels of 17beta-estradiol (Figure 3A). Androstenedione, shown to be capable of interfering with luteal regression in post-partum rat CL [9], did not change with estrogen treatment (Figure 3B). Prolactin, one of the main hormones involved in CL function in rats, had a significant increase in its circulating concentration after treatment with estradiol benzoate (Figure 3C), in parallel with a significant increase in circulating levels of progesterone (Figure 3D).
Figure 3 Serum hormone concentrations in post-partum non-lactating animals treated with vehicle or estradiol benzoate. 17beta-estradiol (A), androstenedione (B), prolactin (C) and progesterone (D) concentrations in serum of non-lactating animals treated daily at 10:00 h with either vehicle (V) or estradiol benzoate (EB; 5 μg/rat s.c.), and killed at 13:00 h on day 4 post-partum. *** p < 0.001 compared with vehicle-treated animals (Student t-test).
Bromoergocryptine abrogates estradiol benzoate-induced increase in serum prolactin and progesterone concentrations in post-partum rats
Because it is known that in pregnant rats prolactin stimulates the production of progesterone by the CL [22], we evaluated whether the enhanced circulating levels of progesterone in response to treatment with estradiol benzoate were the consequence of a direct effect of the estrogen on the ovary, or instead, an indirect effect mediated by prolactin secreted in response to estradiol. To answer this question, we injected post-partum animals with estradiol benzoate plus bromoergocryptine in order to block the production of pituitary prolactin induced by the estrogen. The serum concentrations of prolactin and progesterone, which were increased by administration of estradiol benzoate, were abrogated by bromoergrocryptine, reaching control levels (Table 1). These results suggest that the increase in the production of progesterone in estrogen-treated animals is most likely a consequence of an effect of prolactin rather than a direct effect of estradiol on the ovary.
Table 1 Effect of bromoergrocryptine (BEC) on serum concentrations of progesterone and prolactin in post-partum non-lactating rats treated with estradiol benzoate (EB).
Vehicle EB EB + BEC
Progesterone (ng/ml) 45.9 ± 7.2 (10) 154 ± 13.3 (10) *** 44.6 ± 6.7 (9)
Prolactin (ng/ml) 30.2 ± 4.2 (10) 166 ± 19.2 (10) *** 25.5 ± 6.5 (9)
Values are mean ± SEM for the number of animals in parentheses. *** p < 0.001 compared with groups receiving vehicle or EB + BEC (one-way analysis of variance followed by the Tukey multiple comparison test).
Bromoergocryptine abrogates estradiol benzoate enhancement of ex vivo-induced luteal DNA fragmentation
To further study the effect of estradiol on apoptotic cell death in the rat CL, we used a previously defined ex vivo approach in which CL incubated in serum-free conditions accumulate large number of cells in different stages of apoptosis. This approach is sensitive to study hormonal regulation of apoptosis in the CL [8-10,23]. It takes 2 to 4 h of organ culture to observe DNA fragmentation [8], whereas the length of this period depends upon the hormonal environment to which the gland had been exposed in vivo. In the current work, old CL obtained from non-lactating rats on day 4 post-partum and incubated in serum-free conditions displayed DNA fragmentation in a time-dependent manner (Figure 4A, left panel, and 4B). However, in non-lactating rats that had received a daily injection of estradiol benzoate from days 1 to 4 post-partum, the ex vivo-induced luteal DNA fragmentation was markedly accelerated and more abundant (Figure 4A, middle panel, and 4B). When the animals were treated with estradiol benzoate in conjunction with bromoergocryptine, the temporal pattern of DNA fragmentation ex vivo, as well as the abundance of fragmented DNA, was comparable to that of vehicle-treated animals (Figure 4A, right panel, and 4B). These data indicate that the blockage of estradiol-induced prolactin secretion prevents fragmentation of DNA induced by the estrogen, and suggest that the pro-apoptotic signal triggered by estrogen on the post-partum CL is most likely the consequence of the action of prolactin.
Figure 4 Effect of estradiol benzoate or estradiol benzoate plus bromoergocryptine on DNA fragmentation in CL incubated in serum-free conditions. At the end of the incubation, genomic DNA was extracted from vehicle-treated, estradiol benzoate (EB)-treated or EB plus bromoergrocriptine (BEC)-treated animals and run on a gel (Panel A). Densitometry of the DNA fragments studied from three different experiments with similar outcome is also shown (Panel B). * p < 0.05, ** p < 0.01 and *** p < 0.001 compared with values at time zero (one way analysis of variance followed by the Dunnett multiple-comparison test). † P < 0.05 and †† P < 0.01 compared with the corresponding time across treatments (one way analysis of variance followed by the Tukey multiple comparison test). bp, Base pairs. Only CL of previous pregnancy were evaluated.
Expression of ER in the post-partum rat CL
To determine whether ERalpha and ERbeta genes are expressed in the rat CL post-partum, we performed semiquantitative RT-PCR to specifically detect the ERalpha and ERbeta mRNA transcripts. As a positive control for ER mRNA expression we used CL obtained from day 15 pregnant rats, which were previously described as expressing both ER mRNA transcripts [15]. Figure 5A shows that ERalpha and ERbeta mRNA transcripts are highly expressed 4 days after parturition in both types of CL found within the ovaries. Semiquantitative analysis (Figure 5B) shows no difference between the abundance of each ER transcript on the different days analyzed as well as when comparing new and old types of CL on day 4 post-partum. These data also indicate that the post-partum CL can be direct targets of estrogen action.
Figure 5 Expression of ERalpha and ERbeta mRNA in the two populations of CL found in the post-partum ovaries. Old indicates CL from previous pregnancy. New indicates newly formed CL after post-partum ovulation. Total RNA was isolated from old or new CL on day 4 post-partum, and from CL obtained from animals on day 15 of pregnancy, and was analyzed by RT-PCR. In panel (A) a representative gel is shown. Results were quantified by densitometry and corrected using L19. Normalized mRNA levels are graphically represented in panel (B) as the mean ± SEM (n = 3).
Discussion
The pro- or anti-apoptotic effects of estrogens are controversial. For example, estradiol enhances neuronal survival [24,25], but sensitizes anterior pituitary gland to apoptosis [26]. We have shown that estradiol enhances apoptosis in regressing CL of rats. In rabbits, however, estradiol protects the CL from apoptosis [27]. Thus, the effect of estradiol on apoptosis appears to be cell and species dependent.
A role for estradiol in the induction of apoptosis has been suggested for the CL of primates. Prior to the induction of apoptosis the human CL express high levels of 17beta-hydroxysteroid dehydrogenase type I, the enzyme that converts estrone to estradiol [28], suggesting a role for locally produced estradiol in triggering luteal regression. An increase in the sensitivity to estradiol has also been proposed in the CL of monkeys coincident with luteal regression [29]. Further, in a recent review on apoptosis in the human ovary, a role for estradiol has been suggested in triggering luteal regression [30]. Our results in regressing rat CL support a similar effect of estradiol to that shown in humans, stimulating luteal regression through the promotion of luteal apoptosis, and suggest that the elevated estrogenic environment to which the CL of rats are exposed at the time of parturition [13,14] may play a role in the post-partum facet of luteal regression.
In our studies, the pro-apoptotic effect of estradiol appears to rely on the presence of pituitary prolactin. The stimulation of prolactin secretion by estradiol in rats has been demonstrated in several in vivo settings. For example, estradiol implants made in the arquate nucleus provoke a sustained release of prolactin [31], whereas systemic administration of estradiol benzoate in ovariectomized rats stimulates prolactin secretion by a mechanism that involves the opioid system [32]. Furthermore, estradiol can also stimulate prolactin synthesis and release by directly targeting the lactotrophes [33].
One interesting finding in our study is that exogenous estradiol increased apoptosis in the CL while at the same time increased serum progesterone levels. An interaction between estrogen, progesterone and prolactin has been shown during the regression of the CL of the rat estrous cycle [34]. These authors demonstrated that antiestrogens decrease the detrimental effect of prolactin on the CL, whereas progesterone favors the action of prolactin, suggesting that an estrogenic and progestational hormonal environment favors apoptosis induced by prolactin. Our results are in keeping with this because the increased apoptosis induced by estradiol in the post-partum rat CL also occurred in the presence of elevated concentrations of progesterone and prolactin.
The effect of estrogen on progesterone production in rats varies with the experimental approach. For example, in hypophysectomized and hysterectomized rats estradiol stimulates the production of progesterone [35]. In contrast, in pseudopregnant rats, treatment with estradiol benzoate decreases plasma levels of progesterone [36], and, in pregnant rats, estradiol given from days 7 to 14 of pregnancy reduces the levels of circulating progesterone measured on day 15 of pregnancy [37]. In our study using intact rats after parturition estrogen treatment increased progesterone production and such effect seems mediated by pituitary prolactin because it could be abrogated by bromoergocryptine.
Prolactin has different effects on the CL depending on the prevailing steroid concentrations. For example during lactation, prolactin prevents apoptosis in the CL [8,12]. This effect of prolactin occurs in the presence of high circulating levels of progesterone [8], at least until day 9 of lactation [12]. In this physiological situation, estradiol levels are low and beginning to increase towards the 8th to 10th day of lactation [38]. Conversely, our studies in non-lactating post-partum rats receiving estradiol suggest that prolactin stimulates rather than prevents luteal apoptosis, and that this effect takes place in the presence of high progesterone. Together these studies suggest that prolactin may stimulate progesterone production, but at the same time may either trigger or prevent luteal apoptosis depending upon the estrogenic background. Whereas a high estrogenic background may favor a pro-apoptotic action of prolactin (e.g. as in the cyclic rat CL), a low estrogenic background may favor an anti-apoptotic effect of the hormone (e.g. as in the CL of lactation). Further studies need to be done to prove or disprove this hypothesis, but answering this question may resolve the unclear issue that prolactin appears to have a dual effect in the rat CL, being both a survival factor [8,20,22,39] and a pro-apoptotic factor [3,40-46]. It is feasible that opposite actions of prolactin in the CL, driven by the estrogenic environment, are mediated through different signal transduction pathways activated by the lactogenic hormone upon binding to its luteal receptor subtypes [20,47], or by differential activation and turnover of transcription factors [48].
A direct effect of prolactin on the regressing CL after parturition is supported by the presence of receptors for prolactin in both old CL of pregnancy and newly formed CL after ovulation post-partum [8]. Whereas the expression of prolactin receptors in estrogen-treated animals was not measured in the present study, we can anticipate their presence. This is because prolactin has been shown to up-regulate its own receptors in the CL [20]; thus, it is unlikely that in an environment high in prolactin as observed in estrogen-treated rats, those receptors would have been reduced.
In the present study, we have shown that the ER genes are expressed in post-partum CL making them potential direct targets for the action of estradiol. As we only studied the expression of the ER mRNAs, we cannot conclusively indicate that the ER proteins follow similar expression pattern. Yet, most probably the latter assumption is true because a correlation was shown between expression of ER message and protein in the pregnant rat CL [15]. The rat CL express ERalpha and ERbeta mRNAs and ER immunoreactive proteins along pregnancy, with a decline in expression occurring at parturition [15]. Because the post-partum CL express ER transcripts at levels similar to the levels measured at mid-pregnancy, it can be suggested that the down-regulation of ER expression in the rat CL is limited to parturition.
Previously, we demonstrated that the main circulating androgen in female rats, androstenedione, protects the post-partum CL from undergoing apoptosis [9]. In the present report we show that estradiol enhances apoptosis in the same CL types. Taken together, the results of these two studies suggest that androgens and estrogens oppose each other's action on luteal apoptosis. This opposite effect of androgens and estrogens also occurs in granulosa cells, with however, a different pattern. Rat granulosa cells maintained in culture respond to testosterone with an increase in apoptosis, and to estradiol with a decrease in apoptosis [49]. Whereas an increase in the androgen to estrogen ratio in the follicular fluid favors apoptosis of granulosa cells and overall follicular atresia [50,51], a similar change in ratio within the CL may protect the gland from regression.
Conclusion
We have shown that estradiol increases apoptosis in the rat CL during post-partum luteal regression and presented evidence suggesting that pituitary prolactin is the mediator of such effect. We also demonstrated the expression of ERalpha and ERbeta mRNA transcripts in the two populations of CL found in this species after parturition (i.e. the newly formed CL after post-partum ovulation and the old CL of pregnancy). Moreover, the data suggest that the estrogenic environment might be a key factor driving a detrimental effect of prolactin in the post-partum CL.
Authors' contributions
AAG participated in the design of the study and carried out most of the experiments. CMT conceived the study, measured hormone concentrations, carried out some of the experiments, and wrote the manuscript. Both authors read and approved the final manuscript.
Acknowledgements
We are grateful to Dr. Barbara Goodman for the critical revision of the manuscript. This research was supported by grants 2 P20 RR016479 from the INBRE Program, NIH/NCRR, and Internal Funds from the University of South Dakota School of Medicine.
==== Refs
Juengel JL Garverick HA Johnson AL Youngquist RS Smith MF Apoptosis during luteal regression in cattle Endocrinology 1993 132 249 254 8419126 10.1210/en.132.1.249
Rueda BR Wegner JA Marion SL Wahlen DD Hoyer PB Internucleosomal DNA fragmentation in ovine luteal tissue associated with luteolysis: in vivo and in vitro analyses Biol Reprod 1995 52 305 312 7711200
Matsuyama S Chang KT Kanuka H Ohnishi M Ikeda A Nishihara M Takahashi M Occurrence of deoxyribonucleic acid fragmentation during prolactin-induced structural luteolysis in cycling rats Biol Reprod 1996 54 1245 1251 8724351
Rueda BR Tilly KI Botros IW Jolly PD Hansen TR Hoyer PB Tilly JL Increased bax and interleukin-1beta-converting enzyme messenger ribonucleic acid levels coincide with apoptosis in the bovine corpus luteum during structural regression Biol Reprod 1997 56 186 193 9002648
McCormack JT Friederichs MG Limback SD Greenwald GS Apoptosis during spontaneous luteolysis in the cyclic golden hamster: biochemical and morphological evidence Biol Reprod 1998 58 255 260 9472949
Davis JS Rueda BR The corpus luteum: an ovarian structure with maternal instincts and suicidal tendencies Front Biosci 2002 7 d1949 78 12161347
Telleria CM Goyeneche AA Cavicchia JC Stati AO Deis RP Apoptosis induced by antigestagen RU486 in rat corpus luteum of pregnancy Endocrine 2001 15 147 155 11720240 10.1385/ENDO:15:2:147
Goyeneche AA Martinez IL Deis RP Gibori G Telleria CM In vivo hormonal environment leads to differential susceptibility of the corpus luteum to apoptosis in vitro Biol Reprod 2003 68 2322 2330 12606365 10.1095/biolreprod.102.013540
Goyeneche AA Calvo V Gibori G Telleria CM Androstenedione interferes in luteal regression by inhibiting apoptosis and stimulating progesterone production Biol Reprod 2002 66 1540 1547 11967221
Goyeneche AA Deis RP Gibori G Telleria CM Progesterone promotes survival of the rat corpus luteum in the absence of cognate receptors Biol Reprod 2003 68 151 158 12493707 10.1095/biolreprod.102.007898
Rothchild I The regulation of the mammalian corpus luteum Recent Prog Horm Res 1981 37 183 298 7025133
Takiguchi S Sugino N Esato K Karube-Harada A Sakata A Nakamura Y Ishikawa H Kato H Differential regulation of apoptosis in the corpus luteum of pregnancy and newly formed corpus luteum after parturition in rats Biol Reprod 2004 70 313 318 14522835 10.1095/biolreprod.103.018853
Yoshinaga K Hawkins RA Stocker JF Estrogen secretion by the rat ovary in vivo during the estrous cycle and pregnancy Endocrinology 1969 85 103 112 5815010
Shaikh AA Estrone and estradiol levels in the ovarian venous blood from rats during the estrous cycle and pregnancy Biol Reprod 1971 5 297 307 5163522
Telleria CM Zhong L Deb S Srivastava RK Park KS Sugino N Park-Sarge OK Gibori G Differential expression of the estrogen receptors alpha and beta in the rat corpus luteum of pregnancy: regulation by prolactin and placental lactogens Endocrinology 1998 139 2432 2442 9564855 10.1210/en.139.5.2432
Khan I Belanger A Chen YD Gibori G Influence of high-density lipoprotein on estradiol stimulation of luteal steroidogenesis Biol Reprod 1985 32 96 104 3971012
McLean MP Puryear TK Khan I Azhar S Billheimer JT Orly J Gibori G Estradiol regulation of sterol carrier protein-2 independent of cytochrome P450 side-chain cleavage expression in the rat corpus luteum Endocrinology 1989 125 1337 1344 2474437
Rao MC Palfrey HC Nash NT Greisman A Jayatilak PG Gibori G Effects of estradiol on calcium-specific protein phosphorylation in the rat corpus luteum Endocrinology 1987 120 1010 1018 3803307
Gibori G Richards JS Keyes PL Synergistic effects of prolactin and estradiol in the luteotropic process in the pregnant rat: regulation of estradiol receptor by prolactin Biol Reprod 1979 21 419 423 582804
Telleria CM Parmer TG Zhong L Clarke DL Albarracin CT Duan WR Linzer DI Gibori G The different forms of the prolactin receptor in the rat corpus luteum: developmental expression and hormonal regulation in pregnancy Endocrinology 1997 138 4812 4820 9348210 10.1210/en.138.11.4812
Guo K Wolf V Dharmarajan AM Feng Z Bielke W Saurer S Friis R Apoptosis-associated gene expression in the corpus luteum of the rat Biol Reprod 1998 58 739 746 9510961
Risk M Gibori G Horseman ND Mechanisms of luteal cell regulation by prolactin Prolactin 2001 New York , Kluwer Academic 265 295
Dharmarajan AM Hisheh S Singh B Parkinson S Tilly KI Tilly JL Antioxidants mimic the ability of chorionic gonadotropin to suppress apoptosis in the rabbit corpus luteum in vitro: a novel role for superoxide dismutase in regulating bax expression Endocrinology 1999 140 2555 2561 10342842 10.1210/en.140.6.2555
Green PS Simpkins JW Neuroprotective effects of estrogens: potential mechanisms of action Int J Dev Neurosci 2000 18 347 358 10817919 10.1016/S0736-5748(00)00017-4
Won CK Ha SJ Noh HS Kang SS Cho GJ Choi WS Koh PO Estradiol prevents the injury-induced decrease of Akt activation and Bad phosphorylation Neurosci Lett 2005 387 115 119 16087293
Pisera D Candolfi M Navarra S Ferraris J Zaldivar V Jaita G Castro MG Seilicovich A Estrogens sensitize anterior pituitary gland to apoptosis Am J Physiol Endocrinol Metab 2004 287 E767 71 15172886 10.1152/ajpendo.00052.2004
Goodman SB Kugu K Chen SH Preutthipan S Tilly KI Tilly JL Dharmarajan AM Estradiol-mediated suppression of apoptosis in the rabbit corpus luteum is associated with a shift in expression of bcl-2 family members favoring cellular survival Biol Reprod 1998 59 820 827 9746731
Vaskivuo TE Ottander U Oduwole O Isomaa V Vihko P Olofsson JI Tapanainen JS Role of apoptosis, apoptosis-related factors and 17beta-hydroxysteroid dehydrogenases in human corpus luteum regression Mol Cell Endocrinol 2002 194 191 200 12242042 10.1016/S0303-7207(02)00087-4
Duffy DM Chaffin CL Stouffer RL Expression of estrogen receptor alpha and beta in the rhesus monkey corpus luteum during the menstrual cycle: regulation by luteinizing hormone and progesterone Endocrinology 2000 141 1711 1717 10803581 10.1210/en.141.5.1711
Vaskivuo TE Tapanainen JS Apoptosis in the human ovary Reprod Biomed Online 2003 6 24 35 12626140
Caron RW Deis RP Estradiol implants in the arcuate nucleus induce lactogenesis in virgin rats. Role of progesterone Life Sci 1998 62 229 237 9488101 10.1016/S0024-3205(97)01091-6
Soaje M Deis RP Participation of the opioid system in the regulation of prolactin secretion in androgenized rats: effect of ovarian steroids Eur J Pharmacol 1999 371 43 49 10355593 10.1016/S0014-2999(99)00152-1
Fink G Oestrogen and progesterone interactions in the control of gonadotrophin and prolactin secretion J Steroid Biochem 1988 30 169 178 3290579 10.1016/0022-4731(88)90090-8
Gaytan F Morales C Bellido C Aguilar R Millan Y Martin De Las Mulas J Sanchez-Criado JE Progesterone on an oestrogen background enhances prolactin-induced apoptosis in regressing corpora lutea in the cyclic rat: possible involvement of luteal endothelial cell progesterone receptors J Endocrinol 2000 165 715 724 10828856 10.1677/joe.0.1650715
Khan I Glaser LA Gibori G Reactivation of regressing corpora lutea by estradiol in the pregnant rat: dependence on placental lactogen Biol Reprod 1987 37 1083 1088 3442688
de Greef WJ Uilenbroek JT de Jong FH Effect of oestradiol benzoate or luteinizing hormone on levels of progesterone in the pseudopregnant rat J Endocrinol 1982 94 369 379 7119647
Tamura H Role of the nongravid part of the uterus in the luteolytic effects of estrogen in pregnant rats Endocrinol Jpn 1983 30 615 620 6671462
Smith MS Neill JD Inhibition of gonadotropin secretion during lactation in the rat: relative contribution of suckling and ovarian steroids Biol Reprod 1977 17 255 261 889995
Piekorz RP Gingras S Hoffmeyer A Ihle JN Weinstein Y Regulation of progesterone levels during pregnancy and parturition by signal transducer and activator of transcription 5 and 20alpha-hydroxysteroid dehydrogenase Mol Endocrinol 2005 19 431 440 15471942 10.1210/me.2004-0302
Gaytan F Morales C Bellido C Aguilar E Sanchez-Criado JE Role of prolactin in the regulation of macrophages and in the proliferative activity of vascular cells in newly formed and regressing rat corpora lutea Biol Reprod 1997 57 478 486 9241066
Kanuka H Matsuyama S Ohnishi M Matsumoto Y Nishihara M Takahashi M Prolactin expresses differential effects on apoptotic cell death of luteal cells in vivo and in vitro Endocr J 1997 44 11 22 9152610
Bowen JM Towns R Warren JS Landis Keyes P Luteal regression in the normally cycling rat: apoptosis, monocyte chemoattractant protein-1, and inflammatory cell involvement Biol Reprod 1999 60 740 746 10026125
Gaytan F Bellido C Morales C Sanchez-Criado JE Luteolytic effect of prolactin is dependent on the degree of differentiation of luteal cells in the rat Biol Reprod 2001 65 433 441 11466211
Bowen JM Keyes PL Warren JS Townson DH Prolactin-induced regression of the rat corpus luteum: expression of monocyte chemoattractant protein-1 and invasion of macrophages Biol Reprod 1996 54 1120 1127 8722634
Gaytan F Bellido C Morales C Sanchez-Criado JE Cyclic changes in the responsiveness of regressing corpora lutea to the luteolytic effects of prolactin in rats Reproduction 2001 122 411 417 11603369 10.1530/rep.0.1220411
Kuranaga E Kanuka H Bannai M Suzuki M Nishihara M Takahashi M Fas/Fas ligand system in prolactin-induced apoptosis in rat corpus luteum: possible role of luteal immune cells Biochem Biophys Res Commun 1999 260 167 173 10381361 10.1006/bbrc.1999.0858
Binart N Imbert-Bollore P Baran N Viglietta C Kelly PA A Short Form of the Prolactin (PRL) Receptor Is Able to Rescue Mammopoiesis in Heterozygous PRL Receptor Mice Mol Endocrinol 2003 17 1066 1074 12624115 10.1210/me.2002-0181
Russell DL Richards JAS Differentiation-Dependent Prolactin Responsiveness and Stat (Signal Transducers and Activators of Transcription) Signaling in Rat Ovarian Cells Mol Endocrinol 1999 13 2049 2064 10598581 10.1210/me.13.12.2049
Billig H Furuta I Hsueh AJ Estrogens inhibit and androgens enhance ovarian granulosa cell apoptosis Endocrinology 1993 133 2204 2212 8404672 10.1210/en.133.5.2204
Carson RS Findlay JK Clarke IJ Burger HG Estradiol, testosterone, and androstenedione in ovine follicular fluid during growth and atresia of ovarian follicles Biol Reprod 1981 24 105 113 7470536
Maxson WS Haney AF Schomberg DW Steroidogenesis in porcine atretic follicles: loss of aromatase activity in isolated granulosa and theca Biol Reprod 1985 33 495 501 4041532
|
16131396
|
PMC1236958
|
CC BY
|
2021-01-04 16:37:13
|
no
|
Reprod Biol Endocrinol. 2005 Aug 30; 3:40
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-40
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-441614657010.1186/1477-7827-3-44ResearchRats with steroid-induced polycystic ovaries develop hypertension and increased sympathetic nervous system activity Stener-Victorin Elisabet [email protected] Karolina [email protected] Britt-Mari [email protected]äng Agneta [email protected] Cardiovascular Institute and Wallenberg Laboratory, Sahlgrenska Academy, Göteborg University, SE-413 45 Göteborg, Sweden2 Department of Obstetrics and Gynaecology, Sahlgrenska University Hospital, Sahlgrenska, SE-413 45 Göteborg, Sweden3 Institute of Occupational Therapy and Physical Therapy, Sahlgrenska Academy, Göteborg University, SE-405 30 Göteborg, Sweden2005 7 9 2005 3 44 44 30 6 2005 7 9 2005 Copyright © 2005 Stener-Victorin et al; licensee BioMed Central Ltd.2005Stener-Victorin et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Polycystic ovary syndrome (PCOS) is a complex endocrine and metabolic disorder associated with ovulatory dysfunction, abdominal obesity, hyperandrogenism, hypertension, and insulin resistance.
Methods
Our objectives in this study were (1) to estimate sympathetic-adrenal medullary (SAM) activity by measuring mean systolic blood pressure (MSAP) in rats with estradiol valerate (EV)-induced PCO; (2) to estimate alpha1a and alpha2a adrenoceptor expression in a brain area thought to mediate central effects on MSAP regulation and in the adrenal medulla; (3) to assess hypothalamic-pituitary-adrenal (HPA) axis regulation by measuring adrenocorticotropic hormone (ACTH) and corticosterone (CORT) levels in response to novel-environment stress; and (4) to measure abdominal obesity, sex steroids, and insulin sensitivity.
Results
The PCO rats had significantly higher MSAP than controls, higher levels of alpha1a adrenoceptor mRNA in the hypothalamic paraventricular nucleus (PVN), and lower levels of alpha2a adrenoceptor mRNA in the PVN and adrenal medulla. After exposure to stress, PCO rats had higher ACTH and CORT levels. Plasma testosterone concentrations were lower in PCO rats, and no differences in insulin sensitivity or in the weight of intraabdominal fat depots were found.
Conclusion
Thus, rats with EV-induced PCO develop hypertension and increased sympathetic and HPA-axis activity without reduced insulin sensitivity, obesity, or hyperandrogenism. These findings may have implications for mechanisms underlying hypertension in PCOS.
==== Body
Introduction
Polycystic ovary syndrome (PCOS), a heterogeneous endocrine and metabolic disorder affecting 6% to 10% of women of reproductive age, is associated with ovulatory dysfunction, abdominal obesity, hyperandrogenism, and in many cases hypertension and profound insulin resistance [1]. Several of these factors increase the risk of cardiovascular disease (CVD) in women [2]. The prevalence of hypertension in women with PCOS is around 40% [3,4]. PCOS is also associated with a higher risk of myocardial infarction (relative risk, 7.4) [4] and, in young women, a compromised cardiovascular profile, independent of obesity [5]. Women with PCOS also develop a hyperactive response to stress [6]. The heterogeneity of the disorder suggests that there are subpopulations within the syndrome. Hypertension and insulin resistance are not uniformly present, and hypertension may be absent despite profound insulin resistance and vice versa [7]. It is not clear whether PCOS increases the risk of CVD independently of the metabolic syndrome [8].
The abnormalities detected in PCOS have been attributed to primary defects in the hypothalamic-pituitary-adrenal (HPA) axis, the ovarian microenvironment, the adrenal gland, and the insulin/insulin-like growth factor metabolic regulatory system [1]. Hypertension and defects in insulin action may be related to enhanced sympathetic-adrenal medullary (SAM) activity [9]. Increased sympathetic nervous system activity may be associated with enhanced α-adrenergic responsiveness, contributing to elevated blood pressure and hypertension [10,11], and may play a role in PCOS [6,12,13]. If so, the α1a and α2a adrenoceptor (AR) subtypes are of interest since they are strongly implicated in cardiovascular control [14,15].
The etiology of PCOS is unknown, likely reflecting multiple pathophysiological mechanisms, and an accepted animal model of the disease has not been established. A recent review concluded that multiple models may be needed, depending on whether the goal is to investigate ovarian morphology or a particular PCOS-related disorder [16]. In both mechanistic and treatment studies, we and others have used a model in which PCO is induced by a single intramuscular (i.m.) injection of estradiol valerate (EV) in 8-week-old rats [17]. The rats cease ovulating and develop characteristics of human PCOS, including large cystic follicles in the ovaries and altered concentrations of luteinizing hormone [17].
Central neuronal activity in norepinephrine (NE) neurons is increased in rats with EV-induced PCO [18], suggesting increased central sympathetic outflow. Increased ovarian sympathetic tone in rats with EV-induced PCO has been evidenced by elevations in tyrosine hydroxylase activity and NE concentration, downregulation of the β2 AR, and increased production of ovarian nerve growth factor, a target-derived neurotrophin [19-25].
The stress response is largely mediated by the SAM and HPA axes. Activation of the HPA axis increases the secretion of adrenocorticotropic hormone (ACTH) and leads to glucocorticoid release from the adrenal cortex. The HPA axis and intra-adrenal mechanisms involving the adrenal medulla might also regulate adrenocortical steroidgenesis. Furthermore, adrenocortical steroids influence the development and the function of adrenomedullary chromaffin cells and vice versa [26]. Activation of the SAM axis, however, increases the release of epinephrine and NE from the adrenal medulla and stimulates the sympathetic norepinergic nerves, increasing NE secretion.
These two axes form the key efferent links in the "defeat" (HPA axis) and "defense" (SAM axis) reactions [27,28]. Chronic activation of either axis can lead to stress-related diseases such as hypertension, diabetes, and obesity [27,28].
In this study, we investigated SAM-axis activity and HPA-axis regulation in rats with EV-induced PCO. To estimate SAM activity, we measured systolic blood pressure and the expression of α1a- and α2a-AR mRNAs in the hypothalamic paraventricular nucleus (PVN) and adrenal medulla. To assess HPA-axis regulation, we measured ACTH and corticosterone (CORT) concentrations in response to novel-environment stress. We also measured abdominal obesity, sex steroid levels, and insulin sensitivity.
Materials and methods
Animals
Eighteen virgin adult cycling Wistar-Kyoto rats (B & K Universal AB, Sweden) weighing 190 to 210 g were divided into two groups and housed four to a cage under standard conditions (21 ± 2°C, 50% to 60% humidity, 12-hour light/12-hour dark cycle) for at least 1 week before and throughout the study, with free access to standard chow and tap water. The study was approved by the Animal Ethics Committee of Göteborg University and performed in accordance with the NIH "Guide for the Care and Use of Laboratory Animals."
Hormonal Treatment and Study Procedure
After 1 week of acclimatization, 8-week-old rats (n = 10) each received an i.m. injection of EV (Riedeldehaen, Germany), 4 mg in 0.2 mL of oil (arachidis oleum, Apoteket AB, Umeå, Sweden), to induce PCO [17]; this dose induces persistent estrus and permanent PCO, i.e. no fresh corpora lutea, regressing old corpora lutea and atretic follicles [22]. Controls (n = 8) received vehicle only. Rats were weighed weekly. Mean systolic arterial pressure (MSAP) and heart rate (HR) were measured 2, 3, 5, and 7 weeks after the injection. Stress tests were carried out at 6 weeks. Blood samples for measurement of progesterone, 17β-estradiol, and testosterone were obtained at week 7. Euglycemic hyperinsulinemic clamp tests were performed at 10 to 11 weeks, the time point when PCO morphology is fully developed[22].
Vaginal Smear
For 10 days before the stress and clamp tests, estrous cyclicity was monitored by vaginal smears obtained between 0800 and 1200 hours. The rat estrous cycle (estrus, diestrus 1, diestrus 2, and proestrus) usually lasts about 4 days. In controls, the stress and clamp tests were performed during estrus, and sex hormone levels were measured during estrus. In PCO rats, all test procedures were carried out and blood samples collected during estrus or pseudoestrus as described [19]
Blood Pressure and HR Measurements
Systolic arterial pressure and HR were measured between 0800 and 1200 hours at 2, 3, 5, and 7 weeks after EV injection with a tail cuff monitor (RTBP Monitor, Harvard Apparatus, South Natick, MA) with a light-emitting diode and a photoresistor connected to a dual-channel recorder. The rats were conscious and placed on a heating pad, and their tails were warmed with a heating lamp for 10 minutes to cause vasodilatation for an optimal signal. MSAP was calculated from three consecutive stable recordings.
ACTH and CORT Secretion
At 6 weeks, ACTH and CORT responses to novel-environment stress were assessed as described [29]. All tests started between 0700 and 0900 h, and care was taken to keep the rats undisturbed and fed the night before the experiment. The next morning, each rat was placed in a novel environment (new test cage, loud background noise). Blood (60 μL) for measurement of ACTH and CORT was obtained from a tail vein before and 15, 30, 60, 90, and 120 minutes after exposure to the novel environment.
Euglycemic Hyperinsulinemic Clamp Test
At 10 to 11 weeks after injection, rats underwent euglycemic hyperinsulinemic clamp tests as described [30]. Anesthesia was induced with thiobutabarbital sodium (Inactin, 125 mg/kg body weight; RBI, Natick, MA), and catheters were inserted into the left carotid artery for blood sampling and into the right jugular vein for glucose and insulin infusions. Rats were placed on a heating pad to maintain a constant rectal temperature of 37.0 ± 0.1°C.
After a bolus injection (Actrapid, 100 U/mL, Novo, Copenhagen, Denmark), insulin was infused at 8 mU/kg per minute. A 20% glucose solution in physiological saline was administered to maintain plasma glucose at 7 mM; the infusion rate was guided by glucose measurements in 10-μL blood samples obtained every 5 minutes until a steady state was achieved (~60 minutes) and then every 10 minutes. The mean infusion rate was calculated from values during the last 60 minute. At 0 and 120 minutes, 250-μl blood samples were obtained to measure insulin concentration. Less than 1 mL of blood was taken from each rat.
Tissues
After the clamp test, rats were decapitated, the hypothalamus was quickly removed and dissected on dry ice at the border of -0.8–-3.8 according to rat brain stereotactic coordinates [31]. Tissue punches centered on PVN were taken from the sections using an 18 gauge needle, and the adrenal medullae were dissected and snap frozen in liquid nitrogen until assay. The parametrial, mesenteric, retroperitoneal, and inguinal adipose tissues, one ovary and muscles of the hind limb (extensor digitorum longus, soleus, and tibialis anterior) were rapidly dissected by the same person and weighed.
Ovarian morphology
The ovary was removed, cleaned of adherent connective fat tissue, and fixed in 4% formaldehyde buffer for at least 24 hours. Thereafter the samples were dehydrated and imbedded in paraffin. The ovaries were partially longitudinally sectioned (4 μm, every tenth section mounted on the glass slide) and stained with hematoxylin and eosin. An experienced pathologist, blinded to grouping analysed the follicle population under microscope. There was no intention to quantify the number of growing or atretic follicles but rather to establish whether ovulation with corpora lutea formation had occurred within the given time frame. According to morphometric (stereological) and statistical principles there is no need to perform a statistical analysis in this situation.
Real-Time PCR Analysis
Total RNA from the PVN and adrenal medulla was extracted with RNeasy Mini kits (Qiagen, Hilden, Germany). First-strand cDNA was synthesized from 1 μg of total RNA with TaqMan reverse transcription reagents (Applied Biosystems., Foster City, CA). Each 100 μl RT-PCR reaction contained 1 μg of total, 1 × TaqMan RT buffer, 5 mM MgCl2, 2.5 mM random hexamers, 1 mM dNTP, 0.4 U/ml RNase inhibitor, and 1.25 U/ml Multiscribe RT (PE Applied Biosystems, Foster City, CA, USA). Reverse transcription was carried out in a PTC-200 PCR system (MJ Research., Boston, MA, USA) at 25°C for 10 min, 48°C for 30 min and 95°C for 5 min. PCR was performed with the ABI Prism 7700 sequence detection system and FAM-labeled probes specific for the α1a AR (NM 017191) and α2a AR (rCT57545) (PE Applied Biosystems, Stockholm, Sweden). Designed primers and a VIC-labeled probe for GAPDH (NM_031144) were included as internal standards. cDNA was amplified for one cycle of 50°C for 2 minutes and 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. mRNA levels were calculated with the standard curve method (User Bulletin 2, PE Applied Biosystems) and adjusted for GAPDH expression.
RIA and ELISA
Blood was collected into potassium EDTA tubes (ACTH), heparinized microtubes (CORT and insulin), or ordinary tubes (steroids) and centrifuged immediately at 4°C in a microcentrifuge. RIA kits were used to measure ACTH (DSL-2300, Diagnostic Systems Laboratories, Webster, TX), CORT (RSL125I Corticosterone RIA kit; MP Biomedicals, Costa Mesa, CA), testosterone (DSL-4100, Diagnostic Systems Laboratories), progesterone (DSL-3400, Diagnostic Systems Laboratories), and estradiol-17β (double-antibody estradiol procedure, DPC Scandinavia AB, Mölndal, Sweden). Insulin levels were determined by ELISA (Merkodia, Uppsala, Sweden).
Statistical Analysis
All statistical analyses were performed with SPSS 11.0 software. Body weight and hemodynamic data were analyzed by repeated-measures ANOVA. Weight gain, tissue weight, glucose infusion rate and glucose and insulin levels, stress and sex hormones, and mRNA levels were analyzed by two-tailed t test. Values are expressed as mean ± SE. P < 0.05 was considered significant.
Results
PCO Rats Develop Hypertension
At 5 and 7 weeks, MSAP was significantly higher in the PCO group than in controls (Figure 1). At 7 weeks, HR was also significantly higher in the PCO group (405 ± 13.6 versus 345 ± 10.0 beats/minute, P < 0.001).
Figure 1 MSAP 2, 3, 5 and 7 weeks after EV injection. At both time points, MSAP was significantly higher in PCO rats than controls. Values are mean ± SE. **P < 0.01 versus controls (t test).
Expression of AR mRNA in the Hypothalamic Paraventricular Nucleus and Adrenal Medulla
Expression of α1a AR mRNA in the PVN was significantly higher, and expression of α2a AR mRNA in the PVN and adrenal medulla significantly lower, in the PCO group than in controls (Figure 2).
Figure 2 Expression of α1a and α2a AR mRNA in the hypothalamic paraventricular nucleus (PVN) and adrenal medulla; values are normalized to GAPDH expression. Expression of α1a AR mRNA was upregulated in the PVN. Expression of α2a AR mRNA was downregulated in both the PVN and adrenal medulla. Values are mean ± SE. *P < 0.05, **P < 0.01 versus controls (t test.
ACTH and CORT Responses to Stress Are Higher in PCO Rats
To assess effects of EV exposure on the HPA axis, we measured the ACTH and CORT responses to novel-environment stress at 6 weeks (Table 2). At baseline, ACTH levels were similar in the two groups, while CORT levels tended higher in PCO rats. At 15 and 30 minutes, ACTH levels were higher in PCO rats than controls, but the difference was significant (P < 0.05) only when the two time points were analyzed together (not shown). At 30 and 60 minutes, CORT levels were significantly higher in the PCO group.
Table 2 Plasma ACTH and CORT Concentrations before and in Response to Novel Environment Stress Test 6 Weeks after EV Injection
Hormone Control group PCO group P
ACTH (pmol/L)
0 min 56.48 ± 5.97 52.21 ± 1.99 n.s.
15 min 79.83 ± 3.97 93.59 ± 10.71 n.s.
30 min 76.57 ± 7.04 97.45 ± 12.48 n.s.
60 min 77.85 ± 10.26 76.47 ± 7.84 n.s.
90 min 70.68 ± 3.31 72.23 ± 6.33 n.s.
120 min 68.34 ± 4.09 75.84 ± 4.38 n.s.
CORT (ng/mL)
0 min 251.32 ± 46.05 416.44 ± 93.29 n.s.
15 min 597.9 ± 41.26 756.44 ± 95.47 n.s
30 min 668.63 ± 46.18 1001.85 ± 132.73 0.05*
60 min 657.09 ± 65.02 980.83 ± 151.58 0.05*
90 min 744.51 ± 82.48 754.73 ± 129.68 n.s
120 min 537.63 ± 44.23 873.56 ± 158.80 n.s.
Values are mean ± SE. n.s. = non significant.
*t test.
PCO Rats Have Lower Testosterone and Higher Progesterone Levels
The PCO group had significantly lower testosterone and higher progesterone concentrations than the controls (Table 3). There were no differences in 17β-estradiol concentrations.
Table 3 Estradiol, Testosterone, and Progesterone Concentrations 7 Weeks after EV Injection
Hormone Control group PCO group P
17β-estradiol (pmol/L) 0.23 ± 0.02 0.22 ± 0.03 n.s.
Testosterone (nmol/L) 0.45 ± 0.04 0.27 ± 0.02 0.01
Progesterone (nmol/L) 12.78 ± 1.10 22.54 ± 3.38 0.05
Values are mean ± SE. n.s. = not significant.
PCO Rats Have Normal Insulin Sensitivity
The glucose infusion rate and plasma insulin and glucose concentrations before and during the clamp test are shown in Table 4. Insulin sensitivity was not significantly different in the PCO and control groups, as reflected by the glucose infusion rate.
Table 4 Glucose Infusion Rate and Plasma Insulin and Glucose Concentrations before and at Steady State during the Euglycemic Hyperinsulinemic Clamp Test
Measurement Oil group PCO group P
Glucose infusion rate (mg/kg/min)* 26.6 ± 1.0 26.3 ± 1.1 n.s.
Glucose level (mmol/L)
0 min 6.5 ± 0.2 6.3 ± 0.2 n.s.
120 min 6.8 ± 0.1 6.7 ± 0.1 n.s.
Insulin level (mU/mL)
0 min 85 ± 17 69 ± 10 n.s.
120 min 280 ± 38 274 ± 66 n.s.
Values are mean ± SE. n.s. = not significant (t test).
*At steady state (60–120 min).
Ovarian morphology – day 60
The ovaries in the control group exhibited a typically normal appearance with follicles and corpora lutea in different stages of development and regression. The ovaries in the PCO group displayed typical PCO-like changes [17], and both the number and size of the corpora lutea in these groups were decreased compared with the control group. Typically, no young corpora lutea were present in the PCO groups. The dominant structures were atretic follicles, regressing old corpora lutea, and a few growing "healthy" follicles in the primary, secondary, and tertiary stages (Figures 3).
Figure 3 PCO group control group 60 days after EV injection. Survey view showing atretic follicles (1), regressing old corpora lutea (2), growing "healthy" follicles (3), and atretic secondary follicle with detachment of the oocyte from the cumulus mass of pycnotic granulosa cells (4) ×4 obj.
PCO Rats Gain Less Total Body Weight
From weeks 3 to 10, PCO rats weighed less than controls, but the difference was significant only between weeks 3 and 6 (Table 1). Overall, the PCO group gained significantly less weight than the controls.
Table 1 Total Body Weight during the Study Period
Body weight (g)
Week Control group PCO group P
2 188.9 ± 1.6 192.1 ± 0.9 n.s.
3 205.7 ± 1.7 195.8 ± 1.1 0.001*
4 212.6 ± 1.6 203.6 ± 2.6 0.01*
5 214.9 ± 1.8 208.1 ± 2.6 0.05*
6 221.3 ± 2.3 213.0 ± 2.5 0.05*
7 220.8 ± 2.5 215.5 ± 2.8 n.s.
8 221.3 ± 2.3 216.4 ± 2.5 n.s.
9 225.0 ± 2.5 218.6 ± 2.6 n.s.
10 232.9 ± 3.3 228.5 ± 2.8 n.s.
Weight gain (g/week) 4.7 ± 0.2 3.9 ± 0.3 0.05†
Values are mean ± SE. n.s. = not significant.
*Repeated-measures ANOVA.
†t test.
No Differences in Intraabdominal Fat Depots or Hindlimb Skeletal Muscles
The inguinal fat depot (representing subcutaneous fat), was significantly heavier in the PCO group than in the controls (4.36 ± 0.2 and 3.36 ± 0.2 g/kg body weight, P < 0.01). No differences were observed in the weights of the mesenteric, parametrial, and retroperitoneal fat depots (representing intraabdominal fat) or the hind limb skeletal muscles (not shown).
Discussion
This study shows that rats with EV-induced PCO developed hypertension, consistent with increased sympathetic activity, as in human PCOS. They also had increased expression of α1a AR mRNA in PVN and decreased expression of α2a AR mRNA in PVN and adrenal medulla, consistent with elevated blood pressure and increased sympathetic activity, together with hyperactivity of the HPA and SAM axes. Rats with EV-induced PCO did not have reduced insulin sensitivity or become obese or hyperandrogenic.
Sympathetic Activity and the SAM Axis
Women with PCOS have a high incidence of hypertension [3,4] and are at increased risk of CVD [32]. The sympathetic nervous system might play a crucial role in the development of hypertension and CVD. As an example, hypertension is common in patients with polycystic kidney disease (PKD) and a recent study by Klein et al. demonstrates that their muscle sympathetic nerve activity is increased regardless of renal function [11]. This support the hypothesis that sympathetic hyperactivity also may contribute to the pathogenesis of hypertension in PCOS. In addition, overactivity of the SAM axis increases blood pressure [9]. Since sympathetic activity and hypertension are strongly correlated, we measured MSAP and found that it was increased at 5 and 7 weeks after EV injection. This finding may have implications for the mechanisms underlying the increased risk of developing hypertension, including the higher incidence of CVD later in life, in women with PCOS.
In the PCO group, α1a AR mRNA expression was increased in the PVN, and α2a AR expression was reduced both in the PVN and adrenal medulla, strongly suggesting increased sympathetic activity. In spontaneously hypertensive rats, increased central α1a AR expression and decreased α2a AR expression correlated with elevated blood pressure and increased sympathetic activity [15]. In another study, Peng et al. showed that the α2a AR has a sympathoinhibitory role in the brain [33]. Furthermore, α2 AR subtypes have been demonstrated to control the release of catecholamines from sympathetic nerves and from adrenal medulla [34]. Since the hypothalamus participates in central regulation of MSAP and adrenal medulla is involved in the regulation of SAM-axis, these findings support the conclusion that EV-induced PCO is associated with increased sympathetic activity.
The PCO rats gained significantly less weight than controls. The findings of enhanced adrenal glucocorticoid production following EV-injection in the present study might explain the lack of weight gain in PCO rats via lipolytic actions. Activation of the α1 and α2 ARs in the hypothalamic paraventricular nucleus inhibits food intake [35], suggesting that the dysregulated central expression of α1 and α2 ARs in the PCO group was responsible for the difference in weight gain. Food intake was not measured. It is important to keep in mind that PCOS in women is not always associated with obesity [36]. However, increased sympathetic nervous activity has been associated with increased metabolic activity, fat consumption, and decreased body weight, but not with reduced food intake [37].
In a study of cardiovascular risk factors, young women with PCOS had a higher prevalence of hypertension than population controls [5] and a compromised cardiovascular risk profile, even after adjustment for their higher BMI. In a recent study, low birth size and final height predicted high sympathetic nerve activity in adulthood [38] and this was positively correlated to high MSAP. Interestingly, low birth weight has been pointed out as an increased risk of developing PCOS [39]. Thus, the increased CVD risk associated with PCOS cannot be explained by obesity alone. These findings suggest that the greater risk reflects increased sympathetic nerve activity, which in turn results in hypertension.
Signs of hyperactivation in EV-induced PCO include downregulation of β2 AR in theca-interstitial cells [40]. Reduced expression of the β2 AR also enhances progesterone secretion by cystic ovaries in response to isoproterenol in rats with EV-induced PCO [19]. In the current study, the PCO group had significantly higher progesterone concentrations than the controls. In addition, high levels of P and low levels of T suggest a metabolic shift towards corticosterone synthesis and secretion and a hyper-adrenal state.
All these changes seemed to be independent of 17β-estradiol, since they occurred without any changes in its concentrations [41]. Consistent with this finding, 17β-estradiol concentrations were unaltered in the PCO group.
Interactions Between the SAM and HPA Axes
The novel-environment stress test also provided evidence for increased activity of the HPA axis in EV-induced PCO. The PCO group had higher CORT levels at 30 and 60 minutes than controls. Corticotropin-releasing homone (CRH) and locus ceruleus – NE (sympathetic nervous system) have been pointed out as important components of the stress system as well as their peripheral effectors, the HPA and SAM axes [42]. The CRH and locus ceruleus – NE systems directly modulate the HPA and SAM axes in order to maintain homeostasis [42]. Rats with EV-induced PCO have higher CRH levels in the median eminence than controls [43]. In the present study, the development of hypertension, increased expression of α1a AR and decreased expression of α2a-AR mRNA, and increased CORT responses to stress indicate a close interaction between the HPA and SAM axes, as well as increases in their activity. The interesting finding that the CORT response to stress was higher whereas the ACTH response was lower then expected can be explained by direct stimulation of adrenal glucocorticoids with less pituitary ACTH action [44].
Insulin Resistance, Hyperandrogenism, and Obesity
According to one theory of the pathogenesis of PCOS, a defect in insulin action leads to hyperinsulinemia and insulin resistance [45]. The insulin clamp tests performed 10 to 11 weeks after EV injection, when typical PCO are fully developed [17], showed no insulin resistance in the PCO group.
Why didn't our rats with EV-induced PCO develop insulin resistance? One possibility is that the measurements was done prior to the onset of insulin resistance. Another possibility is absence of intraabdominal obesity, thought to be a contributing factor [1]. The major abnormality in insulin action in PCOS is believed to be a post-receptor defect in the insulin signaling cascade in skeletal muscle, which might be due to interaction with testosterone [1]. In humans, fasting insulin levels correlate positively with androgen levels, and some studies have shown that hyperandrogenism causes insulin resistance in humans [46] and in rats [30]. Plasma testosterone levels were low in the PCO group, consistent with the lack of insulin resistance or hyperinsulinemia. As mentioned before, low levels of T suggest a metabolic shift towards corticosterone synthesis and secretion and a hyper-adrenal state.
It is also important to remember that only about 50% of women with PCOS have insulin resistance and no more than about 40% are obese [47]. Moreover, there are pitfalls in assessing insulin resistance in PCOS, including the lack of consensus on what defines PCOS and "normal" insulin sensitivity, ethnic and genetic variability, confounding factors such as obesity, stress, and aging, and concerns about whether simplified models of insulin sensitivity have the precision to predict treatment needs, responses, and morbidity [48].
Perspectives
Different models of PCO are inevitable, and the model to use depends on the goals of the study. We demonstrated that rats with EV-induced PCO develop hypertension, with increased α1a AR and decreased α2a AR mRNA expression, consistent with increased sympathetic activity. We also found evidence of increased HPA-axis activity.
These findings may have implications for mechanisms underlying hypertension in women with PCOS, since essential hypertension is associated with sympathetic hyperactivity, which is itself known to increase CVD risk. We also demonstrated that rats with EV-induced PCO do not have reduced insulin sensitivity and do not develop obesity or hyperandrogenism, which might be later signs of sympathetic hyperactivity. Hyperglycemia of diabetic PCOS patients has been found to be significantly positively correlated with adrenal hypersecretion of cortisol, dehydroepiandrostenedione (DHEA) and dehydroepiandrostenedione sulfate (DHEAS) [49]. They suggested that enhanced adrenocortical (HPA) activity may be important factor underlying the development of type 2 diabetes in women with PCOS. Since we found exaggerated activity in both HPA and SAM axes in the present EV-induced PCO model, it might mimic an initial stage of PCOS.
These findings might provide important leads for future studies with the EV-induced PCO model.
Authors' contributions
ES-V participated in the design of the study, carried out part of the animal preparation, performed RT-PCR, performed the statistical analysis and drafted the manuscript. KP participated in the design of the study and carried out the animal preparation. B-M carried out the animal preparation. AH participated in the design of the study and in writing the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This study was supported by grants from Wilhelm and Martina Lundgrens's Science Fund, Hjalmar Svensson Foundation, Tore Nilsons Stiftelse, Magnus Bergwalls Stiftelse, the Novo Nordisk Foundation, The Göteborg Medical Society, the Swedish Medical Research Council (Project No. 12206, 2004-6399 and -6827), and the Swedish Heart Lung Foundation.
==== Refs
Tsilchorozidou T Overton C Conway GS The pathophysiology of polycystic ovary syndrome Clin Endocrinol (Oxf) 2004 60 1 17 14678281 10.1046/j.1365-2265.2003.01842.x
Kannel WB The Framingham Study: historical insight on the impact of cardiovascular risk factors in men versus women J Gend Specif Med 2002 5 27 37 11974672
Dahlgren E Janson PO Johansson S Lapidus L Lindstedt G Tengborn L Hemostatic and metabolic variables in women with polycystic ovary syndrome Fertil Steril 1994 61 455 460 8137966
Dahlgren E Janson PO Johansson S Lapidus L Oden A Polycystic ovary syndrome and risk for myocardial infarction. Evaluated from a risk factor model based on a prospective population study of women Acta Obstet Gynecol Scand 1992 71 599 604 1336918
Vrbikova J Cifkova R Jirkovska A Lanska V Platilova H Zamrazil V Starka L Cardiovascular risk factors in young Czech females with polycystic ovary syndrome Hum Reprod 2003 18 980 984 12721172 10.1093/humrep/deg218
Lobo RA Granger LR Paul WL Goebelsmann U Mishell DRJ Psychological stress and increases in urinary norepinephrine metabolites, platelet serotonin, and adrenal androgens in women with polycystic ovary syndrome Am J Obstet Gynecol 1983 145 496 503 6824043
Zimmermann S Phillips RA Dunaif A Finegood DT Wilkenfeld C Ardeljan M Gorlin R Krakoff LR Polycystic ovary syndrome: lack of hypertension despite profound insulin resistance J Clin Endocrinol Metab 1992 75 508 513 1639952 10.1210/jc.75.2.508
Cattrall FR Healy DL Long-term metabolic, cardiovascular and neoplastic risks with polycystic ovary syndrome Best Pract Res Clin Obstet Gynaecol 2004 18 803 812 15380148 10.1016/j.bpobgyn.2004.05.005
Reaven GM Lithell H Landsberg L Hypertension and associated metabolic abnormalities--the role of insulin resistance and the sympathoadrenal system N Engl J Med 1996 334 374 381 8538710 10.1056/NEJM199602083340607
Supiano MA Hogikyan RV Sidani MA Galecki AT Krueger JL Sympathetic nervous system activity and alpha-adrenergic responsiveness in older hypertensive humans Am J Physiol 1999 276 E519 28 10070019
Klein IH Ligtenberg G Oey PL Koomans HA Blankestijn PJ Sympathetic activity is increased in polycystic kidney disease and is associated with hypertension J Am Soc Nephrol 2001 12 2427 2433 11675419
Semenova II Adrenergic innervation of the ovaries in Stein-Leventhal syndrome Vestn Akad Med Nauk SSSR (Abstract in english) 1969 24 58 62
Heider U Pedal I Spanel-Borowski K Increase in nerve fibers and loss of mast cells in polycystic and postmenopausal ovaries Fertil Steril 2001 75 1141 1147 11384640 10.1016/S0015-0282(01)01805-2
Civantos Calzada B Aleixandre de Artinano A Alpha-adrenoceptor subtypes Pharmacol Res 2001 44 195 208 11529686 10.1006/phrs.2001.0857
Reja V Goodchild AK Pilowsky PM Catecholamine-related gene expression correlates with blood pressures in SHR Hypertension 2002 40 342 347 12215477 10.1161/01.HYP.0000027684.06638.63
Szukiewicz D Uilenbroek JTJ Polycystic ovary syndrome - searching for an animal model J Med 1998 29 259 275 10503163
Brawer JR Munoz M Farookhi R Development of the polycystic ovarian condition (PCO) in the estradiol valerate-treated rat Biology of Reproduction 1986 35 647 655 3098314
Luza SM Lizama L Burgos RA Lara HE Hypothalamic changes in norepinephrine release in rats with estradiol valerate-induced polycystic ovaries Biol Reprod 1995 52 398 404 7711208
Barria A Leyton V Ojeda SR Lara HE Ovarian steroidal response to gonadotropins and beta-adrenergic stimulation is enhanced in polycystic ovary syndrome: role of sympathetic innervation Endocrinology 1993 133 2696 2703 8243293 10.1210/en.133.6.2696
Lara HE Ferruz JL Luza S Bustamante DA Borges Y Ojeda SR Activation of ovarian sympathetic nerves in polycystic ovary syndrome Endocrinology 1993 133 2690 2695 7902268 10.1210/en.133.6.2690
Lara HE Dissen GA Leyton V Paredes A Fuenzalida H Fiedler JL Ojeda SR An increased intraovarian synthesis of nerve growth factor and its low affinity receptor is a principal component of steroid-induced polycystic ovary in the rat Endocrinology 2000 141 1059 1072 10698182 10.1210/en.141.3.1059
Stener-Victorin E Lundeberg T Waldenstrom U Manni L Aloe L Gunnarsson S Janson PO Effects of electro-acupuncture on nerve growth factor and ovarian morphology in rats with experimentally induced polycystic ovaries Biol Reprod 2000 63 1497 1503 11058557
Stener-Victorin E Lundeberg T Cajander S Aloe L Manni L Waldenstrom U Janson PO Steroid-induced polycystic ovaries in rats: effect of electro-acupuncture on concentrations of endothelin-1 and nerve growth factor (NGF), and expression of NGF mRNA in the ovaries, the adrenal glands, and the central nervous system Reprod Biol Endocrinol 2003 1 33 12725645 10.1186/1477-7827-1-33
Manni L Holmang A Lundeberg T Aloe L Stener-Victorin E Ovarian expression of alpha (1)- and beta (2)-adrenoceptors and p75 neurotrophin receptors in rats with steroid-induced polycystic ovaries Auton Neurosci 2005 118 79 87 15795180 10.1016/j.autneu.2005.01.004
Manni L Lundeberg T Holmang A Aloe L Stener-Victorin E Effect of electro-acupuncture on ovarian expression of alpha (1)- and beta (2)-adrenoceptors, and p75 neurotrophin receptors in rats with steroid-induced polycystic ovaries Reprod Biol Endocrinol 2005 3 21 15941472 10.1186/1477-7827-3-21
Bornstein SR Ehrhart-Bornstein M Basic and clinical aspects of intraadrenal regulation of steroidogenesis Z Rheumatol 2000 59 Suppl 2 II/12 7
McEwen BS Stress, brain, and behaviour: life-long effects upon health and disease. 1997 Philadelphia, Lippincott-Raven
Folkow B Mental stress and its importance for cardiovascular disorders; physiological aspects, "from-mice-to-man" Scand Cardiovasc J 2001 35 163 172 11515688 10.1080/140174301750305036
Lahti RA Barsuhn C The effect of minor tranquilizers on stress-induced increases in rat plasma corticosteroids Psychopharmacologia 1974 35 215 220 4829322 10.1007/BF00437750
Holmäng A Svedberg J Jennische E Björntorp P Effects of testosterone on muscle insulin sensitivity and morphology in female rats Am J Physiol 1990 259 (Endocinol Metab 22) E555 E560 2221057
Paxinos G The rat brain in stereotaxic coordinates 1982 Sydney, Academic Press
Tíras MB Yalcin R Noyan V Maral I Yìldìrìm M Dörtlemez O Daya S Alterations in cardiac outflow parameters in patients with polycystic ovarian syndrome Human Reproduction 1999 14 1949 1952 10438405 10.1093/humrep/14.8.1949
Peng N Chambless BD Oparil S Wyss JM Alpha2A-adrenergic receptors mediate sympathoinhibitory responses to atrial natriuretic peptide in the mouse anterior hypothalamic nucleus Hypertension 2003 41 571 575 12623961 10.1161/01.HYP.0000056998.83031.22
Brede M Nagy G Philipp M Sorensen JB Lohse MJ Hein L Differential control of adrenal and sympathetic catecholamine release by alpha 2-adrenoceptor subtypes Mol Endocrinol 2003 17 1640 1646 12764077 10.1210/me.2003-0035
Wellman PJ Norepinephrine and the control of food intake Nutrition 2000 16 837 842 11054588 10.1016/S0899-9007(00)00415-9
Hart R Hickey M Franks S Definitions, prevalence and symptoms of polycystic ovaries and polycystic ovary syndrome Best Pract Res Clin Obstet Gynaecol 2004 18 671 683 15380140 10.1016/j.bpobgyn.2004.05.001
Racotta R Ramirez-Altamirano L Velasco-Delgado E Metabolic effects of chronic infusions of epinephrine and norepinephrine in rats Am J Physiol 1986 250 E518 22 3518485
Boguszewski MC Johannsson G Fortes LC Sverrisdottir YB Low birth size and final height predict high sympathetic nerve activity in adulthood J Hypertens 2004 22 1157 1163 15167451 10.1097/00004872-200406000-00017
Ibanez L Valls C Potau N Marcos MV de Zegher F Polycystic ovary syndrome after precocious pubarche: ontogeny of the low-birthweight effect Clin Endocrinol (Oxf) 2001 55 667 672 11894979 10.1046/j.1365-2265.2001.01399.x
Lara HE Dorfman M Venegas M Luza SM Luna SL Mayerhofer A Guimaraes MA Rosa E Silva AAM Ramírez VD Changes in sympathetic nerve activity of the mammalian ovary during a normal estrous cycle and in polycystic ovary syndrome: Studies in norepinephrine release Microsc Res Tech 2002 59 495 502 12467025 10.1002/jemt.10229
Dissen GA Lara HE Leyton V Paredes A Hill DF Costa ME Martinez-Serrano A Ojeda SR Intraovarian excess of nerve growth factor increases androgen secretion and disrupts estrous cyclicity in the rat Endocrinology 2000 141 1073 1082 10698183 10.1210/en.141.3.1073
Chrousos GP The role of stress and the hypothalamic-pituitary-adrenal axis in the pathogenesis of the metabolic syndrome: neuro-endocrine and target tissue-related causes Int J Obes Relat Metab Disord 2000 24 Suppl 2 S50 5 10997609
Stener-Victorin E Lundeberg T Waldenstrom U Bileviciute-Ljungar I Janson PO Effects of electro-acupuncture on corticotropin-releasing factor in rats with experimentally-induced polycystic ovaries Neuropeptides 2001 35 227 231 12030806 10.1054/npep.2002.0878
Turnbull AV Smith GW Lee S Vale WW Lee KF Rivier C CRF type I receptor-deficient mice exhibit a pronounced pituitary-adrenal response to local inflammation Endocrinology 1999 140 1013 1017 9927337 10.1210/en.140.2.1013
Dunaif A Insulin resistance and the polycystic ovary syndrome: mechanism and implications for pathogenesis Endocr Rev 1997 18 774 800 9408743 10.1210/er.18.6.774
Polderman KH Gooren LJ Asscheman H Bakker A Heine RJ Induction of insulin resistance by androgens and estrogens J Clin Endocrinol Metab 1994 79 265 271 8027240 10.1210/jc.79.1.265
Hemmings R Farookhi R Brawer JR Pituitary and ovarian responses to LHRH in a rat with polycystic ovaries. Biol Reprod 1983 29 239 248 6351938
Legro RS Castracane VD Kauffman RP Detecting insulin resistance in polycystic ovary syndrome: purposes and pitfalls Obstet Gynecol Surv 2004 59 141 154 14752302 10.1097/01.OGX.0000109523.25076.E2
Buffington CK Givens JR Kitabchi AE Enhanced adrenocortical activity as a contributing factor to diabetes in hyperandrogenic women Metabolism 1994 43 584 590 8177046 10.1016/0026-0495(94)90200-3
|
16146570
|
PMC1236959
|
CC BY
|
2021-01-04 16:37:13
|
no
|
Reprod Biol Endocrinol. 2005 Sep 7; 3:44
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-44
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-1011615014210.1186/1465-9921-6-101ResearchIncomplete quality of life data in lung transplant research: comparing cross sectional, repeated measures ANOVA, and multi-level analysis Vermeulen Karin M [email protected] Wendy J [email protected] Mark M [email protected] der Bij Wim [email protected]ëter Gerard H [email protected] Elisabeth M [email protected] Office for Medical Technology Assessment, University Medical Center Groningen, the Netherlands2 Department of Pulmonary Diseases, University Medical Center Groningen, the Netherlands2005 8 9 2005 6 1 101 101 6 6 2005 8 9 2005 Copyright © 2005 Vermeulen et al; licensee BioMed Central Ltd.2005Vermeulen et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In longitudinal studies on Health Related Quality of Life (HRQL) it frequently occurs that patients have one or more missing forms, which may cause bias, and reduce the sample size. Aims of the present study were to address the problem of missing data in the field of lung transplantation (LgTX) and HRQL, to compare results obtained with different methods of analysis, and to show the value of each type of statistical method used to summarize data.
Methods
Results from cross-sectional analysis, repeated measures on complete cases (ANOVA), and a multi-level analysis were compared. The scores on the dimension 'energy' of the Nottingham Health Profile (NHP) after transplantation were used to illustrate the differences between methods.
Results
Compared to repeated measures ANOVA, the cross-sectional and multi-level analysis included more patients, and allowed for a longer period of follow-up. In contrast to the cross sectional analyses, in the complete case analysis, and the multi-level analysis, the correlation between different time points was taken into account. Patterns over time of the three methods were comparable. In general, results from repeated measures ANOVA showed the most favorable energy scores, and results from the multi-level analysis the least favorable. Due to the separate subgroups per time point in the cross-sectional analysis, and the relatively small number of patients in the repeated measures ANOVA, inclusion of predictors was only possible in the multi-level analysis.
Conclusion
Results obtained with the various methods of analysis differed, indicating some reduction of bias took place. Multi-level analysis is a useful approach to study changes over time in a data set where missing data, to reduce bias, make efficient use of available data, and to include predictors, in studies concerning the effects of LgTX on HRQL.
==== Body
Background
Lung transplantation has become an accepted treatment option for appropriately selected patients with end-stage lung disease. Besides clinical outcome measures such as survival, Health Related Quality of Life (HRQL) has become an increasingly important endpoint in studies regarding the effectiveness of lung transplantation. Studies in which HRQL was included as an outcome measure generally report improvements across many domains of HRQL after lung transplantation [1-7]. The aim of the present study was twofold. First, to address the problem of missing data in the field of HRQL and lung transplantation, and secondly to compare results from different methods of analysis in a data-set where missing data occur in order to show the value of each type of statistical method used to summarize data.
In many studies, HRQL is assessed longitudinally by means of questionnaires, which are presented to the patients at several predetermined time points in order to evaluate changes over time. Unfortunately, missing assessments are frequently encountered and can be caused by a variety of factors. A possible cause for missingness of data can be poor data management, for example when a research employee 'forgets' to hand out a questionnaire to a patient (logistic reason). When the burden on the patient is too high, for example due to a large number of questionnaires, or question difficulty this can also be a reason for dropping out (methodological reason). In the examples mentioned above, it is unlikely that the reason for missing is related to the patients health status. Other reasons for missingness are health problems or side effects of therapy due to which patients are temporarily unable to complete the questionnaire. An other example of a reason for missingness is the death of a patient. In these cases the missingness is reflects the patients health status. Missingness of data due to logistic or methodological reasons, can be prevented. Consequently, in this case the best way to handle the missing data problem is prevention. Missingness of data caused by patient related factors is more unpreventable.
The missingness of data has two major undesirable effects. First, if missingness is correlated with the outcome one is interested in, ignoring it will bias the results. For example, when missingness is caused by serious health problems, patients with missing assessments will differ on health status from patients who have completed all forms. Consequently, results of patients with complete forms cannot be generalized to the entire population: conclusions are only applicable to the group of 'completers' who have better health status than other patients in the population. A second complication associated with missing is the loss of efficiency. Because most statistical software packages automatically drop subjects with one or more missing assessments, it causes loss of efficiency due to reduced sample sizes in the analysis. Few researchers in the field of lung transplantation have acknowledged the problem of missing HRQL data [1,8]. However, no consensus could be found in the LgTX literature about the appropriate statistical method for dealing with it. Moreover, the choice for a particular statistical method strongly depends on the study objective under investigation.
Irrespective of the reasons for, and the magnitude of the missing data problem, two methods of analyzing data are commonly performed in studies regarding the effects of lung transplantation on HRQL. First, especially in the earlier years when the number of transplanted patients was still relatively small, cross sectional analyses were usually performed. In this type of analyses, at two or more time points, all available data at that specific point are analyzed. These kind of analyses result in conclusions for different groups of patients at the various time points. Thus, in cross sectional analyses, the longitudinal character of the data set is ignored. When the research aim is to assess changes over time, cross sectional analyses are not suitable. However, this method is acceptable for descriptive purposes and has the advantage that it makes efficient use of the available data at each time point.
When studying changes over time, longitudinal analyses are preferred [9]. However, when repeated measures techniques are used, most commonly used software packages exclude the entire patient with one or more missing assessments from the analysis. Consequently, only patients who have completed all questionnaires (complete cases) are included. When the research is aimed at describing a specific subgroup of, for example surviving patients, complete case analysis may be appropriate. In addition, complete case, but also cross sectional methods can be used in case missing forms are completely randomly distributed, and the reduced data represent a randomly drawn sub-sample of the original data-set [10]. However, when patients with incomplete data differ from patients with complete data, and missingness can be predicted from other observed variables, complete case analysis may not be valid. In that case, an alternative method of analysis has to be used to assess changes over time. In our study, the methods we will focus on are likelihood based, which provide estimates based on all available data. These methods have been applied in other fields of research to estimate complex models for data sets with missing observations. Examples of likelihood based methods are multilevel models. Multilevel methods are also called random effects, mixed, or hierarchical models.
Two advantages for using these models are that the dependency between measurements at successive time points is maintained, and that subjects with incomplete data are not excluded from the analysis. This means that, if a patient is missing one or more observations, the remaining available data from the other observations for that particular patient are used in the analysis [11]. When missing depends on the observed data, for example on previous HRQL outcome, the estimates provided by estimation procedures such as those of maximum likelihood used in the multi-level analysis, are unbiased [12]. Therefore, models like this are preferable because they incorporate all available information in the data and are less vulnerable to bias. This in contrast to an analysis confined to the complete cases [13]. Until recently, these modeling procedures were not available in most standard software packages used by the majority of clinical researchers. Some frequently used software programs of today offer this option. However, to our knowledge in the field of lung transplantation and HRQL no studies have been published comparing results obtained with one of these programs to results obtained with the commonly used software packages.
In the present study, we compared results obtained with three different methods of analysis: cross-sectional analysis, repeated measures ANOVA on complete cases, and multi-level analysis. We used the dimension 'energy' of the Nottingham Health Profile (NHP) with a maximum follow-up of almost 10 years after lung transplantation. This dataset was suitable for the present purpose, because it covered a long period of follow-up, it included different types of missing data, and depending on the period of follow-up, there was a rather substantial amount of missing assessments.
Patients and Methods
Patient population and HRQL measure
After lung transplantation patients were asked to fill in HRQL-questionnaires at one, four, seven, and subsequently every six months. The questionnaires consisted of a combination of generic, disease-specific, and domain-specific health status measures, including the Nottingham Health Profile (NHP) [14].
The NHP is a generic measure of health status designed to measure perceived health on six specific domains of life. For illustrative purposes, one outcome measure is considered in this study: the dimension energy of the NHP. NHP-energy scores are shown in the present study because they depict an important dimension of HRQL in LgTX patients. Possible scores range from 0 to 100. When interpreting the results, please note that higher scores represent lower experienced energy levels. Between November 1990 and September 2003, 239 patients filled in one or more HRQL questionnaires after transplantation, and were analyzed in the present study. The maximum period of follow-up was 109 months after transplantation.
Data set
The numbers of completed and missing questionnaires were registered at all time points. For convenience of comparison, numbers of completed and missing questionnaires at 1, 13, 37, 73, and 109 months are shown in table 1. In our data set, three reasons for missingness can be distinguished. First incidental dropout, which means that a person has one or more missing forms in-between a series of completed forms. Secondly, dropout due to censoring, which includes patients that could not complete the questionnaire because their time since transplantation was shorter than that specific period of follow-up. For example, 20 patients did not complete the 13-month questionnaire, because they were transplanted less than 13 months before the moment we analyzed the data set. The last column shows the number of patients that died before a specific time point. For example 48 patients did not complete a questionnaire at 13 months after transplantation, because they had died within 13 months after transplantation.
Table 1 Numbers of completed and missing questionnaires
Time after transplantation Completed questionnaires Missing questionnaires
months number Incidental number Censored number Deceased number
1 133 106 - -
.
13 115 56 20 48
.
37 74 28 72 65
.
73 45 15 103 76
.
109 14 8 127 90
Patients: n = 239
Methods of analyses
By means of a logistic regression model [15] we tested which type of missing occurred in our data. The analysis suggested that the probability a questionnaire was missing was dependent on previous HRQL measurements. Consequently, the use of a likelihood based method was appropriate. For further reading on the subject of testing for different types of missingness we refer to Hedeker and Gibbons [16].
Cross-sectional analyses were performed using descriptive statistics, including mean scores and standard errors, on all available cases at each time point. For these analyses, the SPSS program was used (SPSS 11.0; SPSS, Inc; Chicago). Repeated measures on complete cases were also performed in SPSS, using repeated measures analysis of variance including only those patients who had complete follow-up until 73 months after transplantation.
For the multi-level analysis the MLwiN software package for fitting multi-level models was used (version 1.10; Centre for Multilevel Modelling, Institute of Education, University of London, UK). In an additional analysis, the same results were obtained by using the mixed models option in SPSS (SPSS 12.0; SPSS, Inc; Chicago). For further reading on different software packages see Singer and Willet [17]. An SPSS syntax file is available from the authors on request.
In the modeling process, variables were included in the model sequentially. After each step, the goodness of fit was determined by the difference in deviance (-2*loglikelihood) between the present and the previous model, and the number of additional included variables compared to the previous model. We used the unconditional means model [17] as a starting point. Instead of describing change in the outcome over time, this model simply describes and partitions the outcome variation across patients [17]. Subsequently, time was added to the model (unconditional growth model [17]) based on the observed pattern of results of the cross sectional analysis.
Finally, a number of confounding variables was identified because of their expected influence on experienced energy after transplantation, based on the available literature. Demographic data like gender, age, and diagnosis could be of influence [18,19]. Diagnosis was categorized into 4 categories: 'alpha 1 antitrypsin deficiency', 'cystic fibrosis', 'emphysema' and 'other'. Furthermore, time spent on the waiting list, and the presence or absence of Bronchiolitis Obliterans Syndrome (BOS) which is characterized by a slowly progressive decline in lung function and is also associated with increased morbidity [2,20] were possible predictors. The severety of BOS was not taken into account. Presence of BOS was assessed according to the criteria of the International Society for Heart and Lung Transplantation [21], either on functional data, if there was sustained and significant decline in the forced expiratory volume in 1 second to less than 80% of a previously established baseline value, or on the presence of obliterative bronchiolitis in biopsies, even if the lung function had not deteriorated [2].
Finally, the calendar year in which a patient was transplanted was a possible predictor of NHP-energy scores after LgTX. After the 'unconditional growth model'[17] was built, an advanced model was fitted based on these possible predictors.
Results
Indication of the missing data problem and demographic characteristics
Table 1 shows the magnitude of the missing data problem. One month after transplantation 133 patients completed a HRQL questionnaire. At the end of the follow-up period, approximately 9 years after transplantation (109 months), 14 patients completed a questionnaire, 8 patients had an 'incidental-missing', 127 did not complete the questionnaire because their time since transplantation was shorter than 109 months (censoring), and 90 patients had died.
In table 2, the demographic characteristics of the patients in the study population are depicted.
Table 2 Characteristics of transplanted patients (n = 239)
Gender, Male n(%) 128 (53.6)
Age years, mean (range) 44 (20–64)
Diagnosis, n (%)
alpha1 antitrypsin deficiency 59 (24.7)
Emphysema 41 (17.2)
Cystic fibrosis 48 (20.1)
Miscellaneous 91 (38.0)
Days on waiting list, mean (range) 465 (1–2207)
Patients with BOS, n (%) 67 (28.1)
Two hundred thirty nine patients were included. Mean age of this population was 44 years, and 53.6% were male. In our sample, the main diagnosis before lung transplantation was alpha 1 antitrypsin deficiency. Furthermore, 67 patients developed BOS at some time point after transplantation.
NHP-energy scores
Results of cross-sectional analyses (mean and standard error per time point) are depicted graphically in figure 1. At each time point the analysis is based on a different group of patients, and consequently no changes over time could be assessed. One month after transplantation, mean NHP-energy scores are approximately 25 (range: 0–100), whereas the reference value for the general population is below 15. Four months after transplantation, means scores are below 10 (range: 0–100), and after that mean scores are around 15 (ranges 0–100 and 0–63 at all time points till 103 months and 109 months respectively), and remain more or less stable and within the reference value at the different points in time (in the different subgroups). Towards the end of the follow-up period mean scores seem to fluctuate. However, number of patients in these subgroups are relatively small, and results should be carefully interpreted.
Figure 1 Results of cross sectional analysis
To maintain a reasonable sample to analyze in the repeated measures ANOVA on complete cases we used a follow-up period of 73 months. This allowed for the inclusion of 19 patients in the analysis (figure 2). One month after transplantation, mean NHP-energy scores were just below 20. Between four and approximately 40 months mean scores are between 5 and 10, and after that scores increase, indicating worse health. Changes over time appeared to be not significant in this group and over this period.
Figure 2 Results of repeated measures ANOVA on complete cases
Table 3 shows the three significant models, estimated with the multi-level analysis. The modeling procedure started with an unconditional means model, using of a constant term only. This constant has one fixed and two random parts. The fixed part can be interpreted as the mean score over all patients and time points (in this model approximately 19 points), whereas the random parts represent the variability within and between patients (not shown).
Table 3 Variables in various stages of the model
Explanatory variables Unconditional means model Estimate (SE) Unconditional growth model Estimate (SE) Final model Estimate (SE)
Fixed
Constant 19.40 (1.92) 18.16(2.42) 22.66 (3.05)
Time -5.32 (3.76) -10.71 (3.04)
Time square 4.54 (1.84) 5.69 (1.81)
Time third degree -0.80 (0.34) -0.92 (0.33)
Time forth degree 0.04 (0.02) 0.05 (0.02)
Age 0.56 (0.17)
BOS 23.73 (2.84)
Gender (male) -8.00 (3.58)
-2*loglikelihood (IGLS) 12935.69 12778.50 12698.65
All effects significant, except for time in the unconditional growth model
The unconditional means model was extended by including the time variable, and subsequently time square, time to the third degree, and time to the fourth degree, resulting in the unconditional growth model (figure 3). NHP energy scores that are estimated by the model can be compared to the results from cross-sectional and repeated measures ANOVA on complete cases.
Figure 3 Estimated NHP-energy scores (unconditional growth model)
After having estimated the changes over time, we added possible predictors to the model. First of all the presence of Bronchiolitis Obliterans Syndrome (BOS) was added. It was found that BOS had a statistically significant effect. Diagnosis did not contribute significantly to the model. Furthermore, neither time patients spent on the waiting list, nor calendar year of transplantation, nor the interaction between calendar year and time since transplantation contributed significantly. Age and gender however, provided a significant contribution to the model.
In figure 4 the predictions based on the estimates obtained from the final model are graphically displayed. The lines show mean NHP energy scores over time in transplanted males and females with and without BOS. Age was centered at 44 years (the mean age in our population) so that the lines correspond to 44-year-old subjects. With each year of age, estimated energy scores increased with 0.56 points (table 3), indicating that the experienced energy level declines when patients get older. After the development of BOS, the estimated energy scores increased with 23.73 points (table 3), and overall, male patients had an eight points lower energy score than females. Note that higher scores represent less perceived energy.
Figure 4 Estimated NHP energy scores (final model)
Comparison of the different methods
Figure 5 displays the differences between the results estimated with the three methods of analysis. Patterns over time were comparable. However, clear differences were found concerning the mean scores, the number of included patients, and the period of follow-up.
Figure 5 Comparison of available case, repeated measures ANOVA on complete cases, and multi-level analysis
Cross-sectional analysis of available cases showed mean scores that were more or less in-between the mean scores estimated with the other two methods. Furthermore, with this method, all patients were included, and results were analyzed until the maximum period of follow-up, 109 months after transplantation. However, no changes over time could be assessed.
Repeated measures ANOVA on complete cases showed the lowest scores compared to the other two methods, indicating better health. In this type of analysis, the smallest number of patients could be included, and results were analyzed until 73 months after transplantation, which was the shortest period of follow-up. Changes over time could be assessed.
Multilevel analysis showed higher predicted scores compared to the other two methods, indicating worse health. All patients and measurements were included in the analysis, and results were analyzed up to the maximum period of follow-up. Furthermore, changes over time could be assessed, and this method accounts for dependency between different measurements within a patient. In addition, predictors could be added to the model.
Discussion
Missing data is a common problem in HRQL research. However, only few studies assessing HRQL in lung transplantation patients [1,8] openly addressed the problems associated with missing data: possible bias and loss of efficiency. In the present study, we compared the results of three different methods in a data set where depending on the period of follow-up, there was a substantial proportion of patients that did not complete all questionnaires. Methods were: cross sectional analyses, repeated measures analysis ANOVA on complete cases, and multi-level analysis. The estimated NHP energy scores were used to illustrate differences in results. Analyses showed that in our dataset patients with missing data differed from patients who completed all questionnaires, which means that patients who completed all questionnaires were not representative for the entire population of transplanted patients. Results showed that mean scores on NHP-energy were less favorable when estimated with cross-sectional analysis compared to the repeated measures ANOVA on complete cases.
The unconditional growth model estimated in the multi-level analysis, showed the least favorable energy scores compared to the other two methods. Patterns over time were comparable in all three methods.
The finding that scores estimated with the multi-level method were higher and thus less favorable compared to the complete case, and especially the cross sectional results, may raise questions. This can be explained by the fact that in the multi-level analysis, contrary to the other two methods, patients who have a missing questionnaire at a certain time point are not excluded from the analysis. The model estimates the subjects trend across time on the basis of whatever data that subject has, augmented by the time trend that is estimated for the sample as a whole, and effects of all covariates in the model [16].
Thus, in the multi-level model, scores on previous time points are taken into account in the estimation procedure, whereas in the cross sectional analysis the means are solely based on the observed scores at that point in time. Patients who drop out due to their worse health most likely have less favorable scores on previous time points. Complete exclusion of these patients from the analysis (repeated measures ANOVA) will lead to a lower, more favorable estimation of mean scores compared to the situation were estimations are based on worsening previous scores (multi-level analysis).
In addition, the fact that mean predicted scores were less favorable with the multi-level method compared to the other two methods indicates a reduction of bias. Both cross sectional and longitudinal means are based on results from patients who had better health states. Therefore, in the repeated measures ANOVA on complete cases, the selection of surviving patients that are capable to complete each questionnaire could also explain the lower, more favorable scores.
We have demonstrated with this study that, when analyzing a data set in which missing assessments occur, differences between results obtained with the various methods of analysis do exist. Depending on the research aim each of the three methods has its merits.
Cross sectional analysis are appropriate when health states at separate time points are under study rather than changes over time. When changes over time are relevant longitudinal analysis are preferred [9]. However, exclusion of patients with one or more missing data, which occurs when repeated measures analysis is used, results in conclusions based on, and only applicable to the particular subgroup of patients. This approach, however, may be legitimate or even necessary in order to confine the analysis on a specific subgroup, like surviving patients, who were able to complete all questionnaires. When the focus is on changes over time, multi-level analysis provides a good alternative to repeated measures ANOVA because with this method all available data are used in the analysis. This method gives unbiased estimates for most types of missing data, and, like repeated measures ANOVA, takes into account the dependency between different measurements within a patient. Finally, multi-level analysis proved to be very useful to analyze longitudinal changes, to include all available assessments, to reduce bias, and to include predictors.
When interpreting results from longitudinal studies on HRQL after lung transplantation, it is wise to be informed about the amount and type of missing data, the type of analysis which was performed, and the subgroup of patients the analysis was confined to. All these aspects determine the population and the circumstances, for example surviving patients without major complications, for which the results and conclusions described in the study are valid.
Because in the multi-level analysis all available assessments are used in the analysis, no reduction of power takes place. A result of this more efficient use of data is that predictors can be included in the model. This is in contrast to the repeated measures ANOVA, where due to the selection of patients with complete data, the power is reduced dramatically, and inclusion of predictors is impossible.
In conclusion, when longitudinal changes are under study, and missing data occur in the data set, Multilevel analysis is preferred to cross sectional and complete case analysis.
Declaration of competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KV was involved in acquisition of the HRQL data, carried out the statistical analysis and interpretation of the data, and drafted and revised the manuscript.
WP contributed to the conception and design of the study, supported carrying out the statistical analysis, supervised the analysis and critically revised the manuscript.
MS intellectually supported the research, and critically revised the manuscript.
WB was involved in acquisition and interpretation of the clinical data and critically revised the manuscript.
GK supervised the research and analysis and critically revised the manuscript
ETV supervised acquisition of the HRQL data, contributed to conception and design of the study, and critically revised the manuscript.
All authors read and approved the final manuscript.
==== Refs
TenVergert EM Essink Bot ML Geertsma A van Enckevort PJ de Boer WJ van der Bij W The effect of lung transplantation on health-related quality of life: a longitudinal study Chest 1998 113 358 64 9498952
van den Berg JW Geertsma A van der Bij W Koëter GH de Boer WJ Postma DS TenVergert EM Bronchiolitis obliterans syndrome after lung transplantation and health-related quality of life Am J Respir Crit Care Med 2000 161 1937 41 10852770
Stavem K Bjortuft O Lund MB Kongshaug K Geiran O Boe J Health-related quality of life in lung transplant candidates and recipients Respiration 2000 67 159 65 10773787 10.1159/000029480
Limbos MM Joyce DP Chan CK Kesten S Psychological functioning and quality of life in lung transplant candidates and recipients Chest 2000 118 408 16 10936133 10.1378/chest.118.2.408
TenVergert EM Vermeulen KM Geertsma A van Enckevort PJ de Boer WJ Koëter GH Quality of life before and after lung transplantation in patients with emphysema versus other indications Psychol Rep 2001 89 707 17 11824741
Lanuza DM Lefaiver C McCabe M Farcas GA Garrity E Prospective study of functional status and quality of life before and after lung transplantation Chest 2000 118 115 22 10893368 10.1378/chest.118.1.115
Vermeulen KM Ouwens JP Van der Bij W De Boer W Koëter GH TenVergert EM Long-term quality of life in patients surviving at least 55 months after lung transplantation Gen Hosp Psychiatry 2003 25 95 102 12676422 10.1016/S0163-8343(03)00007-0
Festle MJ Qualifying the quantifying: Assessing the Quality of Life of Lung Transplant Recipients Oral History review 2002 29 59 68 12099304 10.1525/ohr.2002.29.1.59
Diggle PJ Liang KY Zeger SL Analysis of longitudinal data 1996 Clarendon press, Oxford
Curran D Molenberghs G Fayers PM Machin D Incomplete quality of life data in randomized trials:missing forms Statist Med 1998 17 697 709 10.1002/(SICI)1097-0258(19980315/15)17:5/7<697::AID-SIM815>3.3.CO;2-P
Twisk J De Vente W Attrition in longitudinal studies: How to deal with missing data J Clin Epidemiol 2002 55 329 337 11927199 10.1016/S0895-4356(01)00476-0
Rasbash J Browne W Goldstein H Yang M Plewis I Healy M Woodhouse G Draper D Langford I Lewis T A user's guide to MlwiN Version 21d for use with MlwiN 110 centre for Multilevel Modelling 2000 Institute of Education, University of London
Little RJA Modeling the drop-out mechanism in repeated-measures studies J Am Stat Ass 1995 90 112 121
Hunt SM Mc Ewen J McKenna SP Measuring health status 1986 London: Croom Helm
Ridout M Testing for random dropouts in repeated measurement data Biometrics 1991 47 1617 1621 1786335
Hedeker D Gibbons RD Application of random-effects pattern-mixture models for missing data in longitudinal studies Psychol meth 1997 2 64 78 10.1037//1082-989X.2.1.64
Singer JD Willet JB Applied longitudinal data analysis. Modeling change and event occurence 2003 Oxford university press
TenVergert EM Vermeulen KM Geertsma A van Enckevort PJ de Boer WJ van der Bij W Koëter GH Quality of life before and after lung transplantation in patients with emphysema versus other indications Psychol Rep 2001 89 707 17 11824741
Vermeulen KM Van der Bij W Erasmus ME Duiverman EJ Koëter GH TenVerger EM Improved quality of life after lung transplantation in individuals with cystic fybrosis Pediatr Pulmonol 2004 37 419 426 15095325 10.1002/ppul.20009
Reichenspurner H Girgis RE Robbins RC Yun KL Nitschke M Berry GJ Morris RE Theodore J Reitz BA Stanford experience with obliterative bronchiolitis after lung and heart-lung transplantation Ann Thorac Surg 1996 62 5 1467 1472 8893585 10.1016/0003-4975(96)00776-X
Cooper JD Billingham M Egan T Hertz MI Higenbottam T Lynch J Mauer J Paradis I Patterson GA Smith C A working formulation for the standardization of nomenclature and for clinical staging of chronic dysfunction in lung allografts. International Society for Heart and Lung Transplantation J Heart Lung Transplant 1993 12 5 713 6 8241207
|
16150142
|
PMC1236960
|
CC BY
|
2021-01-04 16:23:27
|
no
|
Respir Res. 2005 Sep 8; 6(1):101
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-101
|
oa_comm
|
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-991614657210.1186/1465-9921-6-99ResearchImpediment in upper airway stabilizing forces assessed by phrenic nerve stimulation in sleep apnea patients Sériès F [email protected]érin E [email protected] T [email protected] Centre de recherche, Hôpital Laval, Institut universitaire de cardiologie et de pneumologie de l'Université Laval, Quebec City, Quebec, Canada2 UPRES EA 2397, Université Paris VI Pierre et Marie Curie, Paris, France3 Service de Physiologie, GRHV EA 3830, Université de Rouen, Rouen, France4 Service de Pneumologie, Groupe Hospitalier Pitié-Salpêtrière, Assistance Publique - Hôpitaux de Paris, Paris, France2005 7 9 2005 6 1 99 99 29 6 2005 7 9 2005 Copyright © 2005 Sériès et al; licensee BioMed Central Ltd.2005Sériès et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The forces developed during inspiration play a key role in determining upper airway stability and the occurrence of nocturnal breathing disorders. Phrenic nerve stimulation applied during wakefulness is a unique tool to assess Upper airway dynamic properties and to measure the overall mechanical effects of the inspiratory process on UA stability.
Objectives
To compare the flow/pressure responses to inspiratory and expiratory twitches between sleep apnea subjects and normal subjects.
Methods
Inspiratory and expiratory twitches using magnetic nerve stimulation completed in eleven untreated sleep apnea subjects and ten normal subjects.
Results
In both groups, higher flow and pressure were reached during inspiratory twitches. The two groups showed no differences in expiratory twitch parameters. During inspiration, the pressure at which flow-limitation occurred was more negative in normals than in apneic subjects, but not reaching significance (p = 0.07). The relationship between pharyngeal pressure and flow adequately fitted with a polynomial regression model providing a measurement of upper airway critical pressure during twitch. This pressure significantly decreased in normals from expiratory to inspiratory twitches (-11.1 ± 1.6 and -15.7 ± 1.0 cm H2O respectively, 95% CI 1.6–7.6, p < 0.01), with no significant difference between the two measurements in apneic subjects. The inspiratory/expiratory difference in critical pressure was significantly correlated with the frequency of nocturnal breathing disorders.
Conclusion
Inspiratory-related upper airway dilating forces are impeded in sleep apnea patients.
Sleep apneaupper airwayphrenic stimulation
==== Body
Background
Sleep-related breathing disorders are fairly common in the general population [1]. In the majority of cases, they are obstructive in nature and are caused by recurrent sleep-related episodes of complete or partial upper airway (UA) closure. During these episodes, UA muscle activity progressively increases but the dilating force they develop is not sufficient to restore UA patency and normal ventilation until arousal or awakening. These breathing disturbances lead to intermittent hypoxia and hypercarbia, sleep disruption, and a sympathetic/parasympathetic imbalance that result in the development of daytime sleepiness [2], systemic hypertension [3], ischemic vascular disorders [4], and the clinical consequences of a prothrombotic state [5,6]. Given that the consequences of obstructive sleep apnea (OSA) are dramatically improved by the appropriate treatment [7-10] early diagnosis and treatment are essential.
The maintenance of UA patency depends on the balance between stabilizing and collapsing forces. The contraction of UA dilators generates the only stabilizing force that opposes a series of collapsing forces, including the effects of gravity-induced posterior displacement of UA structures, the negative inspiratory UA transmural pressure gradient, and surface tension forces. Numerous factors are involved in the determination of the effective stabilizing force applied to UA structures. These include the amount of UA neuromuscular activity [11], the physiological and histochemical properties of the muscles [12], the effectiveness of UA muscle contraction (i.e., synergistic or eccentric contractions, mechanical disadvantages) [13,14] and the mechanical coupling of UA muscles to surrounding soft tissues [15] Numerous attempts have been made to evaluate the mechanical importance of each of these individual factors, but the clinical and physiological impacts of combined UA stabilizing forces are unknown. A key observation is that UA compliance is increased in sleep apnea patients despite an higher neuromuscular drive to UA dilators in these patients during wakefulness and sleep [16,17], leading to the concept that the mechanical efficiency of UA dilating forces is impeded in these patients.
Phrenic nerve stimulation (PNS) can be used to study UA dynamics in conscious humans [18,19]. By provoking a diaphragm contraction independently of the neural activation of the UA dilators that normally precedes it, this technique mimics the dissociation between UA and diaphragm activities that is associated with the occurrence of OSA. Applying the PNS during expiration when UA dilators are only tonically active provides a pressure-flow relationship that reflects the mechanical properties of the tonically active UA. This pressure-flow relationship is dramatically modified if the PNS is applied when the UA dilators are phasically active [20]. Comparing the "expiratory" and "inspiratory" PNS-induced pressure-flow relationship provides a unique tool to measure the overall mechanical effects of the inspiratory process on UA stability. Since UA stability is vital in UA closure, the aim of the present work was to compare the ability of UA dilating forces to stabilize UA structures in normals and OSA subjects. Therefore, this work provides unique information on the impediment in upper airway stabilizing forces in sleep apnea patients.
Materials and methods
Patients
Twenty-one men (11 untreated sleep apneic, 10 normal) participated in this study. Normal subjects were recruited by newspaper advertisements and were free of symptoms suggestive of breathing disorders during sleep. The presence or absence of obstructive sleep apnea was always assessed with a polysomnographic recording. No subject was taking medication that could alter sleep or nocturnal breathing. The normal and OSA subjects had similar age and body mass index (BMI) ranges. The protocol was approved by the ethical review board of Hôpital Laval (Université Laval, Quebec City, Quebec, Canada). Subjects provided informed written consent.
Sleep recordings
Polysomnographic recordings consisted of in-lab continuous acquisition of electroencephalogram, electroocculogram, submental electromyogram, arterial oxyhemoglobin saturation by transcutaneous pulsed oxymetry, naso-oral airflow with thermistors, nasal pressure with nasal cannula, chest and abdominal movements by impedance plethysmography (Respitrace™, Ambulatory Monitoring Inc., Ardsley, NY), electrocardiogram, and breath sounds. Sleep position was continuously assessed by the attending technician using an infrared camera. All variables were digitally recorded (Sandman Elite™ System, Mallinckrodt, Kenilworth, NJ).
Phrenic nerve stimulation
Measurements
Surface recordings of right and left costal diaphragmatic EMG activities were obtained using silver cup electrodes placed on the axillary line of the sixth to eighth right and left intercostal spaces and connected to a electromyograph (Biopac System/Biopac, Santa Barbara, CA). A pressure-tipped catheter (Gaeltec, model CT/S X1058, Hackensack, NJ) was inserted in one nare after local anesthesia (1 ml of viscous 2% xylocaine) and located 16 cm from the nares to record hypophrayngeal (retroglossal) pressure (Pphar) [21]. An esophageal balloon was inserted through the other nare and located into the lower third of the oesophagus as assessed by the occlusion technique [22]. A plastic nasal stent (Nozovent; WPM International AB; Göteborg, Sweden) was placed in the anterior nares to prevent nasal collapse and the catheter was secured on the nose. A tightly fitting continuous positive pressure mask (Profile Light Nasal Mask, Respironics, Pittsburg, PA) was then placed over the nose. Airtightness of the mask was assessed by occluding the opening during maximal inspiratory efforts. A third catheter was passed through another opening of the mask to measure pressure inside the mask (Pmask). Flow was obtained from a pneumotachograph (Hans Rudolph, model 112467-3850A, Kansas City, MO) connected to the mask. Pressures and flow were digitally recorded at a 300 Hz sample rate (Digidata 1320, Axon Instruments, Foster City, CA). During the study, the subjects were seated in a comfortable armchair with a 60° inclination and their heads were maintained in a natural "neutral" position by a moulded pillow to ensure that the positions of the head and neck did not change during the experiment.
Phrenic nerve stimulation procedure
Bilateral anterior magnetic phrenic nerve stimulation (BAMPS) was performed using two Magstim 200 stimulators (Magstim Ltd,, Whitland, Dyfed, UK) connected to two 45 mm, figure eight-shaped coils with 90° handles, as previously described [20,23]. In brief, each stimulating coil was positioned antero-laterally over the anatomical landmark of the phrenic nerve in the neck at the posterior border of the strenomastoid muscle at the level of the cricoid cartilage. The handle of the coil was at a 45° angle to the mid-sagital and horizontal planes of the body. The optimal position and orientation of the coils was determined separately for each side at a 80 to 100 % stimulation intensity. A simplified recruitment curve (motor response to stimulation against stimulation intensity) was performed to verify the supramaximal nature of the stimulation. The two stimulators were triggered by a timer driven by the changes in flow direction. The 1 ms twitches were delivered after the operator-selected delay following inspiration or expiration onset had been reached.
Protocol
All measurements were made with the subjects breathing exclusively by the nose. BAMPS was applied at end-expiration (2 s after expiratory onset) or early inspiration (200 ms after inspiratory onset) in random order. Considering that UA muscles activity reaches a plateau within 200 ms of inspiratory onset [24,25], this timing was selected for inspiratory twitch to restrict the influence of differences in lung volumes between inspiratory and expiratory twitches. Subjects were blind to the twitch timing. For each respiratory timing, one series of five stimulations was applied at each stimulation intensity between 70% and 100% of maximal stimulation intensity with a 10% stepwise increase in intensity.
Data and statistical analyses
Data Analysis
Polysomnographic studies
Sleep and respiratory variables were scored according to standard criteria [26]. Normal subjects were defined by an apnea + hypopnea index (AHI) ≤ 15/h.
Flow/pressure curve
Twitch-induced breaths were considered flow-limited when instantaneous flow plateaued or decreased despite a persistent increase in driving pressure. Representative tracings of the , Poeso and Pphar responses to twitch are presented in Figure 1. Flow increased in response to decreasing pharyngeal pressure up to a maximal flow value (Imax lim), which was reached at a pressure value corresponding to Pphar lim. For pressure values below Pphar lim, flow and pressure values were dissociated with a decrease in flow as pressure became more negative. ΔI represented the drop in from Imax lim to flow nadir value (I min) for flow-limited twitches. The UA dynamic response of flow-limited twitches was characterized by modeling the twitch-related pressure-flow relationship from the rise in driving pressure up to its peak value as previously described [27]. According to this model, the pressure-flow relationship is fitted to a polynomial equation of the form = k1 Pphar + k2 Pphar2 using the least square method. Solving this equation for = 0 with Pphar different from zero (Pphar = k1/k2) provides the pharyngeal pressure value at which the UA is totally closed. The k1/k2 ratio is equivalent to the UA critical pressure value during twitch and is thus a descriptor of UA stability (the higher the k1/k2, the higher the UA stability). Polynomial model fitting and determinations of k1 and k2 values were performed semi-automatically using custom-made software (Matlab, The Mathwork Inc., Boston, MA). Imax lim, Pphar lim, ΔI, and k1/k2 were considered to characterize FL twitches. The difference in the inspiratory and expiratory values of the k1/k2 ratio was considered as an index of UA aperture inspiratory efficiency, the more negative the index, the lesser the impediment of UA dilating forces. The following variables were also measured: (1) peak flow values (Imax) of non-flow limited twitches, (2) corresponding Pphar value, (3) peak Pphar and peak Peso (Pphar max and Peso max respectively), (4) Pphar at 400 ml/s, (5) ΔI, and (6) the % of twitches associated with flow limitation.
Figure 1 Example of the changes in inspiratory flow and pharyngeal andesophageal pressures in response to PNS-induced diaphragm twitch applied during expiration and inspiration in a representative normal subject. Time 0 indicates the application of the twitch. I max and I min stand respectively for the maximal and minimal flow values reached during the twitch. Pphar lim indicates the pharyngeal pressure value corresponding to Imax. Peso peak indicates the driving pressure corresponding to I min. See text for abbreviations.
Statistical Analysis
A nested split-plot design was first completed to separately analyze the effects of stimulation intensity and twitch respiratory timing on the measured variables in OSA and normals. Since stimulation intensity increased repeatedly for each subject, the respiratory timing factors were randomly assigned to the subjects (main plot) and the stimulation intensity factors were assigned to the split plot. This second factor was analyzed as a repeated-measures factor. A mixed model analysis was then performed with interaction terms between groups, stimulation intensity, and respiratory timing factors. A first-order autoregressive covariance structure was used to evaluate the influence of stimulation intensity and timing on the measured variables [28]. The heterogeneity in covariance structures was defined based on the effect between groups and the respiratory timing. The univariate normality assumptions were verified with Shapiro-Wilk tests, and multivariate normalities were verified with Mardia tests [29]. The results were considered significant with p values ≤ 0.05. All analyses were conducted using the SAS statistical package (SAS Institute Inc., Cary, NC).
Results
No differences between normals and OSA subjects were found with regard to age, BMI, neck circumference, or pharyngeal resistance measured at iso-flow and peak flow (Table 1). In both groups, obstructive events represented 95 to 100 % of sleep-related breathing disturbances. Flow-limitation was consistently absent during tidal breathing in every subject. BAMPS-induced Peso max progressively rose with increasing stimulation intensity and was significantly higher during inspiratory than expiratory twitches (Table 2). For a given twitch timing, no significant difference was found between values obtained in normals and OSA subjects. The resulting flow was always higher when twitches were applied during inspiration than end-expiration with no difference between the two groups (Table 2), and remained unchanged with increasing stimulation intensity in both groups whatever the BAMPS timing.
Table 1 Anthropomorphic and polysomnographic characteristics of participating subjects. BMI: body mass index, NC: neck circumference, AHI: apnea + hypopnea index. Mean ± SEM.
Normal Subjects (n = 10) OSA Subjects (n = 11) p
Age (y) 50 ± 1 50 ± 2 0.9
BMI (Kg.m-2) 27.2 ± 1.1 26.7 ± 1.2 0.8
NC (cm) 39.6 ± 0.9 40.5 ± 0.8 0.6
Pharyngeal resistance at 0.4 L.s-1 (cm H2O.L-1.s) 3.9 ± 0.8 3.9 ± 0.7 0.3
Peak flow pharyngeal resistance (cm H2O.L-1.s) 4.4 ± 0.8 3.8 ± 0.7 0.2
AHI (n/h) 4.8 ± 1.0 29.4 ± 2.2 10-4
Table 2 Flow and pressure values obtained in response to 100% intensity PNS applied during expiration and inspiration in normals and OSA subjects. *: significant difference between inspiratory and expiratory values in a given group. Mean ± SEM.
Expiratory Inspiratory
Normal Subjects OSA Subjects Normal Subjects OSA Subjects
Imax (ml.s-1) 716 ± 36 739 ± 51 990 ± 42 * 1038 ± 58 *
Peso peak at 100% stimulation I (cm H2O) -17.6 ± 1.8 -14.0 ± 2.0 -21.1 ± 1.0 * -17.7 ± 1.9 *
Pphar peak at 100% stimulation I (cm H2O) -11.2 ± 2.2 -11.1 ± 2.1 -15.9 ± 1.5 * -14.7 ± 1.4 *
R Phar 400 ml/s (cm H2O.ml-1.s) 9.0 ± 0.7 9.5 ± 0.9 10.5 ± 1.3 8.5 ± 0.9
R Phar Imax (cm H2O.ml-1.s) 9.2 ± 0.8 11.1 ± 1.3 10.2 ± 0.9 8.4 ± 0.7
Imax lim (ml.s-1) 742 ± 248 752 ± 352 981 ± 297 * 1014 ± 348 *
Pphar lim (cm H2O) -8.4 ± 4.2 -7.2 ± 3.3 -12.5 ± 4.1 * -9.6 ± 3.6 *
ΔI (ml.s-1) 234 ± 26 256 ± 37 258 ± 35 234 ± 24
% twitches with FL 76.6 ± 3.4 79.0 ± 4.4 75.0 ± 3.5 84.4 ± 4.1
Twitch-induced flow-limited responses
Figures 1 and 2 depict expiratory and inspiratory twitch-induced instantaneous flow and pressure responses and the corresponding pressure/flow curves. The flow-limited nature of the twitch flow is clearly demonstrated by the flow/pressure pattern. Despite a progressive decrease in pharyngeal pressure, a rise in flow was only observed up to a pharyngeal pressure threshold before plateauing occurred and instantaneous flow dropped. The percentage of twitches with a flow-limitation pattern was identical in the two groups for expiratory twitches. The percentage of flow-limited twitches did not differ between expiratory and inspiratory twitches (Table 2). In both groups, Imax lim was significantly higher for inspiratory than expiratory twitches (Table 2), with no influence of stimulation intensity. For a given timing, no difference was found in Imax between normal subjects and OSA. Pphar lim progressively increased with rising stimulation intensity for both inspiratory and expiratory twitches. In both groups, Pphar lim was significantly lower during inspiratory twitches. Pphar lim measured during inspiratory twitches were more negative in normals than in OSA but the difference between the two groups did not reach significance (p = 0.07) (Table 2). P phar peak of flow-limited twitches also decreased with increasing stimulation intensity and reached more negative values during inspiratory twitches. No difference was found in the ΔI between the two groups for each PNS timing (Table 2).
Figure 2 Plot of the flow/pharyngeal pressure relationships obtained with data presented in Figure 1. The flow-limited nature of the twitch-induced flow is clearly demonstrated by the flow drop associated with the pharyngeal pressure decrease once Pphar lim and the corresponding I max have been reached. A clear difference in the flow/pressure curves can be seen between the expiratory and inspiratory twitches.
Driving pressure-flow relationship
The = k1Pphar+k2Pphar2 model adequately described the relationship between pharyngeal pressure and the related flow of flow-limited twitches in all but one subject (r 0.71 to 1.0, mean r: 0.91, p < 0.0001). The model could not be applied on flow-limited twitches in two OSA subjects due to the presence of artefacts. In the other subjects, artefacts occurred in 13.6 % of twitches that prevented data interpretation at some stimulation intensities between 60 to 90%. The index of UA aperture inspiratory efficiency was significantly influenced by stimulation intensity, twitch timing, and subject group. It decreased with rising stimulation intensity, and was lower during inspiratory than expiratory twitches (Figure 3). In normals, k1/k2 significantly decreased from expiratory to inspiratory twitches (-11.1 ± 1.6 and -15.7 ± 1.0 cm H2O respectively, 95% CI 1.6–7.6, p < 0.01), with no difference between the two twitches timings in OSA subjects (-12.0 ± 1.1 and -10.9 ± 1.3 cm H2O respectively, 95% CI -3.2 to 1.1, p = 0.3). Similar results were obtained when only considering data collected at 100 % stimulation intensity. The UA aperture inspiratory efficiency was higher in normals than in OSA subjects. The individual values of this index measured at 100% stimulation intensity positively correlated with the apnea + hypopnea index (r = 0.73, p = 6.10-4, Figure 4).
Figure 3 k1/k2 ratios for expiratory and inspiratory twitches in the normal and OSA groups. There was no difference between the two groups for expiratory twitches while the ratio decreased significantly during inspiratory twitches in normals and increased in OSA subjects. Boxes identify the 25 to 75 th percentiles of the data with the median value indicated. Horizontal lines outside the boxes depict the 10 to 90 th percentiles. Closed circles represent outliers. *: significant difference between inspiratory and expiratory values.
Figure 4 Correlation between individual values of the apnea + hypopnea index and the corresponding UA aperture inspiratory efficiency index.
Discussion
This study showed that inspiratory UA stabilizing forces were altered in OSA subjects compared to normals matched for age, sex, and BMI. The magnitude of the decrease was statistically associated with the frequency of nocturnal obstructive events.
The assessment of UA stability is a complex issue because of its dynamic nature. The role of UA dilating forces in the pathophysiology of UA closure has thus to be considered from a mechanistic point of view that takes into account the entire the sequence of events from the rise in respiratory and UA muscle activity to UA tissue displacement and/or stretching. UA shape and dimension determine the load imposed on phasic stabilizing forces by influencing the amount of tissue displacement resulting from a given increase in dilator activation [30]. As for the dynamic properties of UA during inspiratory twitches, the synergy between agonistic/antagonist muscle activation as well as the interaction with tracheal traction are important determinants of the UA stabilization process [11,13,31-33]. Another important aspect is the occurrence of eccentric contractions due to the drop in transmural pressure gradient [14]. Such contraction pattern has been found to contribute to the development of muscle fatigue in peripheral skeletal muscles [34]. Lastly, the stretching effect of UA tissues during inspiration largely depends on the mechanical properties of the soft tissues surrounding the UA muscles [15]. In this context, the PNS technique applied during expiration or inspiration provides unique information in that it can be used to differentiate UA tonic or phasic mechanical properties. The UA aperture inspiratory efficiency index thus reflects the overall effect of inspiratory UA traction forces and takes into account the efficiency of the contraction of UA muscles to generate tension, the transmission of this tension to surrounding tissues, as well as the effects of inspiratory tracheal traction.
Characterizing the entire pressure-flow response instead of specific flow and pressure values to evaluate twitch-induced UA dynamic improves the sensitivity of this method for detecting changes in UA properties. This is supported by the non-significant differences in I and Pphar measurements observed between the two groups. Indeed, the twitch-induced flow-pressure curve has an M-shape due to the initial rise in inspiratory flow up to a maximal value in response to driving pressure before decreasing to a minimal value that usually corresponds to the peak driving pressure, and then increasing again. This late increase in flow may be caused by a negative pressure-triggered reflex activation of UA dilators. The first part of the flow-pressure relationship was adequately described by a polynomial regression model I = k1Pphar + k2 Pphar2 with a negative k2 value. The k1 and k2 coefficients relate to airway conductance, k1 being the counterpart of RI max and k2 that of RI min. The k1/k2 ratio determines the driving pressure (in this case pharyngeal pressure) that should lead to the complete collapse of the UA and therefore is an index of UA stability in response to an over-imposed negative intra-thoracic pressure pulse. The fact that k1/k2 decreases from expiration to inspiration in normals indicates that phasic activity of UA dilators stabilizes the UA.
It is interesting to note that the difference in UA stability between normals and OSA subjects was influenced by the respiratory time during which the PNS was applied. During expiration, there was no difference in critical pressure between the two groups. This differs from the results previously published by our laboratory [35]. However, such apparent discrepancies can be easily accounted for by (1) the anthropometric differences between the control and apneic groups in our previous study possibly originating from the patients recruitment procedure, (2) the completion of twitches exclusively at 100 % intensity in our previous study and (3) the differences in driving pressure assessment (esophageal pressure vs. pharyngeal pressure). The absence of difference in UA mechanics in non-phasically active UA could reflect the mechanical effect of higher UA muscle tonic activity in sleep apnea patients [36] to compensate for the deleterious effect of the alteration in UA shape and/or dimension. Moving from expiration to inspiration was associated with an increase in critical pressure in normals and with a paradoxical decrease in critical pressure in OSA subjects.
Before elaborating on the possible explanation for such findings, it is first important to consider that in our study, UA stability was assessed by the flow/pressure response to an over-imposed negative intra-thoracic pressure pulse. The amplitude of this one can be considered as in the physiological range since it is generated by the respiratory system following maximal stimulation of the diaphragm via the phrenic nerves. However, the steepness of the rise in driving pressure that is specific to the phrenic nerve stimulation response (square pressure wave) and differs from the pressure profile generated during tidal breathing.
Several factors could account for the inability of UA dilating forces to overcome a super-imposed collapsing force during inspiration in OSA subjects such as a smaller increase in inspiratory EMG activity, an altered EMG response to negative pressure, or altered mechanical efficiency of UA dilator contraction. The phasic activity of UA muscles is known to preceed that of respiratory muscles in normals [37,38] as well as in OSA subjects while awake and during the post-apneic period [39]. Therefore, it is reasonable to assume that UA muscle recruitment was maximal (or almost maximal) at the time of application of inspiratory BAMPS in both groups. Furthermore, previous results suggest that this phasic activity should represent a higher percentage of spontaneous maximal UA neuromuscular activity in OSA subjects than in normals [16]. Concerning the reflex-induced rise in UA muscles activity following twitches, this one should be larger for inspiratory than for expiratory twitches [40], with a greater response amplitude in OSA subjects than normals [41]. Should this reflex response have influenced the twitch-induced flow/pressure response, it would attenuate the difference in UA dynamics that we observed between normals and OSA. In this context, the inspiratory/expiratory difference in UA stability measured with our technique reflected the net effect of UA dilating forces on UA stability. Our results thus support the concept that the splinting effects of inspiratory forces is altered in OSA subjects as a consequence of the circumstances described above and contribute to a decrease in the stretching effect of UA muscle contraction. Given the dramatic impact of sleep on upper airway muscles' responses to negative pressure, it can be speculated that the observed difference in UA aperture inspiratory efficiency between apneic and non-apneic subjects should be larger during sleep.
We found a significant relationship between AHI and the UA aperture inspiratory efficiency index. Even if there is no causal link between these parameters, the rise in AHI with decreasing the UA aperture inspiratory efficiency suggests that net inspiratory dilating forces are involved in UA stability and nocturnal obstructive breathing disorders. Given the importance of weight in determining the frequency of nocturnal breathing disorders and UA collapsibility [42], it should be remembered that our normal and apneic groups had identical anthropomorphic characteristics. Our results thus confirm the concept of the role of UA collapsibility in the development of obstructive breathing disorders [43-45], and also indicate that inspiratory stabilizing forces, as assessed in the present study, are cornerstones in the occurrence of such disorders. This may explain the conflicting results reported in the literature on the relationship between AHI and critical pressure when the relationship is assessed using different techniques.
We feel that the present study provides a useful contribution to the understanding of the effects of inspiratory stabilizing forces on the behaviour of the UA and its role in the pathophysiology of sleep apnea syndrome.
Competing interests
The author(s) declare that they have no financial or non-financial competing interests
Authors' contributions
FS conceived of the study, elaborated its design and contributed to its coordination. TS and EV participated in the revision of the design of the study. EV performed the analysis of the flow/pressure curves. All authors participated in and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
Authors thank S. Simard for the statistical analysis, G Éthier for recruitment of subjects, data collection and analysis, and the subjects for their participation in the study.
This work was supported by CIHR grant MT 13 768. F. Sériès is a scholar of the Fonds de Recherche en Santé du Québec, T. Similowski is supported in part by a three-year "Legs Poix" contract from La Chancellerie des Universités de Paris, Paris, France and by the Association pour le Développement et l'Organisation de la Recherche en Pneumologie (ADOREP), Paris, France;. E Vérin is supported by Université of Rouen, France.
==== Refs
Young T Palta M Dempsey J Skatrud J Weber S Badr S The occurrence of sleep-disordered breathing among middle-aged adults N Engl J Med 1993 328 1230 1235 8464434 10.1056/NEJM199304293281704
Guilleminault C Van den Hoed J Mitler M Guilleminault C, Dement WC, Alan R Liss Clinical overview of the sleep apnea syndrome Sleep apnea syndromes 1978 New York 1 12
Peppard PE Young T Palta M Skatrud J Prospective study of the association between sleep-disordered breathing and hypertension N Engl J Med 2000 342 1378 1384 10805822 10.1056/NEJM200005113421901
Shahar E Whitney CW Redline S Lee T Newman AB Nieto J O'Connor GT Boland LL Schwartz JE Samet J Sleep-disordered breathing and cardiovascular disease Am J respir Crit Care Med 2001 163 19 25 11208620
Eisenher I Noachtar S Haematological aspects of obstructive sleep apnoea Sleep Med Rev 2001 5 207 221 12530987 10.1053/smrv.2001.0158
Arnulf I Merino-Andreu M Perrier A Birolleau S Similowski T Derenne JP Obstructive sleep apnea and venous thromboembolism JAMA 2002 287 2655 2656 12020300 10.1001/jama.287.20.2655
Sullivan CE Issa FG Berthon-Jones M Eves L Reversal of obstructive sleep apnoea by continuous positive airway pressure applied through the nares Lancet 1981 1 862 865 6112294 10.1016/S0140-6736(81)92140-1
Lamphere J Roehrs T Wittig R Zorick F Conway WA Roth T Recovery of alertness after CPAP in apnea Chest 1989 96 1364 1367 2684554
Chin K Ohi M Kita H Noguchi T Otsuka N Tsuboi T Mishima M Kuno K Effects of NCPAP therapy on fibrinogen levels in obstructive sleep apnea syndrome Am J Respir Crit Care Med 1996 153 1972 1976 8665063
Becker HF Andreas J Ploch T Psych D Grote L Penzel T Sullivan CE Peter H Effect of nasal continuous positive airway pressure treatment on blood pressure in patients with obstructive sleep apnea Circulation 2003 107 68 73 12515745 10.1161/01.CIR.0000042706.47107.7A
Remmers JE Degroot WJ Sauerland EK Anch AM Pathogenesis of upper airway occlusion during sleep J Appl Physiol 1978 44 931 938 670014
Sériès F Simoneau JA St Pierre S Marc I Genioglossus and musculus uvulae muscle characteristics in sleep apnea hypopnea syndrome and in snorers Am J Respir Crit Care Med 1996 153 1870 1874 8665048
Kuna S Brennick MJ Effects of pharyngeal muscle activation on airway pressure-area relationships Am J Respir Crit Care Med 2002 166 972 977 12359656 10.1164/rccm.200203-214OC
Brennick MJ Parisi RA England SJ Influence of preload and afterload on genioglossus muscle length in awake goats Am J Respir Crit Care Med 1997 155 2010 2017 9196109
Sériès F Côté C St Pierre S Dysfunctional mechanical coupling of upper airway tissues in sleep apnea syndrome Am J Respir Crit Care Med 1999 159 1551 1555 10228125
Mezzanotte WS Tengel DJ White DP Waking genioglossal electromyogram in sleep apnea patients versus normal controls J Clin Invest 1992 89 1571 1579 1569196
Mezzanotte WS Tangel DJ White DP Influence of sleep onset on upper airway muscle activity in apnea patients versus normal controls Am J Respir Crit Care Med 1996 153 1880 1887 8665050
Sériès F Demoule A Marc I Sanfacon C Derenne JP Similowski T Inspiratory flow dynamics during phrenic nerve stimulation in awake normals during nasal breathing Am J Respir Crit Care Med 1999 160 614 620 10430737
Sériès F Straus C Demoule A Attali V Arnulf I Derenne JP Similowski T Assessment of upper airway dynamics in awake patients with sleep apnea using phrenic nerve stimulation Am J Respir Crit Care Med 2000 162 795 800 10988085
Sériès F Éthier G Assessment of upper airway stabilizing forces with the use of phrenic nerve stimulation in conscious humans J Appl Physiol 2003 94 2289 2295 12598484 10.1063/1.1593223
Hudgel DW Hendricks C Hamilton HB Characteristics of the upper airway pressure-flow relationship during sleep J Appl Physiol 1988 64 1930 1935 3391893
Baydur A Berhrakis P Zin WA Jaeger M Milic-Emili J A simple method for assessing the validity of oesophageal balloon technique Am Rev Respir Dis 1982 126 788 791 7149443
Mills GH Kyroussis D Hamnegard CH Polkey MI Green M Moxham J Bilateral magnetic stimulation of the phrenic nerves from an anterolateral approach Am J Respir Crit Care Med 1996 154 1099 1105 8887614
Pillar G Fogel RB Malhotra A Beauregard J Edwards JK Shea SA White DP Genioglossal inspiratory activation: central respiratory vs mechanoreceptive influences Respir Physiol 2001 127 23 38 11445198 10.1016/S0034-5687(01)00230-4
Hairston LE Sauerland EK Electromyography of the human palate: discharge patterns of the levator and tensor veli palatini Electromyogr Clin Neurophysiol 1981 21 287 97 7262008
American Academy of Sleep Medecine Task Force Report. Sleep-related breathing disorders in adults: Recommendations for syndrome definition and measurement techniques in clinical research Sleep 1999 22 667 689 10450601
Vérin E Similowski T Sériès F CPAP-induced changes in upper airway dynamics assessed with magnetic phrenic nerve stimulation in awake obstructive sleep apnea syndrome J Physiol 2003 546 1 London 279 287 12509495 10.1113/jphysiol.2002.029215
Burnham KP Anderson RD Model Selection and Inference. A Practical Information-Theoretic Approach Springer 1998 Chap. 2.2, 2.4, p. 43–48, 51–54
Mardia KV Applications of some measures of multivariate skewness and kurtosis in testing normality and robustness studies Sankhya B 1974 36 115 128
Leiter JC Upper airway shape. Is it important in the pathogenesis of obstructive sleep apnea? Am J respir Crit Care Med 1996 153 894 898 8630569
Strohl KP Wolin AD Van Lunteren E Fouke JM Assessment of muscle action on upper airway stability in anesthetized dogs J Lab Clin Med 1987 110 221 230 3598349
Van De Graaff WB Thoracic influence of upper airway patency J Appl Physiol 1988 65 2124 2131 3209556
Thut DC Schwartz AR Roach D Wise RA Permutt S Smith PL Tracheal and neck position influence upper airway airflow dynamics by altering airway length J Appl Physiol 1993 75 2084 2090 8307863
Newham DJ Mills KR Quigley BM Edwards RHT Pain and fatigue following concentric and eccentric muscle contractions Clin Sci Lond 1983 64 55 62 6822050
Vérin E Similowski T Teixeira A Sériès F Discriminative power of phrenic twitch-induced dynamic response for the diagnosis of sleep apnea during wakefulness J Appl Physiol 2003 94 31 37 12391097 10.1063/1.1580192
Fogel RB Malhotra A Pillar G Edwards JK Beauregard J Shea SA White DP Genioglossal activation in patients with obstructive sleep apnea versus control subjects Am J Respir Crit Care Med 2001 164 2025 2030 11739130
Strohl KP Hensley MJ Hallett M Saunders NA Ingram RH Activation of upper airway muscles before onset of inspiration in normal humans J Appl Physiol 1980 49 638 642 6777347
Van Der Touw T O'Neill N Amis T Wheatley J Brancatisano A Soft palate muscle activity in response to hypoxic hypercapnia J Appl Physiol 1994 77 2600 2605 7896597
Hudgel DW Harasick T Fluctuation in timing of upper airway and chest wall inspiratory muscle activity in obstructive sleep apnea J Appl Physiol 1990 69 443 450 2228853
Woodall DL Horanson JA Mathew OP Time of application of negative pressure pulses and upper airway muscle activity J Appl Physiol 1998 67 366 370 2759965
Berry RB White DP Roper J Pillar G Fogel RB Stanchina M Malhotra A Awake negative pressure reflex response of the genioglossus in OSA patients and normal subjects J Appl Physiol 2003 94 1875 1882 12533492
Schwartz AR Gold AR Schubert N Stryzak A Wise RA Permutt S Smith PL Effect of weight loss on upper airway collapsibility in obstructive sleep apnea Am Rev Respir Dis 1991 144 494 498 1892285
Suratt PM McTier RF Wilhoit SC Collapsibility of the nasopharyngeal airway in obstructive sleep apnea Am Rev Respir Dis 1985 132 967 971 4062051
Isono S Morrison DL Launois SH Feroah TR Whitelaw WA Remmers JE Static mechanics of the velopharynx of patients with obstructive sleep apnea J Appl Physiol 1993 75 148 154 8376260
Gleadhill IC Schwartz AR Schubert N Wise RA Permutt S Smith PL Upper airway collapsibility in snorers and in patients with obstructive hypopnea and apnea Am Rev Respir Dis 1991 143 1300 1303 2048817
|
16146572
|
PMC1236961
|
CC BY
|
2021-01-04 16:23:25
|
no
|
Respir Res. 2005 Sep 7; 6(1):99
|
utf-8
|
Respir Res
| 2,005 |
10.1186/1465-9921-6-99
|
oa_comm
|
==== Front
Theor Biol Med ModelTheoretical Biology & Medical Modelling1742-4682BioMed Central London 1742-4682-2-311609553910.1186/1742-4682-2-31ResearchAllometric scaling of the maximum metabolic rate of mammals: oxygen transport from the lungs to the heart is a limiting step Painter Page R [email protected] Office of Environmental Health Hazard Assessment, California Environmental Protection Agency, P. O. Box 4010, Sacramento, California 95812, USA2005 11 8 2005 2 31 31 22 3 2005 11 8 2005 Copyright © 2005 Painter; licensee BioMed Central Ltd.2005Painter; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The maximum metabolic rate (MMR) of mammals is approximately proportional to M0.9, where M is the mammal's body weight. Therefore, MMR increases with body weight faster than does the basal metabolic rate (BMR), which is approximately proportional to M0.7. MMR is strongly associated with the capacity of the cardiovascular system to deliver blood to capillaries in the systemic circulation, but properties of this vascular system have not produced an explanation for the scaling of MMR.
Results
Here we focus on the pulmonary circulation where resistance to blood flow (impedance) places a limit on the rate that blood can be pumped through the lungs before pulmonary edema occurs. The maximum pressure gradient that does not produce edema determines the maximum rate that blood can flow through the pulmonary veins without compromising the diffusing capacity of oxygen. We show that modeling the pulmonary venous tree as a fractal-like vascular network leads to a scaling equation for maximum cardiac output that predicts MMR as a function of M as well as the conventional power function aMb does and that least-squares regression estimates of the equation's slope-determining parameter correspond closely to the value of the parameter calculated directly from Murray's law.
Conclusion
The assumption that cardiac output at the MMR is limited by pulmonary capillary pressures that produce edema leads to a model that is in agreement with experimental measurements of MMR scaling, and the rate of blood flow in pulmonary veins may be rate-limiting for the pathway of oxygen.
==== Body
Introduction
The maximum metabolic rate (MMR) of mammals is measured as the rate of oxygen consumption during the maximum sustainable rate of exercise [1]. Unlike the basal metabolic rate (BMR), which consumes oxygen at rates far below the delivery capacity of the cardiovascular system [1,2], the MMR is largely determined by the maximal rate that the cardiovascular system can deliver oxygen to mitochondria in muscle tissue [1].
MMR has been measured in mammals ranging in size, M, from 0.007 kg (pygmy mice) to 575 kg (cattle). Regression of the logarithm of MMR (denoted Q) on the logarithm of M gives a maximum-likelihood estimate (MLE) of the exponent b in the allometric expression
Q = aMb (1)
of 0.872 with a 95% confidence interval (CI) of 0.812–0.931 for MMR data from 32 mammalian species [1]. In contrast, regression analysis of BMR data from 619 mammalian species gives a MLE of the slope, 0.69, with 95% CI, 0.68–0.70 [3]
To explain the scaling of the metabolic rate in mammals, West et al. [4] and Bengtson and Eden [5] model the arterial network as a structure that starts with a single tube (aorta) that repeatedly branches into two (or more) smaller tubes. Branching continues until a tube (small arteriole) that supplies capillaries is reached. They assume that all paths from the heart to capillaries pass through n tubes and that the arterial network is a truncated self-similar fractal (i.e., a fractal-like network). The smallest vessels of the circulatory system have dimensions that vary little with body size, whereas the dimensions of the aorta and other great vessels are highly dependent on size. For convenience, we define level 1 of the arterial tree (or venous tree) as the smallest arterioles (or venules). These have radius r1 and length l1. Each level 2 vascular tube with radius r2 and length l2 is connected to η1 level 1 structures. In general, each level i+1 tube of radius ri+1 and length li+1 is connected to ηi level i tubes. It follows from the assumption of a self-similar fractal that the branching ratio is a constant (denoted η) and that the ratio of tube lengths, li+1/li, is also a constant (denoted L) throughout the network.
The theory of West et al. minimizes the (pressure) × (volume) work of the heart that is required to pump one unit of blood against a difference in pressure equal to the pressure in the aorta minus the pressure in capillaries. This work per unit of blood flow is proportional to the impedance in the arterial network. Minimization of this energy cost for pulsatile flow in arteries is claimed to require area-preserving branching of the network (i.e., the ratio ri+1/ri, termed R, is equal to η1/2) and, as a consequence, to require that the density of capillaries in tissues is proportional to M-1/4 (assuming that the diameter of the aorta scales proportionally to M3/8 or that arterial blood volume scales proportionally to M). The theory's 3/4-power scaling prediction for metabolic rate follows from the assumption that metabolic rate is proportional to the total number of capillaries calculated as tissue capillary density multiplied by M, an assumption that is reasonable for MMR but not for BMR [1]. The theory of Bengtson and Eden assumes that energy dissipation per endothelial surface area is constant, leading to the conclusions that R is equal to η2/5 and that the total number of capillaries is proportional to M15/17 if the volume of blood in arteries scales proportionally to M. If it is assumed that the diameter of the aorta scales proportionally to M3/8, the number of capillaries is proportional to M15/16.
The scaling of the total number of capillaries in skeletal muscle, where over 90% of energy metabolism occurs during MMR exercise, is nearly identical to the scaling of MMR [1], and, as noted above, this scaling is not proportional to M3/4. The 95% CI for the scaling exponent for total capillary volume, 0.909 – 1.0559, contains 15/16 but not 3/4. Moreover, if either of these theories is adequate for predicting capillary density, it should correctly predict the scaling exponent for capillaries in the lung, which is 1.00 with 95% CI of 0.912 – 1.087 [6]. This CI contains 15/16 but not 3/4.
A model for the maximum metabolic rate
While minimization of impedance does not by itself lead to a correct prediction of capillary density in muscle and lung tissue, it is clearly an important principle for design of mammalian vascular systems [7,8]. The potential importance of impedance is most apparent in the pulmonary venous circulation, where the entire output of the heart's right ventricle flows before blood enters the left atrium of the heart. The driving force for pulmonary venous return to the heart is the pressure at the venous end of pulmonary capillaries minus the diastolic pressure in the left atrium (denoted PLA).
The output of oxygen by the left ventricle of the heart into the aorta is equal to the input of oxygen from the lungs to the heart. This is equal to the cardiac blood output rate multiplied by the maximum amount of oxygen per ml of blood multiplied by the percent saturation of blood with oxygen. Pressure in the model is strictly increasing with flow. However, as pressure rises above oncotic pressure, interstitial edema increases and then more and more fluid accumulates within alveoli. Therefore, oxygen saturation is strictly decreasing as a consequence of the increasing barrier to oxygen diffusion from pulmonary air into capillaries. As a result, there is a blood flow rate, denoted Fmax, that produces the maximum uptake of oxygen in the lungs, which is also the maximum output of oxygen to the body. The pressure near the venous end of alveolar capillaries at Fmax is denoted Πmax. Consequently, the pressure gradient that drives the return of blood in pulmonary capillaries back to the heart is
ΔPmax = FmaxIp (2)
where ΔPmax = Πmax - PLA and Ip is the impedance of the pulmonary venous network. It is assumed that Πmax is proportional to the oncotic pressure of blood, denoted Πo. The value of Πmax is assumed to be approximately the same in mammals of different sizes because Πo appears to be nearly invariant in mammalian species, being approximately 20 mm Hg [9-11] and PLA is approximately 1 mm Hg. (All pressures in this article are measured relative to ambient pressure.) Therefore, the scaling of Fmax with body size depends largely on the scaling of Ip.
The impedance of the pulmonary venous network is a consequence of its physical structure and the viscosity of blood (termed ν). The pulmonary arteries and veins form parallel fractal-like networks in each lung with arteries and veins of the same level having similar dimensions [12,13]. Small venules have dimensions that are body-size-invariant (r1 approximately 10-5 m and l1 approximately 10-4 m). These vascular tubes receive blood from the capillaries in pulmonary acini, the structures that comprise approximately 10,000 alveoli and that appear to be body-size-invariant in mammals [14].
The impedance of a fractal-like network is the sum of impedances contributed by each level of the network. We assume that the impedance Ii due to level i is the value calculated from the Poiseuille theory for non-turbulent fluid flow, , where Ni is the number of level i vessels [4]. Consequently, Ii+1 is equal to . The observation that dimensions within acini are size-invariant leads to the conclusion that η, R and L are size-invariant in acini. We assume that these ratios remain constant throughout the network. Therefore, the factor ηL/R4 (denoted α) is assumed to be size-invariant, and the expression for Ip is a geometric series (when α ≠ 1) that simplifies to
Substitution of this formula into Equation (2) gives
The assumption that the acinus is a size-invariant structure implies that the number of level 1 venules per acinus is independent of body size. Consequently, the total number of level 1 venules, N1, is proportional to lung volume, which is proportional to body mass M [6]. The parameter n is the number of branchings from the pulmonary vein to level 1 venules. Therefore ηn = N1∝ M, which is written as ηn = M /M1. The constant M1 is the mass of body tissue supplied with the oxygen in blood flowing through a single level 1 venule. This is estimated to be approximately 10-5 kg [15,16] leading to the equation n = [log(M)-log(10-5)]/log(η). Substitution for N1 and n in Equation (4) gives Fmax = KM/ [1-ζlog(M)-log(0.00001)], where ζ = α1/log(η) and K is the constant . The maximal rate oxygen uptake in the lungs, Q, is UoFmax,, where Uo is the oxygen uptake in the lungs per unit of blood. Therefore, when α ≠ 1,
Q = UoC M/ [1-ζlog(M)-log(0.00001)] (5)
where C is a constant. Note that ζ depends on the base used to define the logarithm. The base 10 is used in the following regression analysis. When α = 1, we have
Q = UoC M/ [log(M)-log(0.00001)]/ log(η) (6)
Equation (5) is termed the general pulmonary venous flow capillary pressure model (PVFCP model), and Equation (6) is termed the constrained PVFCP model.
Testing model predictions
The conventional method for determining the best fit of Equation (5) or Equation (1) to oxygen uptake data is to find the values of the two parameters in the model that correspond to a minimum of the sum of squares of residuals (SSR), where a residual is defined as the logarithm of a measured value of the uptake rate minus the logarithm of the uptake rate predicted by the model for a mammal of the experimentally measured weight M. The technique is termed least squares logarithmic regression (LSLR). Figure 1 shows the best fit of the standard allometric model, Equation (1), to the data in Table 1. The minimal SSR occurs when b is 0.872 and the SSR is 1.6307. Figure 2 shows that the model of Equation (5), assuming that Uo is constant, fits the data equally well: the minimal SSR occurs when the parameter ζ, which determines the slope of this scaling expression, is 1.193, and the SSR is 1.6269.
Figure 1 Regression analysis of MMR data in Table 1 (VO2 max in ml/min and body weight in kg) using the standard linear model, Equation (1). The minimum SSR is 1.6308.
Table 1 Maximum metabolic rates (VO2 max) of mammals from Weibel et al.[1].
Mammal M (kg) VO2 max (ml/min)
Pygmy mouse 0.0072 1.884
Woodmouse 0.02 5.28
Deer mouse 0.022 4.928
Mouse 0.026 3.884
Chipmunk 0.09 21.485
Mole rat 0.136 14.58
Rat 0.278 23.13
Dwarf mongoose 0.43 54.44
Guinea pig 0.584 32.59
Rat kangaroo 1.1 194.7
Banded mongoose 1.14 130
Genet cat 1.38 146.6
Spring hare 3 291.6
Agouti 3.22 328.4
Suri 3.3 317.8
Dik-dik 4.2 228.1
Fox 4.51 897.5
Grant's gazelle 10.1 539.3
Coyote 12.4 2283.3
Pig 18.5 1731.6
African sheep 21.8 1013.7
Goat 24.3 1344.7
Dog 25.9 3825
Wolf 27.6 4310
Pronghorn 28.4 8435
Lion 30 1800
Wildebeest 102 4468
Waterbuck 110 5172
Calf 141 5161
Pony 171 15185
Zebu cattle 193 5660
Eland 240 8640
Horse 453 56005
Steer 475 24225
Figure 2 Regression analysis of MMR data in Table 1 (VO2 max in ml/min and body weight in kg) using the model of Equation (5). The closed circles are the data points from Table 1, and the open circles are the graph of the physiologically-based model, Equation (5), with parameters calculated from LSLR. The minimum SSR is 1.6263.
In the analysis of data in Table 1, it is assumed that maximum oxygen uptake is proportional to cardiac output (i.e. Uo is constant). A more reasonable assumption is that oxygen uptake is proportional to cardiac output multiplied by the hemoglobin concentration of blood. The data in Table 2 include values of the hematocrit, which is nearly proportional to hemoglobin concentration. Therefore, the maximal rate of oxygen uptake multiplied by 0.42 and divided by the hematocrit (i. e., the oxygen uptake adjusted to a hematocrit of 0.42) is now assumed to be proportional to maximum cardiac output.
Table 2 Maximum metabolic rates of mammals adjusted to a standard hematocrit of 0.42 from Weibel et al.[1].
Mammal Body mass (kg) Hematocrit VO2 max (ml/min)
Measured value Adjusted value
Woodmouse 0.02 0.42 5.28 5.28
Mole rat 0.129 0.42 13.61 13.61
Rat 0.148 0.42 15.55 15.55
Guinea pig 0.595 0.5 33.2 27.888
Agouti 3.22 0.42 328.44 328.44
Fox 4.4 0.42 955.7 955.7
Goat 21 0.299 1386 1946.89
Dog 23.7 0.5 3455.5 2902.62
Pronghorn 28.4 0.456 8434.8 7768.895
Horse 446 0.55 60745.2 46387.24
Steer 475 0.4 24225 25436.25
LSLR using the data in Table 2 and the model of Equation (1) gives the value of 0.957 for b (Rc2 = .9697) and SSR = 0.5890) when the SSR is minimized. LSLR using Equation (5) finds that the SSR is minimized when ζ equals 0.801 (SSR = 0.5833). LSLR of predicted values of cardiac output from Equation (5) using values of M from Table 2 and the estimate for ζ of 0.801 gives b = 0.958 and Rc2 = 0.9991. Clearly the predictions from Equation (5) are again nearly indistinguishable from those of Equation (1), and Equation (5) fits these data as well as Equation (1) does.
While the logarithm of the function Q defined in Equation (5) is a nonlinear function of the logarithm of M, it is clear from Figure 2 that the logarithm of Q closely approximates a linear function of the logarithm of M. This observation is confirmed by substituting first-order approximations into Equations (5) and (6): The scaling of Q when α = 1 can be predicted directly from Equation (6). Multiplying and dividing by log(M1 ) gives Q ∝ (M/log(M1))/(1 - log(M)/log(M1)). Using logarithms to the base e and the first-order approximation loge(1+x) = x shows that loge(Q) is approximately equal to loge(M) + loge(M)/loge(M1) plus a constant , i.e., Q is approximately proportional to Mb where b = 1 + 1/loge(M1). For M1 = 0.00001 b = 0.914, which is close to the value from LSLR of data simulated using Equation (6). A similar approximation analysis of Equation (5) shows that it too is approximately a power function when α is approximately equal to 1. Figure 3 shows that, with the parameters used in Figure 2, the logarithm of Q defined in Equation (5) is nearly identical to a linear function of the logarithm of M.
Figure 3 Predicted values of MMR from Equation (5) for mammals with the body weights in Table 1. The straight line is the best fit of the standard allometric model, Equation (1), to the predicted values.
Comparison with Murray's law
The estimate of α = ηL/R4 corresponding to ζ is ζlog(η). For a branching ratio of 2 and ζ = 1.193, α is estimated to be 1.054. For a volume-filling fractal distribution network, it has been conjectured that [4]
L = η1/3, (7)
and this equation for L leads to the formula
R3 = 1.04η. (8)
Equation (8) is remarkably similar to Murray's law for the scaling of radii of arterial or venous networks, which states that flow rate is proportional to the third power of vessel radius [7]. For our network model, Murray's law implies R3 = η, and this equation together with the condition L = η1/3 implies α = 1. With this value of α, the slope of the logarithm of Equation (6) depends only on the estimate of M1. For M1 = 0.00001 kg, Equation (6) is nearly identical to a power function with b = 0.916. Therefore, Murray's law and the fractal length scaling relationship lead to the constrained PVFCP model and predict that the slope parameter of the scaling function is in the range of observed values.
Discussion
The PVFCP model predicts that the logarithm of maximum oxygen uptake in mammals is approximately proportional to the logarithm of body mass. If the radii of veins in the pulmonary venous tree obey Murray's law, then the constant of proportionality is in the range of experimentally observed values for MMR. The PVFCP model, like other published explanations for MMR scaling, focuses on the supply of oxygen to the tissues. However, the PVFCP model differs from other explanations for MMR scaling because it focuses on pulmonary blood flow.
The PVFCP model and the model of Bengtson and Eden [5] use the same mathematical description of pressure-flow relationships in a vascular tree. While the model of Bengtson and Eden [5] is consistent with current data on MMR, the model's assumption of energy dissipation that is proportional to vascular surface area is questionable as a principle of mammalian design. For example, a hypothetical mammalian species that replaces the R = η2/5 requirement of their theory with the R = η1/3 relationship of Murray's law would reduce total energy dissipation in arteries. This replacement would also give a higher predicted capillary density and consequently a higher MMR.
It is instructive to compare the number of independent parameters and assumptions in the PVFCP model with the number of parameters and assumptions in the two fractal-like models of the arterial network that predict metabolic scaling [4,5]. All three models describe the vascular network as a self-similar fractal-like tubular structure with pressure gradients determined by Poiseuille's law. All assume that the size of terminal (smallest) network tubes is the same in mammals of different size and that blood viscosity does not vary with body size. All contain the branching ratio parameter η and the network length parameter n. In the PVFCP model, a relationship between η, n and body mass is derived from the assumption that the number of terminal segments is proportional to body mass, an assumption that is supported by observations. In the other two models, a relation between these parameters is derived from the assumption that arterial blood volume is proportional to body mass, an assumption without direct observational support. Network structure is related to metabolic rate in the PVFCP model by Equation (5), which specifies the maximum rate of blood flow that does not compromise pulmonary function. In the other models, such a relation is derived from the assumption that metabolic rate is proportional to the number of capillaries in the systemic circulation. In the PVFCP model, there is one more independent parameter, α, which is defined by fitting experimental data. The other models have two additional parameters, L and R. Both models specify L indirectly using the assumption of Equation (7). The parameter R is specified by an energy minimization principle in one model [4] and by an energy dissipation principle in the other [5]. While the number of parameters and assumptions in the PVFCP model is relatively large, it is less than the number in the fractal-like network models previously published. Another recent mathematical description of metabolic scaling, the "Allometric Cascade" model [2], is not discussed here because it is not a mechanistic model. Indeed, the two models appear compatible because the PVFCP model could be integrated into the "Allometric Cascade" model to provide a mechanism-based scaling term for the maximum rate of blood flow.
Weibel et al. [1] argue that it is the volume of mitochondria in muscle tissue and the blood supply in capillaries in muscle tissue that determine the scaling of MMR. This view is supported by their demonstration that MMR is remarkably correlated with and is proportional to mitochondrial volume (b = 1.09, Rc2 = 0.9939) and to estimated capillary blood volume in muscle tissue (b = 0.975, Rc2 = 0.9846). However, total mitochondrial volume and blood volume in muscle capillaries can be increased by exercise conditioning, and the correlation between capillary surface area and MMR or between mitochondrial volume and MMR may arise from such conditioning.
In the formulation of the PVFCP model, the role of gravity in facilitating or impeding the return of pulmonary blood to the heart has been ignored. Blood that is one inch higher than the left atrium has potential energy to facilitate its return to the heart that is approximately equivalent to a 2 mm Hg pressure gradient. For small mammals (e.g., mice), gravitational effects would be small compared with the approximately 20 mm Hg pressure gradient that we assume drives blood return during MMR exercise. However, for large mammals (e.g., elephants and whales), the effects of gravity will significantly increase blood return from regions of lung above the heart, but decrease blood return from regions below the heart. Therefore, Equation (5) may not adequately describe MMR blood flow in large mammals.
A second reason for doubting the validity of Equation (5) for large mammals is that intervals of the heart cycle increase with body size. The minimum length of the heart cycle (at maximum heart rate) is largely composed of the time required for the ventricles to fill plus the time required for the ventricles to eject blood into the pulmonary artery and aorta. At maximal heart rate, ventricular filling time is nearly equal to the PR interval, which is approximately proportional to the 1/4-power of body mass [17]. If the sum of the QRS interval and the ST segment, which is nearly equal to the time required to eject blood from the ventricles, has similar scaling, then the scaling exponent for maximum heart rate is less than the scaling exponent for the MMR divided by body mass, i.e., the specific maximum metabolic rate (SMMR). Thus, maximum heart rate, not the limitation posed by pulmonary venous impedance, may limit MMR for very large mammals.
The biological plausibility of the relation between MMR and Ip proposed in the PVFCP model depends on whether pressures in lung capillaries approach the oncotic pressure of blood during periods of maximal exertion. In healthy humans at rest, the pressure difference between pulmonary capillaries and the left atrium ranges from approximately 5 to 11 mm Hg [18]. Assuming that the value of 5 mm Hg occurs when pulmonary veins are dilated, this pressure difference is predicted to increase by a factor of approximately 4 during heavy exercise in a trained athlete when cardiac output increases by a factor of 4 (assuming that the pulmonary veins are in a comparable state of dilation). This would require the capillary pressure to rise to approximately 21 mm Hg. It is noteworthy that signs of pressure stress are sometimes observed in pulmonary tissue from trained endurance athletes [19].
Studies of human patients with narrowing of the mitral valve, the valve between the left atrium and left ventricle, are consistent with the hypothesis that Ip limits maximum metabolic rate. This condition, termed mitral stenosis, causes an increase in PLA. Patients with a PLA below 20 mm Hg usually do not have pulmonary edema at rest but may develop it with exercise. Furthermore, women with a PLA between 18 and 20 mm Hg are at risk for developing pulmonary edema during pregnancy where the cardiac output at rest increases on average by approximately 50% [20-22].
Additional support for the proposed role of pulmonary impedance in determining MMR comes from studies of horses, which have an MMR well above the value predicted by the allometric equation fitted to the data in Table 1[1]. Horses at rest have pulmonary capillary blood pressures that are above those in humans with mitral stenosis and pulmonary edema with exercise. Horses are apparently able to exercise without developing pulmonary edema because they are able to "concentrate" their blood during periods of exertion. The concentration of erythrocytes (measured as the hematocrit) is increased during exercise [23]. This requires a preferential loss of water that likely occurs in capillaries of the systemic circulation. As a result, the concentration of albumin in blood is increased and the oncotic pressure of blood is increased. This adaptation enables a horse at a gallop to tolerate pulmonary capillary pressures as high as 38 mm Hg [24].
Horses possess a second adaptation that allows them to increase their SMMR. Their ratio of lung volume to body mass is approximately 20% greater than the average value for mammals [6]. To pump blood through their large lungs at an unusually high rate per unit lung volume, horses possess a heart that is larger (as a fraction of body mass) than the average value for mammals [25]. This enables them to achieve a SMMR that is more than twice that of a cow of similar size. However, even with its remarkable adaptations, no horse can sustain the SMMR that pygmy mice and other small mammals can achieve [1].
Competing interests
The author(s) declare that they have no competing interests.
Acknowledgements
I thank Charles Salocks and Danielle Ketchum for their careful reviews and helpful comments.
==== Refs
Weibel ER Bacigalupe LD Schmitt B Hoppeler H Allometric scaling of maximal metabolic rate in mammals: muscle aerobic capacity as determinant factor Respir Physiol Neurobiol 2004 140 115 132 15134660 10.1016/j.resp.2004.01.006
Darveau C-A Suarez RK Andrews RD Hochachka PW Allometric cascade as a unifying principle of body mass effects on metabolism Nature 2002 417 166 170 12000958 10.1038/417166a
White CR Seymour RS Mammalian basal metabolic rate is proportional to body mass2/3 Proc Natl Acad Sci USA 2003 100 4046 4049 12637681 10.1073/pnas.0436428100
West GB Brown JH Enquist BJ A general model for the origin of allometric scaling laws in biology Science 1997 276 122 126 9082983 10.1126/science.276.5309.122
Bengtson H-U Eden P A simple model for the arterial system J Theoret Biol 2003 221 437 443 12642118 10.1006/jtbi.2003.3198
Gehr P Mwangi DK Ammann A Maloiy GM Taylor CR Weibel ER Design of the mammalian respiratoey system. V. Scaling morphometric pulmonary diffusing capacity to body mass: wild and domestic animals Respir Physiol 1981 44 61 86 7232887 10.1016/0034-5687(81)90077-3
Murray CD The physiological principle of minimum work. I. The vascular system and the cost of blood volume Proc Natl Acad Sci USA 1926 12 207 214 16576980
LaBarbera M Principles of design of fluid transport systems in zoology Science 1990 249 992 1000 2396104
Stohrer M Boucher Y Stangassinger M Jain RK Oncotic pressure in solid tumors is elevated Cancer Res 2000 60 4251 4255 10945638
Gabel JC Scott RL Adair TH Drake RE Traber DL Errors in calculated oncotic pressure of dog plasma Am J Physiol 1980 239 H810 H812 7446756
Madigan JE Rahal CJ Measurement of Plasma Colloid Osmotic Pressure in Normal Thoroughbred Neonatal Foals 2003 Davis California: University of California
Agur AM Ming JL Grant JC Grant's Atlas of Anatomy 2004 Baltimore: Lippincott, Williams and Williams
Bergman RA Afifi AK Heidger PM Atlas of Microscopic Anatomy 1996 Philadelphia: WB Saunders
Weibel ER Fractal geometry: a design principle for living organisms Am J Physiol 1991 261 L361 L369 1767856
Turcotte DL Pelletier JD Newman WI Networks with side branching in biology J Theoret Biol 1998 193 577 592 9745754 10.1006/jtbi.1998.0723
Fung YC Biomechanics 1990 New York: Springer-Verlag
Noujaim SF Lucca E Munoz V Persaud D Berenfeld O Meijler FL Jalife J From mouse to whale: a universal scaling relation for the PR Interval of the electrocardiogram of mammals Circulation 2004 110 2802 2808 15505092 10.1161/01.CIR.0000146785.15995.67
Tabata T Oki T Fukuda N Iuchi A Manabe K Kageji Y Sasaki M Yamada H Ito S Influence of left atrial pressure on left atrial appendage flow velocity patterns in patients in sinus rhythm J Am Soc Echocardiogr 1996 9 857 864 8943446 10.1016/S0894-7317(96)90478-2
West JB Vulnerability of pulmonary capillaries during exercise Exerc Sport Sci Rev 2004 32 24 30 14748546 10.1097/00003677-200401000-00006
Reis G Motta MS Barbosa MM Esteves WA Souza SF Bocchi EA Dobutamine stress echocardiography for noninvasive assessment and risk stratification of patients with rheumatic mitral stenosis J Am Col Cardiol 2004 43 393 401 10.1016/j.jacc.2003.09.037
Desai DK Adanlawo M Naidoo DP Moodley J Kleinschmidt I Mitral stenosis in pregnancy: a four-year experience at King Edward VIII Hospital, Durban, South Africa BJOG 2000 107 953 958 10955424 10955424
van Oppen AC Stigter RH Bruinse HW Cardiac output in normal pregnancy: a critical review Obstet Gynecol 1996 87 310 318 8559545 10.1016/0029-7844(95)00348-7
Weber J-M Dobson GP Parkhouse WS Wheeldon D Harman JC Snow DH Hochachka PW Cardiac output and oxygen consumption in exercising Thoroughbred horses Am J Physiol 1987 253 R890 R895 3425767
Hackett RP Ducharme NG Gleed RD Mitchell L Soderholm LV Erickson BK Erb HN Do Thoroughbred and Standardbred horses have similar increases in pulmonary vascular pressures during exercise? Can J Vet Res 2003 67 291 296 14620866
Hoppeler H Lindstedt SL Claassen H Taylor CR Mathieu O Weibel ER Scaling mitochondrial volume in heart to body mass Respir Physiol 1984 55 131 137 6729269 10.1016/0034-5687(84)90018-5
|
16095539
|
PMC1236962
|
CC BY
|
2021-01-04 16:39:26
|
no
|
Theor Biol Med Model. 2005 Aug 11; 2:31
|
utf-8
|
Theor Biol Med Model
| 2,005 |
10.1186/1742-4682-2-31
|
oa_comm
|
==== Front
Theor Biol Med ModelTheoretical Biology & Medical Modelling1742-4682BioMed Central London 1742-4682-2-361614455410.1186/1742-4682-2-36ResearchBoundary effects influence velocity of transverse propagation of simulated cardiac action potentials Sperelakis Nicholas [email protected] Bijoy [email protected] Lakshminarayanan [email protected] Dept. of Molecular & Cellular Physiology, University of Cincinnati College of Medicine, Cincinnati, OH 45267-0576, USA2 Dept. of Electrical Computer Engineering and Computer Science, University of Cincinnati, College of Engineering, Cincinnati, OH 45221, USA2005 6 9 2005 2 36 36 18 7 2005 6 9 2005 Copyright © 2005 Sperelakis et al; licensee BioMed Central Ltd.2005Sperelakis et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We previously demonstrated that transverse propagation of excitation (cardiac action potentials simulated with PSpice) could occur in the absence of low-resistance connections (gap – junction channels) between parallel chains of myocardial cells. The transverse transmission of excitation between the chains was strongly dependent on the longitudinal resistance of the interstitial fluid space between the chains: the higher this resistance, the closer the packing of the parallel chains within the bundle. The earlier experiments were carried out with 2-dimensional sheets of cells: 2 × 3, 3 × 4, and 5 × 5 models (where the first number is the number of parallel chains and the second is the number of cells in each chain). The purpose of the present study was to enlarge the model size to 7 × 7, thus enabling the transverse velocities to be compared in models of different sizes (where all circuit parameters are identical in all models). This procedure should enable the significance of the role of edge (boundary) effects in transverse propagation to be determined.
Results
It was found that transverse velocity increased with increase in model size. This held true whether stimulation was applied to the entire first chain of cells or only to the first cell of the first chain. It also held true for retrograde propagation (stimulation of the last chain). The transverse resistance at the two ends of the bundle had almost no effect on transverse velocity until it was increased to very high values (e.g., 100 or 1,000 megohms).
Conclusion
Because the larger the model size, the smaller the relative edge area, we conclude that the edge effects slow the transverse velocity.
Propagation of cardiac action potentialtransverse propagation velocityPSpice simulationsedge/boundary effectselectric field transmission of excitation.
==== Body
Introduction
Computer simulation of the propagation of impulses in cardiac muscle shows that the electric field generated in the narrow junctional clefts when an action potential occurs at the prejunctional membrane depolarizes the postjunctional membrane to threshold [1]. Thus, the postjunctional cell is excited after a brief delay at the junction and propagation in cardiac muscle is saltatory. We have modeled APs in this tissue using the PSpice program for circuit design and analysis, and we have corroborated earlier reports that the EF developed in the junctional cleft is sufficiently large to allow transfer of excitation to the contiguous cell without the requirement for a gap-junction [2-6]. To date, however, we have only used small-sized models for these simulation studies.
When our paper on transverse propagation of cardiac action potential (APs) simulated by PSpice in a 5 × 5 model [4,5] was reviewed by the journal, one unanswered question was whether edge (boundary) effects were important. The purpose of the present study was to address this question. To do this, we expanded the model to a 7 × 7 size (7 parallel chains of 7 cells each). Thus, we could compare transverse velocity in 2-dimensional models of 4 sizes: 7 × 7, 5 × 5, 3 × 4, and 2 × 3. It was essential that all circuit parameters were the same in all four models. It was found that the larger the model, the faster the transverse velocity of propagation, up to a presumed saturation point.
Methods
The detailed methods and circuit parameters used for cardiac muscle were described previously [2,4,5]. As shown in Figure 1 (7 × 7 model), there were two surface membrane units in each cell (one facing upwards and one inverted) and one unit for each junctional membrane (intercalated disk). The values for the circuit parameters used (standard conditions) are listed in Table 1 (footnote) for both the surface units and the junctional units. Under standard conditions, Rol2 was 500 KΩ, Ror2 was 100 Ω, and Rjc was 25 MΩ (50 MΩ ÷ 2). The Rol2/ Ror2 ratio of 5000 was calculated from the equation relating absolute resistance to the resistivity of the interstitial fluid (ρ) (50 Ω – cm) and the distance (L) and cross-sectional area (Ax);
Table 1 Transverse Propagation Velocity (antegrade (A) and retrograde(R)) of Simulated Cardiac Action Potentials in 2-D Sheets at a Rol2 of 500 KΩ.
Model Size Stimulations No. of Chains Responding TPT ms Transv. Velocity cm/sec
7 × 7 A Entire A Chain 7 1.2 8.0
Cell A1 Only 7 1.5 6.4
R Entire G Chain 7 1.2 8.0
Cell G1 Only 6 (A failed) 1.5 5.4
5 × 5 A Entire A Chain 5 1.6 4.0
Cell A1 Only 5 1.7 3.8
R Entire E Chain 5 1.7 3.8
Cell E1 Only 5 1.8 3.6
3 × 4 A Entire A Chain 3 1.0 3.2
Cell A1 Only 3 1.1 2.9
R Entire C Chain 3 1.2 2.7
Cell C1 Only 3 1.2 2.7
2 × 3 A Entire A Chain 2 0.7 2.3
Cell A1 Only 2 0.8 2.0
R Entire B Chain 2 0.9 1.8
Cell B1 Only 2 0.9 1.8
A = antegrade R = retrograde
Circuit Parameters: RBT = 10 KΩ Rjc = 25 MΩ (50 MΩ/2); Cj = 16 pF; Cs = 0.2 pF
To improve the performance of the 7 × 7 model (for Rol2 of 500 KΩ), Rjc was decreased slightly to 24.5 MΩ (49.0 MΩ/2).
Figure 1 7 × 7 Model for Cardiac Muscle: Block diagram of the 7 × 7 model for cardiac muscle. These were 7 parallel chains (A-G) of 7 cells each (1–7). The cells longitudinally were separated by high-resistance cell junctions, with a radial junctional cleft resistance (Rjc) of 25 MΩ (50 MΩ/2). The parallel chains were separate and also not connected by gap-junction channels. The longitudinal resistance of the interstitial space between the parallel chains (Rol2) had values of 200 KΩ and 500 KΩ. Both ends (termination) of the tissue bundles were connected by transverse resistances (RBT); the value was 1.0 KΩ, but much higher values were tested. There was only little effect of varying RBT over a wide range, until very high values of 500 MΩ were inserted. As shown, there were 4 basic units for each cell: two for the surface membrane (one facing upwards and one downwards) and one for each of the two junctional membranes. The resistive and capacitive elements of the surface membrane and junctional membranes were prorated based on the relative areas. Electrical stimulation (rectangular current pulse of 0.5 nA and 0.5 ms) was applied intracellularly, either to the entire chain (A or G) or to the first cell only of these chains (cell A1 or cell G1).
The myocardial cell was assumed to be a cylinder 150 μm long and 16 μm in diameter. The cell capacitance was assumed to be 100 pF, and the input resistance to be 20 MΩ. A junctional tortuosity (interdigitation) factor of 4 was assumed for the cell junction [1,2]. The junctional cleft potential (Vjc) is produced across Rjc, the radial resistance of the narrow and tortuous junctional cleft. The junctional cleft contained two longitudinal resistances of 7Ω each and two radial resistances (Rjc) of 50 MΩ each in parallel.
The tortuosity factor does not interact with the packing factor. The tortuosity factor concerns the complex interdigitation of contiguous cells longitudinally (end-to-end), whereas the packing factor deals with how closely the cell chains are packed transversely (or radially) within a tissue bundle. The value assigned to Rol2 reflects the closeness of this packing. The value assigned to Rjc reflects the thickness of the junctional gap (end-to-end) and the tortuosity factor.
The circuit used for each unit was kept as simple as possible, using only those ion channels that set the resting potential (RP) and predominate during the rising phase of the AP. We wanted only to inscribe the rising phase of the APs to study propagation in the 2-dimensional sheet. The RP was -80 mV and the overshoot potential was +30 mV (AP amplitude of 110 mV). Transverse propagation velocity was calculated from the measured total propagation time (TPT) (measured as the difference between when the APs of the first cell and last cell crossed -20 mV) and cell width (number of chains minus one gives the number of transverse junctions traversed).
Because the PSpice program does not have a V-dependent resistance to represent the increase in conductance for Na+ ions in myocardial cells during depolarization and excitation, this function was simulated by a V-controlled current source (our "black-box") in each of the basic circuit units. The current output of the black-box at various membrane voltages was calculated assuming a sigmoidal relationship between membrane voltage and resistance over the range of -60 mV to -30 mV. The V values used in the GTABLE were those recorded directly across the membrane. The excitability of the basic units was the same as in our previous papers, i.e., it was set at the moderate level [6].
The upper chain of cells was assumed to be bathed in a large volume of Ringer solution connected to ground. The external resistance (Ro) of this fluid was divided into two components: a radial resistance (Ror) and a longitudinal resistance (Rol). The longitudinal resistance value between the chains (Rol2) was increased over a wide range to reflect closer packing of parallel chains into a bundle of fibers (Fig. 1). The transverse resistance of the interstitial fluid space (Ror2) was found to have almost no effect on the transverse velocity. The cells in each chain were not interconnected by low-resistance pathways (gap-junction channels), so that transmission of excitation from one cell to the next had to be by the electric field (EF) developed in the narrow junctional cleft. In our previous papers, we presented a number of references demonstrating that propagation velocity is slowed only slightly in the absence or paucity of gap junctions [e.g., see refs [1,3] and [7]]. There were seven parallel chains (chains A-G) of seven cells each in the 7 × 7 model. The block diagrams and detailed circuitry for the other models (5 × 5; 3 × 4; 2 × 3) were previously published [4,5]. The ends of each chain had a bundle termination resistance (RBT) of 1.0 KΩ to mimic the physiological condition. However, variation of RBT over a wide range had almost no effect, until very high values of about 500 MΩ were inserted.
Electrical stimulation (rectangular current pulses of 0.50 nA and 0.50 ms duration) was applied to the inside of either the first cell of chain A (cell A1) or simultaneously to all cells of the A-chain. For retrograde propagation, stimulation was applied either to cell G1 or to all cells of the G-chain. For some measurements, the V-recording markers were placed on only one chain at a time. To minimize confusion, the voltage was recorded from only one surface unit (upward-facing) in each cell.
Results
The results to be illustrated here will be from the 7 × 7 model only, because this model is new. However, the results from the smaller models (5 × 5, 3 × 4, 2 × 3), previously published, are summarized in Table 1. Thus, Table 1 enables transverse propagation velocities to be compared in four models differing in size but with identical circuit parameters used in the basic units. Through this comparison, it can be ascertained whether edge (boundary) effects are important in transverse propagation.
In the 7 × 7 model, with all circuit parameters having the standard values, including Rol2 of 500 KΩ, either the entire A-chain was stimulated simultaneously (Fig. 2A) or only cell A1 was stimulated (Fig. 2B) (for antegrade propagation). Since the circuit was symmetric, the terms "antegrade" and "retrograde" are arbitrary, and are used simply to denote direction of propagation. For retrograde propagation, the entire G-chain was simultaneously stimulated (Fig. 2C) or only cell G1 was stimulated (Fig.2D). As can be seen, the A-chain failed in the retrograde (antidromic) direction (Fig. 2D) when a single cell was stimulated. However, there were no failures when the entire G-chain was stimulated (Fig. 2C). Increasing Rol2 caused fewer chains to fail, and propagation velocity was increased substantially (TPT decreased). Thus, retrograde propagation is not always identical to the antegrade propagation, though it is always very close. Since the PSpice program generates a netlist error indicating the presence of any floating node, we suggest that the aberrant retrograde propagation behavior is a limitation in the PSpice computational algorithm rather than a property of the model. Activation maps would have revealed the patterns in more detail, but the software for obtaining such maps was not available when these experiments were performed.
Figure 2 Rising phases of the simulated APs recorded from the 7 × 7 model for cardiac muscle when Rol2 was 500 KΩ. A-B: Antegrade propagation. A: Stimulation of the entire A chain. No chains failed, and TPT was short. Many traces overlap. B: Stimulation of only cell A1. Again, there were no failures. TPT was prolonged (compare to Panel A). C-D: Retrograde propagation. C: Stimulation of entire G-chain. No failures occurred. TPT was about the same as in panel A (for orthodromic). D: Stimulation of only cell G1. The last chain (A) failed.
Table 1 summarizes all these data, not only for the 7 × 7 model, but also for the smaller models of 5 × 5, 3 × 4, and 2 × 3. These data include antegrade (A) and retrograde (R) propagation, with stimulation of the entire chain or single cell only, for the Rol2 value of 500 KΩ. As can be seen, the calculated transverse propagation velocities were highest in the large 7 × 7 model and slower in the smaller models. This was true for all values of Rol2. Transverse propagation velocities were faster at Rol2 of 500 KΩ than at 200 KΩ.
To help clarify how propagation spreads through the 7 × 7 model, recordings were made from one chain at a time when stimulation was applied to cell A1 (Fig. 3). Records from selected chains are illustrated to allow appreciation of the time sequence of firing of the various chains. The stimulated chain always begins to respond first, but there is some overlap of firing from the adjacent chain. Thus, transverse propagation between chains occurs simultaneously with longitudinal propagation within each chain when only one cell is stimulated. When the middle chain (D-chain) of the network was stimulated, transverse spread of excitation occurred simultaneously in both directions (not illustrated). Transverse spread occurs at multiple points along the length of the chain.
Figure 3 Recordings of the APs from one chain at a time, so that the transverse spread of excitation can be more clearly seen. 7 × 7 model of cardiac muscle. Standard conditions for all circuit parameters; Rol2 was 500 KΩ. Stimulation was applied to cell A1 (first cell of A-chain). All 49 cells responded. To reduce the number of panels, records from every other chain are illustrated. A: Records from the A- chain. B: Records from the C- chain. C: Records from the E- chain. D: Records from the G- chain. As can be seen, the stimulated A-chain responded earlier, followed by C-, E-, and G-chains. But there was some overlap between the traces from the various chains, indicating that transverse propagation between chains occurs simultaneously with longitudinal propagation within each chain.
Discussion
The present results, comparing the velocities of transverse propagation (θtr) in cardiac models differing in size but with identical circuit parameters, demonstrate that edge/ boundary effects have a strong action on transverse velocity (Fig. 4). θtr was slowest in the smaller models and fastest in the larger models. In our new large 7 × 7 model, θtr was about double the value in the 5 × 5 model (at Rol2 of 500 KΩ) (Table 1). Since the larger the model, the less the relative area of edges and the faster the propagation velocity, this means that edges must slow down θtr.
Figure 4 Plot of transverse velocity of propagation of the simulated cardiac action potentials as a function of the ratio of relative edge area to interior area (see Table 2).
In fact, θtr is almost inversely proportional to the ratio of edge to interior areas for the four models compared in the present study (Table 2). This table relates the ratio of velocities to the inverse ratio of the relative edge area (or volume), using the 7 × 7 model as the base for comparison (A/Y for velocity and Y/A for relative edge area). The comparisons are: 1.38 vs 1.40; 1.88 vs 2.05; and 2.47 vs 2.93 (for Rol2200 KΩ). These comparisons are strikingly close. Comparisons for a Rol2 of 500 KΩ were also close: 2.00 vs 1.40, 2.50 vs 2.05, and 3.48 vs 2.93 (Table 1). The data for Rol2 of 200 KΩ are plotted in Figure 4.
Table 2 Comparison of the inverse ratios of the relative edge areas of the various-sized cardiac models with the ratio of the transverse propagation velocities (θtr)
Model Size Ratio of relative edge area to interior area θtr cm/sec Velocity A/Y Area Y/A
A 7 × 7 28/49 = 0.57 4.7 -- --
B 5 × 5 20/25 = 0.80 3.4 1.38 1.40
C 3 × 4 14/12 = 1.17 2.5 1.88 2.05
D 2 × 3 10/6 = 1.67 1.9 2.47 2.93
Orthodromic direction; stim of entire A-chain; values for Rol2 of 200 KΩ.
Y equals value of B, C, or D.
The equation relating these parameters is:
Where a single asterisk superscript (*) denotes the values for the smaller model compared to the referral 7 × 7 model (denoted by double asterisk**).
This means that one can predict the transverse propagation velocities in yet-larger models. For example, in a 10 × 10 model, θtr should be approximately 6.70 cm/s (if compared with the 7 × 7 model) or 6.80 cm/s (if compared with the 5 × 5 model). From the equation given in the footnote of Table 2:
Hence, the two calculations are in close agreement. However, this relationship between transverse velocity and the inverse of the relative edge area probably saturates and levels off at some point, i.e., a maximum θtr is reached. In the intact heart, the velocity of transverse propagation is difficult to measure accurately because of the complicated geometry of bundles, but estimates that θtr is about 1/5th to 1/10th that of θlo (longitudinal velocity) have been given (see references given in ref 1). If θlo is taken to be 0.40 m/s, then θtrshould be between 4.0 and 8.0 cm/s. Thus, the values calculated in the present simulations are in good agreement with physiological measurements.
The ratio of propagation velocities, longitudinal (θlo) to transverse (θtr), is almost what is expected based on the cell geometry (cylinder 150 μm long and 16 μm wide). These dimensions would predict a θlo / θtr ratio of 9.4 (150/16), provided that the longitudinal and transverse transfer function are equal (i.e, the delay time at the two types of junctions were equal). If the average cell length were taken to be only 100 μm, then the θlo / θtr ratio would be 6.3. Thus, for a θlo value of 40 cm/s and a θlo / θtr ratio of 7.9 (average of 9.4 and 6.3), then θtr would be 5.1 cm/s, which is close to the value of 4.7 cm/s measured in the 7 × 7 model (for Rol2 of 200 KΩ). However, it has been reported that the anisotropic conduction velocity observed in the heart is not a result of cell geometry [8].
Another observation in the large 7 × 7 model is that some chains distal to the point of stimulation failed when Rol2 was only 200 KΩ. Such failures did not occur when the model was smaller (e.g., 5 × 5). Failure of distal chains occurred in both the orthodromic and antidromic directions, but was greater in the antidromic direction. However, increasing Rol2 to 500 KΩ allowed all chains to respond, with the exception of failure of one chain (the most distal A-chain) in the retrograde direction (Table 1). Therefore, in the largest model, there is an increase in probability of failure of one or more distal chains.
Although we don't know the mechanism for this effect, we may speculate about two possibilities. First, if some current leaked out at the ends of each chain, then less current would be available for downstream depolarization. Second, if the phenomenon of reflection occurred at the longitudinal edge of the last chain (G), then this would act to slow the transverse velocity. Thus, both of these mechanisms may be involved in explaining why transverse propagation was faster in the larger models.
In summary, the present results using our enlarged 7 × 7 model for cardiac muscle, with comparisons with our prior smaller models, demonstrate that edge effects are important in determining the transverse velocity of propagation, when all circuit parameters are identical. θtr increased with the inverse of the ratio of the relative edge areas in the various-sized models. This relationship likely levels off at some point, such that a maximum velocity is reached. The transverse velocities measured in the largest model (7 × 7), and estimated for a 10 × 10 model, give values in the same range as the physiological values.
==== Refs
Sperelakis N McConnell K Electric field interactions between closely abutting excitable cells IEEE-Eng Med Biol 2002 21 77 89 10.1109/51.993199
Sperelakis N Ramasamy L Propagation in cardiac muscle and smooth muscle based on electric field transmission at cell junctions: An analysis by PSpice IEEE-Eng Med Biol 2002 21 130 143 10.1109/MEMB.2002.1175149
Sperelakis N Murali KPV Combined electric field and gap junctions on propagation of action potentials in cardiac muscle and smooth muscle in PSpice simulation J Electrocardiol 2003 36 279 293 14661164 10.1016/j.jelectrocard.2003.08.001
Sperelakis N Propagation of action potentials between parallel chains of cardiac muscle cells in PSpice simulation Can J Physiol Pharmacol 2003 81 1 112 12665251 10.1139/y03-019
Sperelakis N Kalloor B Transverse propagation of action potentials between parallel chains of cardiac muscle and smooth muscle cells in PSpice simulations Biomed Eng Online 2004 3 5 14998434 10.1186/1475-925X-3-5
Sperelakis N Kalloor B Effect of variation in membrane excitability on propagation velocity of simulated action potentials for cardiac muscle and smooth muscle in the electric field model for cell to cell transmission of excitation IEEE-Eng Med Biol 2004 51 2216 2219 10.1109/TBME.2004.836528
Thomas SP Kucera JP Bircher-Lehmann L Rudy Y Saffitz JE Kleber AG Impulse propagation in synthetic strands of neonatal cardiac myocytes with genetically reduced levels of connexin43 Circ Res 2003 91 1209 1216 12730095 10.1161/01.RES.0000074916.41221.EA
Darrow BJ Fast VG Kleber AG Beyer EC Saffitz JE Functional and structural assessment of intercellular communication. Increased conduction velocity and enhanced connexin expression in dibutyryl cAMP-treated cultured cardiac myocytes Circ Res 1996 79 174 183 8755993
|
16144554
|
PMC1236963
|
CC BY
|
2021-01-04 16:39:26
|
no
|
Theor Biol Med Model. 2005 Sep 6; 2:36
|
utf-8
|
Theor Biol Med Model
| 2,005 |
10.1186/1742-4682-2-36
|
oa_comm
|
==== Front
Theor Biol Med ModelTheoretical Biology & Medical Modelling1742-4682BioMed Central London 1742-4682-2-371614656810.1186/1742-4682-2-37ResearchA comparative study of a theoretical neural net model with MEG data from epileptic patients and normal individuals Kotini A [email protected] P [email protected] AN [email protected] D [email protected] Laboratory of Medical Physics, Medical School, Democritus University of Thrace, University Campus, Alex/polis, 68100, Greece2 General Hospital of Chania, Crete, Greece2005 7 9 2005 2 37 37 27 4 2005 7 9 2005 Copyright © 2005 Kotini et al; licensee BioMed Central Ltd.2005Kotini et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Objective
The aim of this study was to compare a theoretical neural net model with MEG data from epileptic patients and normal individuals.
Methods
Our experimental study population included 10 epilepsy sufferers and 10 healthy subjects. The recordings were obtained with a one-channel biomagnetometer SQUID in a magnetically shielded room.
Results
Using the method of x2-fitting it was found that the MEG amplitudes in epileptic patients and normal subjects had Poisson and Gauss distributions respectively. The Poisson connectivity derived from the theoretical neural model represents the state of epilepsy, whereas the Gauss connectivity represents normal behavior. The MEG data obtained from epileptic areas had higher amplitudes than the MEG from normal regions and were comparable with the theoretical magnetic fields from Poisson and Gauss distributions. Furthermore, the magnetic field derived from the theoretical model had amplitudes in the same order as the recorded MEG from the 20 participants.
Conclusion
The approximation of the theoretical neural net model with real MEG data provides information about the structure of the brain function in epileptic and normal states encouraging further studies to be conducted.
Poisson distributionGauss distributionMEG
==== Body
Introduction
Epilepsy is a disorder involving recurrent unprovoked seizures: episodes of abnormally synchronized and high-frequency firing of neurons in the brain that result in abnormal behaviors or experiences. This is a fairly common disorder, affecting close to 1% of the population. The lifetime risk of having a seizure is even higher, with estimates ranging from 10 to 15% of the population. Epilepsy can be caused by genetic, structural, metabolic or other abnormalities. Epileptic disorders can be generalized, partial (focal) or undetermined. A primary generalized seizure starts as a disturbance in both hemispheres synchronously, without evidence of a localized onset. Partial forms of epilepsy start in a focal area of the brain and may remain localized without alteration of consciousness.
MEG is a noninvasive imaging technique, applicable to the human brain with temporal resolution approximately ~1 ms [1]. Several authors during the last decade have demonstrated the importance of MEG in the investigation of normal and pathological brain conditions [2,3]. The major advantage of MEG over electroencephalography (EEG) is that MEG has higher localization accuracy. This is because the different structures of the head (brain, liquor cerebrospinalis, skull and scalp) influence the magnetic fields less than the volume current flow that causes the EEG. Also, MEG is reference free, so that the localization of sources with a given precision is easier for MEG than it is for EEG [4].
The goal of this study is to compare the theoretical model that follows Poisson or Gauss distributed connectivity [5-12] with experimental MEG data from epileptic patients and healthy volunteers.
Methods
Description of the model
Neural nets are assumed to be constructed of discrete sets of randomly interconnected neurons of similar structure and function. The neural connections are set up by means of chemical markers carried by the individual cells. Thus, the neural population of the net is treated as a set of subpopulations of neurons, each of them characterized by a specific chemical marker. We attribute the appropriate Poisson or Gauss distribution law to each subsystem to describe connectivity.
The elementary unit, the neuron, is bistable. It can be either in the resting or in an active (firing) state. The transition from the resting to the firing state occurs when the sum of postsynaptic potentials (PSPs) arriving at the cell exceeds the firing threshold θ of the neuron. PSPs may be excitatory (EPSPs) or inhibitory (IPSPs), shifting the membrane potential closer to or further away from θ, respectively. Each neuron may carry an electrical potential of a few millivolts, which it passes on to the neurons to which it is connected.
In this model, a net with N markers is assumed to be constructed of A formal neurons. A fraction h (0<h<l) of these are inhibitory with all the axon branches generating IPSPs, while the rest are excitatory with all their axon branches generating EPSPs. Each neuron receives, on average, μ+ EPSPs and μ- IPSPs. The size of the PSP produced by an excitatory (inhibitory) unit is K+ (K-). The neurons are also characterized by the absolute refractory period and the synaptic delay τ. If a neuron fires at time t, it produces the appropriate PSP after a fixed time interval τ, the synaptic delay. PSPs arriving at a neuron are summed instantly, and if this sum is greater or equal to θ, then the neuron will fire immediately, otherwise it will be idle. PSPs (if below θ) will persist with or without decrement for a period called the summation time, which is assumed to be less than the synaptic delay. Firing is momentary and causes the neuron to be insensitive to further stimulation for a time interval called the (absolute) refractory period [5-12].
The mathematical formalism of this study is based on the equations for the expectation values of the activity derived in previous papers [5-12]. A brief mathematical analysis for each case is given below.
a) Expectation value of neural activity in noiseless and noisy neural nets with Poisson distributed connectivities
Following the assumptions of previous papers it was shown that the expectation value of the neural activity <αn+1> at t = (n+1) τ, i.e. the average value of αn+1 generated by a collection of netlets with identical statistical parameters (μ+, μ-, h, K+, K-, A, θ) and the same αn at t = nτ, is given by:
<αn+1> = (1-αn) P (αn, θ) (1)
where P(αn, θ) is the probability that a neuron receives post synaptic potentials (PSPs) exceeding its threshold at time t = (n+1)τ. Thus:
Here Pl and Qm are the probabilities that a neuron will receive l and m EPSPs and IPSPs respectively, and are given by (3):
PI = exp (-αn (1-h) μ+) (αn (1-h) μ+)l/l!
Qm = exp (-αn h μ-) (αn hμ-)m/m! (3)
In addition, the upper limits in the double sum mmax and lmax are given by (4):
lmax = A αn (1-h) μ+
mmax = Aαnhμ- (4)
Taking into account equations (2) and (3), equation (1) takes the form:
Similarly for Poisson nets with noise: if Pl and Qm are the probabilities that a given neuron receives I EPSPs and m IPSPs at time t = (n+1)τ, they are given by equation (3). But if Tδ (θ) is the probability that the instantaneous threshold value is θ or less than θ, this is given by (6):
Therefore the firing probability per neuron is then given by (7):
where lmax and mmax are given by equation (4).
Finally, the expectation value of the activity is given by (8):
<αn+1> = (1-αn) P (αn, θ) (8)
b) Expectation value of neural activity in neural nets with Poisson distributed connectivities with chemical markers and noise
Similarly, the expectation value of the activity <αn+1> for an isolated neural net with two chemical markers a and b is given by (9):
where PI, Qi, P'l, Q'i', are the probabilities that a given neuron will receive l EPSPs, i IPSPs or l'-EPSPs, i'-IPSPs, at time t = (n+1)τ in the subsystems a or b respectively. These probabilities are given by (10):
Pl = exp (-αn μa+ (1-ha) ma) (-αn μa+ (1-ha) ma)l/l!
Qi = exp (-αn μa- ha ma) (-αn μa- ha ma)i/i!
P'l' = exp (-αn μb+ (1-hb) (1-ma)) (-αn μb+ (1-hb) (1-ma))l'/l'!
Q'i' = exp (-αn μb- hb (1-ma)) (-αn μb- hb (1-ma))i'/i'! (10)
The upper limits in the sums in equation (9) are given by (11):
lmax = A αn μa+ (1-ha) ma
lmax' = A αn μb+ (1-hb) (1-ma)
imax = A αn μa- ha ma
imax' = A αn μb- hb (1-ma) (11)
Finally, (θa) and (θb) are defined as the probabilities that the instantaneous neural thresholds are equal to or less than θa and θb in subsystems a and b respectively and are given by (12):
b) Expectation value of neural activity in neural nets with Gaussian connectivities in the absence of chemical markers
Let the total PSP of a neuron at t = (n+1)τ be given by:
en+1 = lK+ + mK- (13)
where l and m are the numbers of EPSPs and IPSPs respectively. If both l and m are large, their distributions may be approximated by Gaussian distributions about their respective average values and . The distribution of en+1 is therefore also normal, since the probabilities for l and m are mutually independent, and its variance is the sum of the variances of l and m. Therefore the average PSP will be given by (14):
where K = [μ+ (1-h) K+ + μ-h K-] (14)
The variance of en+1, call it , is then given by (15):
= αn [μ+ (1-h) (K+)2 + μ-h (K-)2] (15)
The probability that the PSP exceeds a threshold now becomes:
Equation (16) in conjunction with equation (1) gives values for <αn+1> at t = (n+1)τ.
Let T(θ') be the probability that the instantaneous threshold of a neuron is θ' or less than θ'. This is given by (17):
Here δ is the standard deviation of the Gaussian distribution of the noise. Finally, the probability that a neuron will receive PSPs that will exceed the threshold at time t = (n+1)τ is given by (18):
Since l and m are very large numbers, the double sum can be approximated by and therefore:
Then the expectation value of <αn+1> of the activity at time t = (n+1)τ will be:
<αn+1> = (1-αn) P(αn, δn+1, δ) (20)
c) Expectation value of neural activity in noisy neural nets with chemical markers and Gaussian distributed connectivities
In a neural netlet of A neurons with two chemical markers a and b, let the fractional numbers corresponding to each chemical marker be ma and mb, and the fractions of inhibitory neurons for each chemical marker be ha and hb, respectively. Also, let αnA be the active neurons in the netlet at t = nτ. Then at t = (n+1)τ the numbers of EPSPs and IPSPs that will appear in the subsystems with a and b markers will be:
la = A αn μa+ (1-ha) ma
ia = A αn μa- ha ma
lb = A αn μb+ (1-hb) mb
ib = A αn μb- hb mb (21)
On the average, the numbers of EPSPs and IPSPs that appear per neuron in subnets with a and b markers will be:
= αn μa+ (1-ha) ma
= αn μa- ha ma
= αn μb+ (1-hb) mb
= αn μb- hb mb (22)
As stated in our previous papers [5-12] the total PSP input to a neuron with a and b markers at t = (n+1)τ will be given by (23):
ea,n+1 = laK+ + iaK-
eb,n+1 = lbK+ + ibK- (23)
(Here it is assumed that )
If the quantities la, lb, ia and ib are sufficiently large, their distributions may be approximated by Gaussian distributions about their average values, given by (22). Then the average PSPs for the two markers a and b will be given by (24):
and their variances will be given by (25):
Therefore the probability that a neuron with marker a or b will receive a certain number of EPSPs or IPSPs that will shift the membrane potential closer to or further away from the instantaneous threshold will be given by (26):
where:
Thus, the probabilities and that the instantaneous threshold of a neuron in subsystems a and b is equal to or less than or will be given by (28):
Consequently, as stated in our previous paper [8], the firing probabilities P(αn, δn+1, δa) and P'(αn, δn+1, δb) that a neuron in subpopulations a and b, respectively, will receive PSPs exceeding threshold at time t = (n+1)τ will be given by (29):
Since the quantities la, ia, lb and ib are sufficiently large, the double sum in equations (29) will be substituted by the probabilities of the average values of la, ia and lb, ib for each marker a and b and will be given by (30):
Then according to our previous papers [5-12], the expectation value of activity in this netlet with two markers a and b at time t = (n+1)τ will be given by (31):
The general case for an isolated noisy net with N markers m1, m2,..., mN, where mi is the fraction of neurons with the ith marker, is described by an equation analogous to the equation for two markers (31). This general equation for such a netlet at time t = (n+1)τ is:
Theoretical analysis
The electromagnetic fields generated in neural networks with Poisson or Gauss connectivities
Let us consider an isolated neural network with structural parameters A, μ+, μ- and h, and initial activity αn at time t = nτ. The potential generated in this network due to this initial activity will be equal to the summation of all the PSPs [7] and will be given by (33):
Vn = αn (A μ+ (1-h) - A μ-h) (33)
Similarly, the potential generated by the neural activity αn+1 at the next time interval t = (n+1)τ will be given by (34):
Vn+1 = αn+1 (A μ+ (1-h) - A μ-h) (34)
By combining equations (33) and (34) and assuming spherical brain symmetry, the potential difference ΔV can be obtained. As is known from classical physics, this generates a magnetic field Bn given by (35):
Choosing Δt = 1 ms, the above equation takes the following form:
where μo and εo are the magnetic permeability and dielectric constant of the medium respectively.
When the neural network is characterized by two chemical markers a and b, the potentials created at the synapses of the neurons with the a and b markers will be given by (37):
Vna = αn (A μa+ (1-ha) ma - A μa- ha ma)
Vnb = αn (A μb+ (1-hb) mb - A μb- hb mb) (37)
On the other hand, the total voltages created at the synapses of the neurons at times t = nτ and t = (n+1)τ will be given by (38):
Vn = Vna + Vnb = αn A [(μa+ (1-ha) ma + μb+ (1-hb) mb) - (μa- ha ma + μb- hb mb)]
Vn+1 = αn+1 A [(μa+ (1-ha) ma + μb+ (1-hb) mb) - (μa- ha ma + μb- hb mb)] (38)
Therefore the potential difference between these two time intervals, taking into account equations (38), is given by (39):
ΔV = Vn+1 - Vn = (αn+1 - αn) A [(μa+ (1-ha) ma + μb+ (1-hb) mb) - (μa- ha ma + μb- hb mb)] (39)
Thus, as stated previously, this potential difference will create a magnetic field Bn, which is given by (40):
where the neural activity αn+1 refers to a Poisson or Gauss distribution of connectivities as given in the previous section.
In the general case, where the neural net has N chemical markers, equation (40) takes the form:
Experimental procedure
We compared the theoretical results with the experimental findings obtained using MEG measurements from 10 epileptic patients and 10 healthy volunteers. Informed consent for the methodology and the aim of the study was obtained from all participants prior to the procedure.
Biomagnetic measurements were performed using a second order gradiometer SQUID (Model 601, Biomagnetic Technologies Inc.), which was located in a magnetically shielded room with low magnetic noise. The MEG recordings were performed after positioning the SQUID sensor 3 mm above the scalp of each patient using a reference system. This system is based on the International 10–20 Electrode Placement System [13] and uses any one of the standard EEG recording positions as its origin; in this study we used the P3, P4, T3, T4, F3, and F4 recording positions [14-16]. Around the origin (T3 or T4 for temporal lobes) a rectangular 32-point matrix was used (4 rows × 8 columns, equidistantly spaced in a 4.5 cm × 10.5 cm rectangle) for positioning of the SQUID [14-16]. The MEG was recorded from each temporal lobe at each of the 32 matrix points of the scalp for 32 s and was band-pass filtered with cut-off frequencies of 0.1 and 60 Hz. The MEG recordings were digitized using a 12 bit precision analog-to-digital converter with a sampling frequency of 256 Hz, and were stored in a PC peripheral memory for off-line Fourier statistical analysis. The method, by its nature (i.e. temporal and spatial averaging), eliminates short-term abnormal artifacts in any cortical area, while it retains long-lasting localized activation phenomena. We used the x2 – fitting method to analyze the MEG data [17].
This method was based on the following equation (42):
where:
Qi: is the number of elements in the ith interval of the normalized MEG histogram
Ti: is the number of elements in the ith interval of the normal distribution with the same mean value and standard deviation as the normalized MEG histogram
k: is the number of intervals
n = k-1: the degrees of freedom of the system
In our case n = 7 and the critical value for distinguishing the Poisson from the Gauss distribution was 14.1 (xcr2 = 14.1). If the estimated value of the x2 was greater than 14.1, the distribution was Poisson; otherwise it was Gauss.
Results
Using the x2-fitting method it was found that the MEG recordings from epileptic patients had Poisson distributions whereas those from normal subjects had Gauss distributions. The Poisson connectivity derived from the theoretical model represents the state of epilepsy, whereas the Gauss connectivity represents normal behavior. The magnetic field derived from the theoretical model was approximately in the same order as the recorded MEG in both conditions. Furthermore, the MEG data obtained from epileptic areas had higher amplitudes than those from normal regions and were comparable with the theoretical magnetic fields from Poisson and Gauss distributions.
Figure 1 shows the MEG recorded from an epileptic patient; figure 2 illustrates the MEG recorded from a healthy volunteer.
Figure 1 The MEG recorded from an epileptic patient over an interval of 1 s duration. The x-axis represents the time sequence and the y-axis the magnetic field.
Figure 2 The MEG recorded from a healthy subject over an interval of 1 s duration. The x-axis represents the time sequence and the y-axis the magnetic field.
Figures 3 and 4 show the magnetic fields derived from the theoretical model with Poisson and Gauss distributions respectively.
Figure 3 The magnetic field derived from the theoretical model with Poisson distribution. The x-axis represents the time sequence and the y-axis the magnetic field. Parameters: ma = 0.6, θa = 5, = 15, ha = 0; mb = 0.2, θb = 4, = 192, hb = 0.01; mc = 0.1, θc = 3, = 34, hc = 0.01; md = 0.1, θd = 3, = 32, hd = 0; K± = 1.
Figure 4 The magnetic field derived from the theoretical model with Gauss distribution. The x-axis represents the time sequence and the y-axis the magnetic field. Parameters: ma = 0.7, θa = 6, = 14, ha = 0; mb = 0.08, θb = 4, = 240, hb = 0.02; mc = 0.02, θc = 4, = 400, hc = 0.0; md = 0.1, θd = 4, = 337, hd = 0; me = 0.1, θe = 4, = 294, he = 0.03; K± = 1.
Discussion
Over the past three decades, neural nets have been intensively studied from several points of view. An area of considerable importance is that of biological nets, i.e. models of nets designed to imitate the structures and functions of human and other living brains and thus enhance our understanding of learning, memory, understanding etc. Widely used models include the pioneering work of McCulloch and Pitts [18], which treats assemblies of neurons as logical decision elements, the mathematical formalism of Caianiello [19] using the "neuronic equation", and probabilistic neural structures [5,6] that monitor the net activity, i.e. the fraction of neurons that become active per unit time. All these models have had a measure of success in improving our understanding of functions such as those mentioned above.
The effect of structure on function and dynamic behavior in neural nets has been also a subject of considerable interest in recent years. In the so-called probabilistic nets we have an assembly comprising a large number of neurons, randomly positioned in space, that have only partial connectivity; i.e. each neuron is connected to only a very small fraction of the total number of neurons in the system, randomly chosen. The principal idea is that this connectivity is given by the binomial distribution. In earlier work, probabilistic neural nets were investigated using Poisson or Gauss distributions of interneuronal connectivity; the main conclusion was that when a neuron was connected to a relatively small number of units, a Poisson distribution law was appropriate but if it was connected to a large number of units then a Gaussian law was a fairly good approximation [10-12]. Thus, Poisson neuronal nets may be viewed as approximately Gaussian whenever the number of synaptic connections is relative large.
In this study we measured the MEG of epileptic patients and normal subjects in order to compare the theoretical neural net model [10-12] with real data. Analyzing the MEG data by x2-fitting revealed that the MEG recordings from epileptic areas had Poisson distributions [17]. This finding is consistent with the correspondence between Poisson distributions and low numbers of internal neural connections, and with the synchronization of neural activity during an epileptic seizure [20,21]. Moreover, the MEG recordings from epileptic areas showed higher amplitudes than those from normal regions, comparable with the results from the theoretical neural model with Poisson and Gauss distributions respectively (Figs. 1, 2, 3, 4).
If a nerve cell is characterized by a given firing threshold which, when exceeded, results in spike discharge, two anatomical situations can be contrasted: one in which only a few synaptic contacts reach the cell in question, and a second in which the cell receives a large number of synaptic inputs. Suppose, in either case, that firing is dependent on the simultaneous excitation of a certain percentage of the total synaptic input (assuming that the ratio of excitatory and inhibitory synapses is the same in both situations so that the inhibitory inputs may be disregarded for the moment). Then it is clear that firing in neurons with a large number of synaptic inputs would require the synchronized activation of a substantial number of synapses; whereas in neurons with few synapses, firing may ensue even from a single excitatory synapse. Thus, a system in which neurons receive small numbers of synaptic connections is likely to exhibit a less "controlled" pattern of activity – and also "spontaneous" discharges [22]. The inverse problem in MEG measurements is the search for unknown sources by analysis of the measured field data. To handle this task one must first study the forward problem, i.e. how the magnetic field and the electrical potential arise from a known source. For practical purposes one also has to choose appropriate models for the source and the biological object as a conductor. Sarvas [23] described basic mathematical and physical concepts relevant to the forward and inverse problems and discussed some new approaches. Especially, he described the forward problem for both homogenous and inhomogenous media. He referred to the Geselowitz's formulae and presented a surface integral equation to handle a piecewise homogenous conductor and a horizontally layered medium. Furthermore, he discussed the non-uniqueness of the solution of the magnetic inverse problem and studied the difficulty caused by the contribution of the electric potential to the magnetic field outside the conductor.
The Poisson distribution corresponds to epileptic areas and the Gauss distribution to normal regions. The approximation of the theoretical neural net model to real MEG data provides a mathematical approach to the structure of brain function and indicates the need for further studies.
Appendix
The subscript i is a marker label and indicates the properties of a subpopulation in the netlet characterized by the ith marker.
Structural parameters of the neural net
τ Synaptic delay
A Total number of neurons in the netlet
hi Fraction of inhibitory neurons
The average number of neurons receiving excitatory postsynaptic potentials (EPSPs) from one excitatory neuron
The average number of neurons receiving inhibitory postsynaptic potentials (IPSPs) from one inhibitory neuron
The size of PSP produced by an excitatory neuron of the netlet
The size of PSP produced by an inhibitory neuron of the netlet
mi Fractions of neurons carrying the ith marker in the netlet
θi Firing thresholds of neurons
Statistical parameters
δi Standard deviation of the Gaussian distribution of the neural firing thresholds in the ith subpopulation
Dynamical parameters
n An integer giving the number of elapsed synaptic delays
αn The activity, i.e. the fractional number of active neurons in the netlet at time t = nτ
==== Refs
Mitra PP Pesaran B Analysis of dynamic brain imaging data Biophys J 1999 691 708 9929474
Timmermann L Gross J Dirks M Volkmann J Freund HJ Schnitzler A The cerebral oscillatory network of parkinsonian resting tremor Brain 2003 126 199 212 12477707 10.1093/brain/awg022
Volkmann J Joliot M Mogilner A Ioannides AA Lado F Fazzini E Ribary U Llinas R Central motor loop oscillations in parkinsonian resting tremor revealed by magnetoencephalography Neurol 1996 46 1359 1370
Kristeva-Feige R Rossi S Feige B Mergner T Lucking CH Rossini PM The bereitschaftspotential paradigm in investigating voluntary movement organization in humans using magnetoencephalography (MEG) Brain Res Protocol 1997 1 13 22 10.1016/S1385-299X(97)80327-3
Anninos PA Beek B Csermely TJ Harth EM Pertile G Dynamics of neural structures J Theor Biol 1970 26 121 148 5411107
Anninos PA Kokkinidis M A neural net model for multiple memory domains J Theor Biol 1984 109 95 110 6471872
Kokkinidis M Anninos PA Noisy neural nets exhibiting epileptic features J Theor Biol 1985 113 559 588 3999786
Kotini A Anninos PA Dynamics of noisy neural nets with chemical markers and Gauss-distributed connectivity Connect Sci 1997 9 381 403 10.1080/095400997116603
Fournou E Argyrakis P Kargas B Anninos PA Neural nets with chemical markers and firing threshold fluctuations Trends Biol Cybernet 1991 2 123 131
Fournou E Argyrakis P Anninos PA Neural nets with markers and Gauss-distributed connectivity Connect Sci 1993 5 77 94
Fournou E Argyrakis P Kargas B Anninos PA A Gaussian approach to neural nets with multiple memory domains Connect Sci 1995 7 331 339
Fournou E Argyrakis P Kargas B Anninos PA Hybrid neural nets with Poisson and Gaussian connectivity J Stat Phys 1997 89 847 867
Jasper HH The ten-twenty electrode system of the International Federation Electroenceph Clin Neurophysiol 1958 10 367 380
Anninos PA Anogianakis G Lehnertz K Pantev C Hoke M Biomagnetic measurements using SQUID Int J Neurosci 1987 37 149 168 3692698
Anninos PA Tsagas N Jacobson JI Kotini A The biological effects of magnetic stimulation in epileptic patients Panminerva Med 1999 41 207 215 10568117
Anninos P Adamopoulos A Kotini A Tsagas N Nonlinear Analysis of Brain Activity in Magnetic Influenced Parkinson Patients Brain Topogr 2000 13 135 144 11154103 10.1023/A:1026611219551
Spiegel MR Schiller J Srinivasan RA Schaum's outline of Probability and Statistics 2000 2 Mc Graw – Hill Companies Inc, USA
McCulloch WS Pitts W A logical calculus of the ideas immanent in nervous activity Bull Math Biol 1943 5 115 133
Caianiello ER Outline of a theory of thought-processes and thinking machines J Theor Biol 1961 2 204 235 13689819 10.1016/0022-5193(61)90046-7
Anninos PA Zenone S Elul R Artificial neural nets: dependence of the EEG amplitude's probability distribution on statistical parameters J Theor Biol 1983 103 339 348 6312198 10.1016/0022-5193(83)90290-4
Leake B Anninos PA Effect of connectivity on the activity of neural nets models J Theor Biol 1976 58 337 363 940330
Anninos PA Elul R Effect of structure on function in model nerve nets Biophys J 1974 14 8 19 4811818
Sarvas J Basic mathematical and electromagnetic concepts of the biomagnetic inverse problem Phys Med Biol 1987 32 11 22 3823129 10.1088/0031-9155/32/1/004
|
16146568
|
PMC1236964
|
CC BY
|
2021-01-04 16:39:25
|
no
|
Theor Biol Med Model. 2005 Sep 7; 2:37
|
utf-8
|
Theor Biol Med Model
| 2,005 |
10.1186/1742-4682-2-37
|
oa_comm
|
==== Front
Thromb JThrombosis Journal1477-9560BioMed Central London 1477-9560-3-111611150010.1186/1477-9560-3-11Case ReportCatheter-related septic thrombophlebitis of the great central veins successfully treated with low-dose streptokinase thrombolysis and antimicrobials Volkow Patricia [email protected]árez Patricia [email protected] Ana Berta [email protected]ía-Méndez Jorge [email protected] Enrique [email protected]érez-Padilla Rogelio [email protected] Mexican National Institute of Cancer (INCan), Mexico City, Mexico2 Mexican National Institute of Respiratory Diseases (INER), Mexico City, Mexico2005 22 8 2005 3 11 11 31 5 2005 22 8 2005 Copyright © 2005 Volkow et al; licensee BioMed Central Ltd.2005Volkow et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Septic thrombophlebitis is an iatrogenic life-threatening disease associated with use of central venous devices and intravenous (IV) therapy. In cancer patients receiving chemotherapy, vein resection or surgical thrombectomy in large central venous lines is time-consuming, can delay administration of chemotherapy, and therefore can compromise tumor control. Experience with thrombolysis has been published for catheter-related thrombosis but for septic thrombosis, this experience is scarce.
Results
We describe three patients with cancer and septic thrombophlebitis of central veins caused by Staphylococcus aureus treated with catheter removal, thrombolysis, and intravenous (IV) antibiotics. In our reported cases, an initial bolus of 250,000 international units (IU) of streptokinase administered during the first h followed by an infusion of 20,000–40,000 IU/h for 24–36 h through a proximal peripheral vein was sufficient to dissolve the thrombus. After thrombolyisis and parenteral antibiotic for 4–6 weeks the septic thrombosis due to Staphylococcus aureus solved in all cases. No surgical procedure was needed, and potential placement of a catheter in the same vein was permitted.
Conclusion
Thrombolysis with streptokinase solved symptoms, cured infection, prevented embolus, and in all cases achieved complete thrombus lysis, avoiding permanent central-vein occlusion.
==== Body
Background
Septic thrombophlebitis is an iatrogenic life-threatening disease associated with use of central venous devices and intravenous (IV) therapy. [1-3] Sole use of antimicrobials is rarely effective for controlling infection, requiring removal of the device and anticoagulation but in some cases a more aggressive approach such as resection of the affected vein [2,4-7] or trombectomy is needed [8]. Vein resection or surgical thrombectomy is time-consuming in large central venous lines, has a high rate of complications, can delay administration of chemotherapy, and therefore delay or impede tumor control. Experience with thrombolysis has been published for catheter-related thrombosis [9-13] but for septic thrombosis, this experience is scarce. [14,15] Herein, we describe three women with cancer and septic thrombophlebitis due to Staphylococcus aureus methicillin sensitive, who failed to resolve with catheter removal, parenteral antibiotics, and anticoagulation therapy and who were successfully treated with low-dose streptokinase fibrinolysis. All patients were receiving chemotherapy through non-tunneled polyurethane, single-lumen catheters placed in the subclavian-vein, but none of them had coagulopathy or septic shock.
Results
Case 1
This was the case of a woman 59-years-of-age with papillary ovarian adenocarcinoma. She had been treated during the previous 10 years with 10 mg/day of prednisone for rheumatoid arthritis. A first central-vein catheter was placed for adjuvant chemotherapy and removed 4 months later with no complications. A new catheter was set in place 8 months later after documenting tumor relapse; however, one day later the patient developed pain at the insertion site and fever (39°C), and chills. An abscess at insertion site was found and the catheter was removed. Blood cultures, purulent secretion, and catheter tip were positive for Staphylococcus aureus. Intravenous dicloxacillin was initiated and amikacin was added 1 day later, but fever and positive blood cultures persisted. Echo-Doppler documented thrombosis of brachiocephalic trunk and computed tomography (CT) scan showed a thrombus reaching brachiocephalic trunk and superior vena cava (Figure 1a); subcutaneous (SC) enoxaparin was initiated. Vancomycin was started because fever and bacteremia persisted, with no clinical improvement. Seven days after beginning with antibiotics, the patient received an initial bolus of 250,000 international units (IU) of streptokinase administered in 1 h followed by an infusion of 40,000 IU per h for 24 h through a peripheral vein. One day after thrombolysis began, fever and positive blood cultures disappeared. Full permeability of right brachiocephalic vein and superior vena cava was documented by CT scan (Figure 1b). The patient completed 4 weeks of parenteral antibiotics, but died 1 month later with peritoneal carcinomatosis-related intestinal occlusion. Necropsy study showed neither thrombosis nor obstruction of great central veins.
Figure 1 CT scan showing thrombus before and after thrombolysis. 1a. CT scan of superior vena cava with intravenous contrast infusion showing a thrombus before streptokinase infusion. 1b. CT scan of superior vena cava with intravenous contrast infusion three days after thrombolysis showing no remaining thrombus.
Case 2
A 42-year-old woman with breast cancer stage IIIB was started on chemotherapy. Four months later, she was admitted with fever, shivering, and painful erythematous lesions disseminated in legs and arms. The patient was initiated on IV vancomycin and amikacin, and after initial blood cultures grew oxacillin-sensitive S. aureus. The catheter was removed and antibiotics were changed to dicloxacillin and amikacin. Initial echo-Doppler for all limbs did not demonstrate obstruction to blood flow. Chest roentgenogram showed bilateral, multiple, rounded, irregular, non-cavitated opacities. Transthoracic echocardiography did not show heart valve vegetations, but did show a mobile hyper-reflectant image in superior vena cava extending to right atrium, suggestive of thrombus. Fever persisted and new painful nodular erythemathous lesions appeared in both limbs that evolved into abscesses, but neither purulent skin lesions nor blood cultures grew microorganisms. A second transthoracic echocardiogram performed 9 days later showed persistence of the same pediculated lesion that measured 40 × 10 × 10 mm. The patient was thrombolyzed with the same doses of streptotokinase; 24 h later, she had no fever, all symptoms resolved and transthoracic echocardiography performed 9 days later showed no lesions. She completed 6 weeks with IV antimicrobials. A month later, the scheduled mastectomy was performed the patient received 11 courses of weekly paclitaxel. Ten months later, the patient is asymptomatic with no evidence of tumor activity.
Case 3
A 57-year-old woman with ovarian adenocarcinoma stage IV metastatic to lungs with a catheter placed in the right subclavian vein, through which she received four cycles of carboplatin and paclitaxel. One week after the last chemotherapy cycle, she developed fever and pain in the right shoulder and two days later presented to the emergency room. At admission, the patient had persistent shoulder pain, anorexia, and an enlarging, painful mass in right shoulder, with an indurated, extremely tender area in right sternoclavicular joint and edema in right arm. She was febrile, hypotensive, and tachycardic. Oxacillin-sensitive Staphylococcus aureus grew in blood and catheter tip, and was started on dicloxacillin and amikacin. Echo-Doppler revealed a 4-cm long thrombus within the right subclavian vein partially occluding the right jugular vein. No intracardiac thrombus or valvular lesions were observed in the Echocardiogram. CT scan showed a large collection of liquid in right shoulder sternoclavicular joint. The patient received an initial streptokinase bolus of 250,000 IU and one 12-h infusion of 20,000 U/h, followed by enoxaparin 60 mg BID. Twenty four hours later, Echo-Doppler showed patent right subclavian and jugular veins. She completed 3 weeks of enoxaparin and then changed to oral anticoagulation. Technetium bone scan showed evidence of ipsilateral clavicle osteomyelitis. She received IV antibiotics for 4 weeks followed by oral dicloxacillin plus rifampin for 28 weeks. After treatment the bone scan did not have evidence of osteomyelitis and 8 months later the patient had normal shoulder function without arm edema.
Discussion
Intravascular infection and thrombosis are two of the most serious complications related to central venous catheter use. Central vein thrombosis was described as a complication of catheters in 1971 [2]. Symptoms were present in <4% of patients with central venous catheter when venography showed thrombosis [2]. Neoplastic disease often creates a thrombogenic state, through inflammation mediators, tumor necrosis factor, platelet activation, as well as a procoagulant substances released by tumor cells [15]. In addition, long indwelling lines increase risk for thrombosis, reported in 0.06-32% of patients, although the risk changes with type of catheter, neoplasia, chemotherapy regimes and radiation [16]. The complications of catheter-related thrombosis are similar although not as frequently as has been described for lower limb thrombosis [16]. It can produce pulmonary embolism. The trombus can become infected with persistent bacteremia and septic embolization ensue [17]. It has been recognized that CVC infection increases the risk of thrombosis [18] even though we believe that the incidence of septic thrombosis with persistent refractory bacteremia as the cases herein described is uncommon, in a recent review CVC associated thrombosis this complication is not mentioned [17]. For the last 6 years at a hospital where we placed >1,100 long indwelling catheters for cancer patients annually we have observed approximately one case every two years.
Standard therapy for catheter associated septic thrombosis includes antibiotics, catheter removal, full heparin anticoagulation, and venotomy. The latter is technically impossible for great central veins, although surgical thrombectomy has been successfully performed [8] and medical lysis of the thrombus is feasible [13,14]. We describe successful lysis of septic thrombosis with low-dose streptokinase infusion through a peripheral vein proximal to central great vein affected and no surgical or invasive procedure performed. This approach was first reported with a high percentage of success in catheter-related thrombosis in the early 1980s, allowing to maintain vein patency [4,6] using streptokinase, urokinase, and more recently, recombinant-tissue plasminogen activator, [13,14]. In our reported cases, streptokinase administered as initial bolus of 250,000 IU during 1 hour followed by infusion of 20,000-40,000 IU/h for 24-36 hours through a proximal peripheral vein was sufficient to dissolve the thrombus Table 1. This treatment solved symptoms, cured infection, prevented embolus, and in all cases achieved complete thrombus lysis, avoiding permanent central-vein occlusion. The episode of septic thrombosis due to Staphylococcus aureus solved with continued parenteral antibiotic for 4 to 6 weeks in all cases and no surgical procedure was required.
Table 1 Demographic and clinical characteristics of the patients described
Case 1 2 3
Gender F F F
Age 59 42 57
Cancer site Ovarian (relapse) Breast stage III-B Ovarian stage IV
Time of catheter stay (days) before symptoms 1 155 63
Thrombus site Superior vena cava Superior vena cava extended to right atrium Subclavian and yugular veins
Days of antimicrobials before thrombolysis 7 19 6
Indication for thrombolyis Persitent fever and bacteremia. Persistent fever and septic embolization Persitent fever and septic embolization
Staphylococcus aureus methicillin sensible + + +
Streptokinase dose 250,000 IU hr. bolus + 40,000 us/hr for 24 hrs. 250,000 IU us/hr bolus. 250,000 us 1 hr. bolus + 25,000 us/hr for 12 hrs.
Thrombus lysis 100% 100% 100%
ARDS* No No No
* Acute respiratory distress syndrome
Conclusion
Fibrinolytic therapy with streptokinase is a therapeutic option in the management of catheter-related septic thrombophlebitis of the great central veins. This therapeutic approach, mantain central vein patency, allowing potential to place a new long indwelling catheter, the cornerstone for cancer patients who need chemotherapy.
List of abbreviations
IV – intravenous
IU – International units
CT – Computed tomography
SC – subcutaneous
ARDS – Acute respiratory distress syndrome
Competing interests
The authors declare that we have not competing interests in the interpretation of data or presentation of information influenced by our personal or financial relationship with other people or organizations.
Authors' contributions
PV – Participated in the design of the study, collected data, wrote the manuscript, general supervision of the research group
PCJ – Collected data, wrote the manuscript
AAB – Collected data and looked for the researches made before related with the manuscript
JGM – Collected data and looked for the researches made before related with the manuscript
EBL – Revised the manuscript and analysis of data
RPP – Revised the manuscript, analysis of data and final approval of the version
All authors read and approved the final manuscript.
==== Refs
Bayer AS Scheld WM Mandell GL, Bennett JE, Dolin R Endocarditis and intravascular infections. In: Principles and practice of infectious diseases 2000 1 5 New York: Churchill Livingstone 857 902
Andes DR Urban AW Acher ChW Maki DG Septic thrombosis of basilic, axillary, and subclavian veins caused by a peripherally inserted central venous catheter Am J Med 1998 105 446 450 9831430 10.1016/S0002-9343(98)00287-3
Raad I Luna M Khalil S Costerton J Lam C Bodey G The relationship between the thrombotic and infectious complications of central venous catheters JAMA 1994 271 1014 16 8139059 10.1001/jama.271.13.1014
Verghese A Widrich W Arbeit R Central venous septic thrombophlebitis – The role of medical therapy Medicine 1985 64 394 400 3932817
Slagle DC Gates RH Unusual case of central vein thrombosis and sepsis Am J Med 1986 81 351 4 3090881 10.1016/0002-9343(86)90278-0
Kaufman J Demas C Stark K Flancbaum L Catheter-related septic central venous thrombosis – Current therapeutic options West J Med 1986 145 200 3 3765599
Hoffman MJ Greenfield LJ Central venous septic thrombosis managed by superior vena cava Greenfield filter and venous thrombectomy: a case report J Vasc Surg 1986 4 606 11 3783835 10.1067/mva.1986.avs0040606
Kniemeyer HW Grabitz K Buhl R Wust HJ Sandmann W Surgical treatment of septic deep venous thrombosis Surgery 1995 118 49 53 7604379
Seigel EL Jew AC Delcore R Iliopoulos J Thomas J Thrombolytic therapy for catheter-related thrombosis Am Surg 1993 166 716 9
Fraschini G Jadeja J Lawson M Holmes F Carrasco H Wallace S Local infusion of urokinase for the lysis of thrombosis associated with permanent central venous catheters in cancer patients J Clin Oncol 1987 5 672 8 3559656
Rubenstein M Creger WP Successful streptokinase therapy for catheter-induced subclavian vein thrombosis Arch Intern Med 1980 140 1370 1 6775611 10.1001/archinte.140.10.1370
Ponec D Irwin D Haire WD Hill PA Li X McCluskey ER Recombinant tissue plasminogen activator (alteplase) for restoration of flow in occluded central venous access devices: a double-blind placebo-controlled trial – The Cardiovascular Thrombolytic to Open Occluded Lines (COOL) Efficacy Trial J Vasc Interv Radiol 2001 12 951 5 11487675
Haire WD Techniques in dosing for thrombolysis of occluded central venous catheters Tech Vasc Interv Radiol 2001 4 127 30 11981800 10.1016/S1089-2516(01)90008-3
Schranz D Haugwitz D Zimmer B Schumacher R Successful lysis of a septic thrombosis of the superior vena cava using recombinant tissue-plasminogen activator Klin Padiatr 1991 203 363 5 1942943
Lewis JA LaFrance R Bower RH Treatment of an infected silicone right atrial catheter with combined fibrinolytic and antibiotic therapy: case report and review of the literature J Parenter Enteral Nutr 1989 13 92 8
Van Rooden CJ Tesselaar ME Osanto S Rosendaal FR Huisman MV Deep vein thrombosis associated with central venous catheters – a review J Thromb Haemost 2005 15975139
Devie-Hubert I Carlier M Pozzo Di Borgo C Venous thrombosis on central catheters in oncology Rev Med Intern 1996 17 821 5
Van Rooden CJ Schippers EF Barge RM Rosendaal FR Guiot HF van der Meer FJ Meinders AE Huisman MV Infectious complications of central venous catheters increase the risk of catheter-related thrombosis in hematology patients: a prospective study J Clin Oncol 2005 23 2655 60 15837979 10.1200/JCO.2005.05.002
|
16111500
|
PMC1236965
|
CC BY
|
2021-01-04 16:36:24
|
no
|
Thromb J. 2005 Aug 22; 3:11
|
utf-8
|
Thromb J
| 2,005 |
10.1186/1477-9560-3-11
|
oa_comm
|
==== Front
J Transl MedJournal of Translational Medicine1479-5876BioMed Central London 1479-5876-3-331613140010.1186/1479-5876-3-33CommentaryImplementations of translational medicine Sonntag Kai-Christian [email protected] Harvard Medical School, Center for Neuroregeneration Research, MRC 102, McLean Hospital, 115 Mill Street, Belmont, MA 02478, USA2005 30 8 2005 3 33 33 5 8 2005 30 8 2005 Copyright © 2005 Sonntag; licensee BioMed Central Ltd.2005Sonntag; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
New developments in science are rapidly influencing and shaping basic and clinical research and medicine. This has led to the emergence of multiple opportunities and challenges on many levels in the bio-medical and other associated fields. To face these opportunities and challenges, new concepts and strategies are needed. These can be provided by translational research/medicine as an integrative concept based on a multidirectional understanding of research and medicine embedded in a socio-economical environment. Although the implementation of translational research/medicine faces many obstacles, some of its goals have already been part of new programs in local institutions and in medical or scientific societies. These implementations are important in creating a unified national and international system of translational research/medicine.
==== Body
The rapid evolutions in science have generated a tremendous spectrum of new technologies and tools in both basic and clinical research/medicine. This includes the constant improvement of old and the discovery of new diagnostics and therapies, which increasingly contain and integrate elements from different fields, such as biomedical and other sciences, modern and traditional medicine and various technology branches. In addition, the application of these developments in clinical settings have created a "feed-back-loop" providing crucial information about their feasibility and success in improving human health. This network of scientific and clinical research/medicine has become one of the factors in shaping modern societies not only by being a major economical factor (see [1] for details), but also by challenging basic values and traditional thinking. To face the emerging challenges of creating a balanced and effective healthcare system, new concepts are needed for providing a framework of integrative strategies and solutions that efficiently combine basic and clinical research/medicine.
So far, translational research/medicine has rather been a linear concept rooted in traditional (academic) approaches to provide therapies for diseases (from bench to bedside), while paying little attention to patient-oriented research that involves understanding the underlying cause of disease and its treatments (from bedside to bench). Moreover, not much attention has been paid to many socio-economical aspects that are associated with research or medicine. Therefore, new definitions based on a bi- or multi-directional understanding of translational research/medicine have been proposed [2-5] and, recently, a strategic outline to successfully implement the goals of this new concept has been outlined in an article by Hörig et al. in Nature Medicine [1].
This commentary aims to add a few additional aspects by emphasizing two major factors that strongly influence and, in turn, are influenced by translational research/medicine: (1) The rapid evolution of new technologies and therapeutic approaches, and (2) examples of attempts to create programs, which are based on a translational understanding of basic and clinical research and/or medicine.
In the past two decades, biomedical and other research fields like gene therapy, the human genome project (HUGO), stem cell research, cloning, nanotechnology and others have revolutionized medicine and generated entirely new fields and approaches in treatment of diseases [6]. Thus, it is becoming more and more obvious that new forms of therapies, such as gene- or cell-replacement-therapy, will have a place in future therapeutic approaches. In addition, the rapid expansion of bio- and other technology development are expected to lead to an advanced understanding of diseases [7,8]. These new fields not only offer multiple new opportunities in the research and biomedical sector, but also require attention on multiple other levels involving religious and ethical issues [9] including new definitions of life, disease, and treatment, the power and limitations of technology, the availability of resources, and the adjustment of social and political actions.
How can these fields be combined and integrated in the concept of translational research/medicine? Are there any attempts to develop programs based on a "translational" thinking in research and medicine? So far, the concepts of translational research/medicine have been described and discussed in the context of different medical fields, e.g. in obstetrics and gynecology [10], radiation therapy and radiobiology [11,12], cancer therapy [8], psychiatry [13], pain research [14], gastroenterology [15] or in the broader context of research and medicine [16]. And, so far, it seems that there is an overall agreement in defining the goals of translational research/medicine and delineating the obstacles in their implementations [1,17-19]. In addition, many of these articles have provided suggestions and strategies how to translate these concepts into practice (e.g. [1,8,12,20-23]). However, it should be emphasized that translational research/medicine does not affect specialized fields alone. Rather it should be an integrative concept that, with implications to everybody in the field, should be part of a broader understanding of how research and medicine should be organized. A major part of these processes refers to education and training opportunities and there have been examples published how this can be implemented by personal experience (e.g. [16,24]), by the quest for specialized institutions (e.g. [20]), or by using existing educational systems (e.g. [25-29]). It should also be noted that scientific/medical societies have started to adopt and integrate many aspects of the concept of translational research/medicine in their agendas and are in the process to find new creative and innovative strategies on multiple levels, including education, public discussions, funding opportunities, career development, etc. This especially applies to societies, which integrate a vast spectrum of bioresearch, biotechnology and biomedical fields such as the American Society for Gene Therapy (ASGT; ) or the International Society for Stem Cell Research (ISSCR; ). On the local level, there are existing clinical programs that aim to improve the clinical education by emphasizing translational thinking (e.g. [25,27-30]) or other programs with a broader approach of being integrative between researchers, clinicians and health care providers. The latter programs – like the medical societies – offer new opportunities to facilitate collaborations between researchers and clinicians and represent a platform to process the multilevel challenges emerging from the rapid advances in the biomedical fields. For example, the Harvard Medical School has launched two new centers, the Harvard Center for Neuroregeneration (HCNR; ) and the Harvard Stem Cell Institute (HSCI; ), whose goals are to create an environment that combines investigators on multiple levels and address many features of translational research/medicine like identification of new and alternative funding opportunities including academia-industry synergies, healthcare delivery, the improvement of infrastructural hurdles by creating core facilities, training opportunities, etc. These "enterprises" require significant logistical efforts and often rely on established systems, which can provide, space, informatics, administration, and skilled personnel. Although these are local efforts and might not reflect the overall situation, it shows that efforts are made and many aspects of translational research/medicine can be realized. The experiences gained from these local "enterprises" might be then translatable and integrated into a bigger national and international network creating a more unified system of translational research/medicine.
In summary, translational research/medicine is not a single standing idea that can be applied whenever it is needed. Rather, it is a concept that needs the attention from everyone and should be the foundation of a modern understanding of health provision. It is essential to face the challenges emerging from the rapid advances in the biomedical and other associated fields. Its implementation needs a multilevel effort in many different areas, which together create a health system as a whole. Some of the goals in translational research/medicine are already part of new programs in local institutions and in medical or scientific societies. These implementations are important in creating an overall and unifying network of translational research/medicine.
Acknowledgements
The author wishes to thank Dr. Christian Brander for review and helpful comments of the manuscript.
==== Refs
Horig H Marincola E Marincola FM Obstacles and opportunities in translational research Nat Med 2005 11 705 708 16015353 10.1038/nm0705-705
Marincola FM Translational Medicine: A two-way road J Transl Med 2003 1 1 14527344 10.1186/1479-5876-1-1
Rustgi AK Translational research: what is it? Gastroenterology 1999 116 1285 10348809
Phillipson EA If it's not integrative, it may not be translational Clin Invest Med 2002 25 94 96 12137259
Dauphinee D Martin JB Breaking down the walls: thoughts on the scholarship of integration Acad Med 2000 75 881 886 10995608
Russell JH Stahl PD Stephenson J Whelan A Biomedical education in the 21st century Mo Med 2004 101 484 486 15535023
Molidor R Sturn A Maurer M Trajanoski Z New trends in bioinformatics: from genome sequence to personalized medicine Exp Gerontol 2003 38 1031 1036 14580855 10.1016/S0531-5565(03)00168-2
Bast RCJ Mills GB Young RC Translational research--traffic on the bridge Biomed Pharmacother 2001 55 565 571 11769968 10.1016/S0753-3322(01)00144-5
Kagarise MJ Sheldon GF Translational ethics: a perspective for the new millennium Arch Surg 2000 135 39 45 10636345 10.1001/archsurg.135.1.39
Baker PN Where is translational research? Bjog 2004 111 395 396 15104600
Dische S Saunders M Translational Research--a new entity? Acta Oncol 2001 40 995 999 11845966 10.1080/02841860152708297
Coleman CN International Conference on Translational Research ICTR 2003 Conference Summary: marshalling resources in a complex time Int J Radiat Oncol Biol Phys 2004 58 307 319 14751498 10.1016/j.ijrobp.2003.09.056
Licinio J Wong ML Translational research in psychiatry: pitfalls and opportunities for career development Mol Psychiatry 2004 9 117 14966470 10.1038/sj.mp.4001479
Mao J Translational pain research: bridging the gap between basic and clinical research Pain 2002 97 183 187 12044614 10.1016/S0304-3959(02)00109-4
Sartor RB Translational research: Bridging the widening gap between basic and clinical research Gastroenterology 2003 124 1178 12730855 10.1016/S0016-5085(03)00341-X
Nathan DG The several Cs of translational clinical research J Clin Invest 2005 115 795 797 15841166 10.1172/JCI200524753
Mankoff SP Brander C Ferrone S Marincola FM Lost in Translation: Obstacles to Translational Medicine J Transl Med 2004 2 14 15149545 10.1186/1479-5876-2-14
Ioannidis JP Materializing research promises: opportunities, priorities and conflicts in translational medicine J Transl Med 2004 2 5 14754464 10.1186/1479-5876-2-5
Pober JS Neuhauser CS Pober JM Obstacles facing translational research in academic medical centers Faseb J 2001 15 2303 2313 11689456 10.1096/fj.01-0540lsf
Hotez PJ Should we establish a North American school of global health sciences? Am J Med Sci 2004 328 71 77 15311164 10.1097/00000441-200408000-00001
Sung NS Crowley WFJ Genel M Salber P Sandy L Sherwood LM Johnson SB Catanese V Tilson H Getz K Larson EL Scheinberg D Reece EA Slavkin H Dobs A Grebb J Martinez RA Korn A Rimoin D Central challenges facing the national clinical research enterprise Jama 2003 289 1278 1287 12633190 10.1001/jama.289.10.1278
Mark AL Kelch RP Clinician scientist training program: a proposal for training medical students in clinical research J Investig Med 2001 49 486 490 11730083
Fenton W James R Insel T Psychiatry residency training, the physician-scientist, and the future of psychiatry Acad Psychiatry 2004 28 263 266 15673819 10.1176/appi.ap.28.4.263
Liang MH Translational research: getting the word and the meaning right Arthritis Rheum 2003 49 720 721 14558060 10.1002/art.11370
Gordon SM Heft MW Dionne RA Jeffcoat MK Alfano MC Valachovic RW Lipton JA Capacity for training in clinical research: status and opportunities J Dent Educ 2003 67 622 629 12856962
Schwartz-Bloom R Science education: a neuroscientist's view of translational medicine J Neurosci 2005 25 5667 5669 15958732 10.1523/JNEUROSCI.1693-05.2005
Herold BC McArdle P Stagnaro-Green A Translational medicine in the first year: integrative cores Acad Med 2002 77 1171 12431953 10.1097/00001888-200211000-00040
DePaolo LV Leppert PC Providing research and research training infrastructures for clinical research in the reproductive sciences Am J Obstet Gynecol 2002 187 1087 1090 12389010 10.1067/mob.2002.125896
Crowley WFJ Thier SO A program to facilitate clinical research in an AHC: the first five years Acad Med 2001 76 403 409 11346512
Halpain MC Jeste DV Trinidad GI Wetherell JL Lebowitz BD Intensive short-term research training for undergraduate, graduate, and medical students: early experience with a new national-level approach in geriatric mental health Acad Psychiatry 2005 29 58 65 15772406 10.1176/appi.ap.29.1.58
|
16131400
|
PMC1236966
|
CC BY
|
2021-01-04 16:39:28
|
no
|
J Transl Med. 2005 Aug 30; 3:33
|
utf-8
|
J Transl Med
| 2,005 |
10.1186/1479-5876-3-33
|
oa_comm
|
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-731612238810.1186/1743-422X-2-73ResearchThe SARS Coronavirus S Glycoprotein Receptor Binding Domain: Fine Mapping and Functional Characterization Chakraborti Samitabh [email protected] Ponraj [email protected] Xiaodong [email protected] Dimiter S [email protected] Protein Interactions Group, LECB, CCR, NCI-Frederick, NIH, Frederick, MD 21702-12012005 25 8 2005 2 73 73 18 7 2005 25 8 2005 Copyright © 2005 Chakraborti et al; licensee BioMed Central Ltd.2005Chakraborti et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The entry of the SARS coronavirus (SCV) into cells is initiated by binding of its spike envelope glycoprotein (S) to a receptor, ACE2. We and others identified the receptor-binding domain (RBD) by using S fragments of various lengths but all including the amino acid residue 318 and two other potential glycosylation sites. To further characterize the role of glycosylation and identify residues important for its function as an interacting partner of ACE2, we have cloned, expressed and characterized various soluble fragments of S containing RBD, and mutated all potential glycosylation sites and 32 other residues. The shortest of these fragments still able to bind the receptor ACE2 did not include residue 318 (which is a potential glycosylation site), but started at residue 319, and has only two potential glycosylation sites (residues 330 and 357). Mutation of each of these sites to either alanine or glutamine, as well as mutation of residue 318 to alanine in longer fragments resulted in the same decrease of molecular weight (by approximately 3 kDa) suggesting that all glycosylation sites are functional. Simultaneous mutation of all glycosylation sites resulted in lack of expression suggesting that at least one glycosylation site (any of the three) is required for expression. Glycosylation did not affect binding to ACE2. Alanine scanning mutagenesis of the fragment S319–518 resulted in the identification of ten residues (K390, R426, D429, T431, I455, N473, F483, Q492, Y494, R495) that significantly reduced binding to ACE2, and one residue (D393) that appears to increase binding. Mutation of residue T431 reduced binding by about 2-fold, and mutation of the other eight residues – by more than 10-fold. Analysis of these data and the mapping of these mutations on the recently determined crystal structure of a fragment containing the RBD complexed to ACE2 (Li, F, Li, W, Farzan, M, and Harrison, S. C., submitted) suggested the existence of two hot spots on the S RBD surface, R426 and N473, which are likely to contribute significant portion of the binding energy. The finding that most of the mutations (23 out of 34 including glycosylation sites) do not affect the RBD binding function indicates possible mechanisms for evasion of immune responses.
==== Body
Background
Viral envelope glycoproteins initiate entry of viruses into cells by binding to cell surface receptors followed by conformational changes leading to membrane fusion and delivery of the genome to the cytoplasm [1]. The spike (S) glycoproteins of coronaviruses are no exception and mediate binding to host cells followed by membrane fusion; they are major targets for neutralizing antibodies and form the characteristic corona of large, distinctive spikes in the viral envelopes [2,3]. Such 20 nm complex surface projections also surround the periphery of the SCV particles [4]. The level of overall sequence similarity between the predicted amino acid sequence of the SCV S glycoprotein and the S glycoproteins of other coronaviruses is low (20–27% pairwise amino acid identity) except for some conserved sequences in the S2 subunit [5]. The low level of sequence similarity precludes definite conclusions about functional and structural similarity.
The full-length SCV S glycoprotein and various soluble fragments have been recently cloned, expressed and characterized [6-11]. The S glycoprotein runs at about 170–200 kDa in SDS gels suggesting posttranslational modifications as predicted by previous computer analysis and observed for other coronaviruses [6,11]. S and its soluble ectodomain, Se, were not cleaved to any significant degree [6]. Because the S protein of coronaviruses is a class I fusion protein [12], this observation classifies the SCV S protein as an exception to the rule that class I fusion proteins are cleaved exposing an N-terminal fusogenic sequence (fusion peptide) although cleavage of S could enhance fusion [9].
Because S is not cleaved, it is difficult to define the exact location of the boundary between S1 and S2; presumably it is somewhere between residues around 672 and 758 [6,7]. Fragments containing the N-terminal amino acid residues 17 to 537 and 272 to 537 but not 17 to 276 bound specifically to Vero E6 cells and purified soluble receptor (ACE2) molecules [6]. Together with data for inhibition of binding by antibodies, developed against peptides from S, these findings suggested that the receptor-binding domain (RBD) is located between amino acid residues 303 and 537 [6]. Two other groups obtained similar results and found that independently folded fragments containing residues 318 to 510 [8] and 270 to 510 [10] can bind receptor molecules. Currently, these fragments are being further characterized to better understand the interactions of the virus with its receptor as well as their potential as inhibitors of the virus entry by blocking these interactions. Here, we present evidence that glycosylation of these and other fragments containing the S RBD does not affect to any measurable degree their binding to the receptor (ACE2), and analyze the S RBD-ACE2 interaction.
Results
A short RBD fragment containing only two potential glycosylation sites folds independently and binds ACE2
We and others have previously identified the RBD by using fragments containing three potential glycosylation sites – at residues 318, 330 and 357 [6,8,10]. To find the minimal number of potential glycosylation sites and shortest length required for expression and folding of S RBD fragments, we cloned in pSecTag 2B fragments with various number of potential glycosylation sites and length including S317–518, S319–518, S329–518, S364–537, S399–518, S317–493, and S329–458, where the numbers after S denote the amino acid residues confining the fragment. Note that these fragments were not constructed as fusion proteins with Fc as in a previous report [8]. This is why we also designed and tested several fragments with deleted portions of the RBD that have already been shown to be important for binding to ACE2 including regions between residues 327 and 490 [8]. The S317–518 and S319–518 fragments were secreted in the culture supernatant (Fig 1A), and bound to ACE2-expressing cells (Fig 1B) and purified ACE2 (Table 1 and data not shown). The difference in the molecular weights of the two fragments (about 3 kD) is much larger than the calculated weight due to the two additional amino acids contained in S317–518, and is likely due to glycosylation. Both fragments bound to ACE2 at comparable levels (Fig. 1B). The other fragments were not secreted (Fig. 1A) but could be detected by Western in cell lysates (data not shown). These results suggest that a short fragment (S319–518), which is not a fusion protein, with only two glycosylation sites can be independently folded and secreted in a soluble form, and can bind ACE2.
Figure 1 Expression and binding of soluble S fragments containing the RBD. A) Soluble S proteins concentrated using Ni-NTA agarose beads from the supernatants of 293 cells transfected with various constructs were run, blotted onto a nitrocellulose membrane and detected with anti-c-myc epitope antibody. B) Cell binding assay data using supernatants described above, shown as a percentage of the reading of S272–537 that has been used in this experiment as a positive control.
Table 1 S RBD mutants, expression levels and binding to ACE2.
Mutant Mutation Expression Binding ASA
1 E327 98 83 123
2 K333 86 90 176
3* K344 95 102 159
4 K390 104 1 44
5 D392 110 95 69
6 D393 30 100 10
7 K411 90 103 33
8 D414 120 130 113
9 D415 90 102 97
10* R426 73 7 95
11 N427 100 111 121
12 D429 103 0 9
13 T431 131 64 59
14 K439 85 87 65
15 R441 10 15 3
16* Y442 105 110 68
17* R444 80 86 52
18 H445 124 103 113
19 K447 87 85 138
20 R449 96 101 178
21 F451 69 71 64
22 D454 50 4 25
23 I455 77 6 89
24 D463 87 81 70
25* L472 95 99 172
26 N473 100 0 70
27 W476 80 76 126
28 F483 91 3 2
29 Q492 95 3 5
30 Y494 50 7 21
31 R495 97 19 7
32 E502 110 84 175
33 S17–276 90 0
34 S319–518 100 100
The mutants that significantly decrease binding to ACE2 are shown in bold. The * denotes mutant residues that are naturally occurring in various SCV strains (see Fig. 6A). The binding and expression values for the individual mutants are expressed as a percentage of the value for the S319–518 (wt) that is assumed 100%. The values of accessible surface area (ASA, Å2) for mutant residues were calculated from the crystal structure of the S RBD-ACE2 complex (coordinates provided by S. Harrison) by using the Lee and Richards' algorithm [23] with a probe radius of 1.4 Å.
The potential glycosylation sites in RBD fragments are functional and glycosylation does not affect binding to ACE2
To find whether the potential glycosylation sites in the RBD fragments are functional we constructed mutants, where the three residues N318, N330 and N357 in S317–319 were mutated individually from asparagine to alanine. As is shown in Fig. 2A all three mutants were expressed and ran on SDS-PAGE at molecular weights of about 3 kD smaller than the unmodified fragment. They all bound to ACE2 (Fig. 2B). Similar results were obtained with the shorter fragment (S319–518) where asparagines were also mutated to glutamines, which better mimic asparagines (Fig. 3). These results suggest that all glycosylation sites in the RBD are functional, and that the lack of glycosylation in any of the glycosylation sites does not interfere with binding to ACE2.
Figure 2 Glycosylation of S fragment containing the RBD. A) Expression of the three mutants on S317–518 where the potential sites of glycosylation at N318, N330 and N357 were individually converted to alanine. All the mutants appear to have similar molecular weights when compared to the wild type protein S317–518. B) Cell binding data of the same mutants.
Figure 3 Effects of glycosylation on expression and binding of RBD-containing fragments. A) Expression of the four mutants on S319–518 where the two sites of glycosylation at N330 and N357 have been individually converted to either alanine or glutamine. The various mutants have similar molecular weights, a little less than the wild type indicating that the level of glycosylation at each residue might be similar. B) Cell binding data for the same mutants.
Only one glycosylation site is required for secretion of functional RBD fragments
To find the minimal number of functional glycosylation sites required for secretion of the RBD we generated double mutants of S319–518 where the asparagines N330 and N357 were mutated to either alanines (Ala 2) or glutamines (Gln 2). These mutants were not detected in the culture supernatants (Fig. 4A) and the culture supernatants did not exhibit any binding activity to ACE2 (Fig. 4B). These results suggest that at least one glycosylation site is required for secretion of functional RBD fragments.
Figure 4 Glycosylation of at least one residue in RBD-containing fragments is required for expression. A) Expression pattern of two mutants on S319–518 in which both the glycosylation sites at N330 and N357 have been mutated either to alanine or to glutamine. No expression is seen when both the sites have been mutated indicating that glycosylation of at least one of the sites is important. In the last lane, purified S317–518 protein has been loaded as a control. B) Cell binding results of the same mutants.
Identification of 11 RBD amino acid residue mutations that affect its binding to ACE2, and 20 – that do not
To identify RBD amino acid residues that might affect binding to ACE2, we converted 32 residues in S319–518 to alanine, expressed the mutants and tested their binding to ACE2. Eleven mutants, K390, R426, D429, T431, D454, I455, N473, F483, Q492, Y494, and R495 exhibited decreased binding to ACE2 at comparable levels of expression (Table 1). Note that RBD fragment mutated at D454 or Y494 was expressed at somewhat lower levels but binding was much more significantly reduced. In addition, one of these mutations, D454, was previously shown to affect the RBD-ACE2 interaction [8]. The T431 mutation reduced binding but to lesser extent than the other mutations that decreased very significantly (more than 10-fold) the RBD-ACE2 interaction. The protein mutated at R441 expressed poorly and we were not able to assess its role in the RBD binding, although because of the similar levels of decrease in binding and expression, it is likely that this mutation does not affect binding. Interestingly, it appears that the D393 mutation enhanced binding – the mutated fragment expressed at low concentration but its binding equaled the binding of the non-mutated protein. The mutated residues that affect RBD binding include positively and negatively charged, polar and hydrophobic residues, indicating a role of electrostatic and hydrophobic interactions in the RBD-ACE2 interactions. These results also demonstrate that the mutations for the selected panel of residues that do affect binding are significantly (about 2-fold) more than those that do not, suggesting possible mechanisms of immune evasion.
Analysis of the S RBD sequence and the role of critical residues in S RBD
In order to further characterize the RBD and its interaction with ACE2 we analyzed the sequence and secondary structure, and how they relate to the mutations that affect binding to the receptor. A sequence-based secondary structure analysis of the S RBD predicted mostly β-sheets (data not shown), connected by loops or turns, where most of the residues affecting the RBD-ACE2 interactions are located. To find out additional residues that are not likely to affect binding significantly we aligned multiple RBD sequences of various non-redundant SCV strains. Figure 5A shows the identified 13 amino acid residues, which can be mutated without affecting the function of the virus to cause infection. Interestingly, one of these residues, R426, which decreases binding to ACE2 about 10-fold if mutated to A, is mutated to G in one of the strains. Four of the other 12 mutations (indicated with * in Table 1) do not affect binding to ACE2 when converted to A. To examine the extent of similarities of the SCV RBD sequence with related sequences of other coronaviruses from different organisms, which share only about 20–35% sequence identities, we performed multiple alignments using BLAST. Strikingly, six cysteine residues are conserved (Fig. 5B) indicating the possibility for up to three possible disulphide bridges within the S RBD that can help to keep the structural integrity of this domain. Most of the residues we found important for binding are highly variable except T431, Q492 and R495, which are highly conserved (Fig. 5B). The multiple sequence alignment score was then used to build a phylogram by using the ClustalW software. The results suggested that the SCV S RBD is much more distant than the respective regions of the other tested coronaviruses (Fig. 5C).
Figure 5 Multiple sequence alignment of S fragment (RBD) with SARS CoV-related and other coronaviruses/spike glycoproteins. A) The table shows 13 amino acid residues in the region of S RBD (319–518) which have sequence variations as identified from the multiple sequence alignment of S RBD with 19 SARS CoV-related sequences (97–99% identities with S RBD) using BLAST. B) Multiple sequence alignment of S RBD and 7 other related proteins from different organisms which share 20–35% identities: bovine coronavirus (BCoV, 327–622), canine respiratory coronavirus (CCoV, 327–622), human coronavirus (OC43, 331–612), equine coronavirus (ECoV, 327–622), porcine hemagglutinating encephalomyelitis virus (PHEV, 327–608), rat sialodacryoadenitis coronavirus (RtCoV, 325–610) and murine hepatitis virus (MHV, 325–611). Dark and gray colors indicate the identity and similarity of residues aligned. Arrowheads on the S RBD sequence show the 13 sites, which are found to have sequence variations. C) The phylogram tree is shown with distances along the protein names and note that S RBD has the highest distance. Multiple sequence alignment and phylogram were constructed using ClustalW program.
Recently, the crystal structure of S RBD-ACE2 complex was solved and the coordinates became available after the completion of this study, kindly provided by Stephen Harrison (Li, F, Li, W, Farzan, M, and Harrison, S. C., submitted). We have mapped the S RBD mutations on the surface of the crystal structure by using InsightII software. The Connolly molecular surface of the S RBD as viewed from the receptor ACE2 is shown in Fig. 6A. The S RBD is in yellow color in which the mutants that significantly affect the binding to ACE2 are shown in red and those that do not affect the binding are in cyan. The two glycosylation sites at 330 and 357 positions are colored in green. In the right panel the structure is rotated by 180° to show the opposite side of the RBD surface.
Figure 6 Mapping of the S RBD mutants on the structure. The molecular surface diagrams of S RBD are shown as the top views in the solid and translucent models. The S RBD surface is in yellow, mutations that significantly affect the binding to ACE2 are in red and those do not affect the binding in cyan. (A) Shown are the solid surface diagrams using the structure of S RBD (left panel) and related by 180° rotations (right panel). The residues that decrease the receptor binding as observed in the experiment and exposed in the structure are labeled (R426, N473). (B) The same surface diagrams as in A but with transparency which are related by 180° rotations. The buried residues, which reduce the receptor binding as observed in the experiment, are seen as blurred red.
In the structure of the S RBD-ACE2 complex two of the mutants with very significantly reduced binding to ACE2, R426A and N473A, make contacts with ACE2 residues and are completely exposed (Table 1). They are separated by residues whose mutations do not affect the S RBD binding to ACE2. Interestingly, six of the mutations we identified to reduce binding are buried but at close proximity to R426 as shown by the translucent surface highlighting in Fig. 6B indicating sensitivity of this area to mutations and likely involvement of other residues. Residues D454 and I455, whose mutation reduced binding to ACE2, do not make contacts with ACE2 and are located on the side opposing the side facing the receptor (right panel of Fig. 6); it is likely that the mutations decrease binding by inducing conformational changes. Other mutations including mutations of the two glycosylations sites on that side do not affect binding to ACE2 (right panels of Fig. 6). These results suggest the existence of two hot spots on the S RBD surface, R426 and N473, which are likely to contribute significant portion of the binding energy.
Discussion
The major results of this work are the demonstration of the functionality of the potential glycosylation sites of the S RBD and the requirement of at least one of them for its proper expression as well as the identification of two hot spots on the S RBD surface, R426 and N473, which are likely to contribute significant portion of the binding energy to ACE2. ACE2 was previously identified as a receptor for the SCV [7] and this finding was confirmed [6,13]. ACE2 binds with high (nM) affinity to S and is expected to induce conformational changes required for membrane fusion [6-8,14]. Its crystal structure was recently reported [15] and is in general agreement with two homology models previously developed [16,17]. It was proposed that the S binding domain on ACE2 involves residues on the ridges surrounding the enzymatic site [17]. Recently, several ACE2 regions and amino acid residues were identified as important for its binding to the S RBD [18].
Currently, the three-dimensional (3D) structure of the S RBD in free unbound form is unknown. We performed sequence analysis and developed a 3D model of a fragment containing the S RBD (the model will be described elsewhere). According to this model the S RBD like RBDs from other viruses contains predominantly β-sheets. Most of the residues affecting the ACE2 interactions are exposed on the surface of the beta sheets and inter-connecting loops. These predicted observations are consistent with the recently solved crystal structure of S RBD complexed with ACE2 (Li, F, Li, W, Farzan, M, and Harrison, S. C., submitted). The nature of the residues, which include charged, hydrophobic and polar residues indicated that all these types of interactions could be involved either directly or indirectly in the S RBD binding to ACE2. Notable are the complementarities in the charges of several residues in S, e.g. R426 and N473 with those of ACE2, e.g. E329 and Q24, respectively. One can reason that these residues might contribute significantly for the on rate constant and proper orientation of the two molecules in the complex, as well as to the low dissociation rate constant. We identified two hot spots, residues R426 and N473, which are likely to contribute to the bulk of the free energy of interaction. Further studies are required for the elucidation of the energy profile of the S RBD-ACE2 interaction.
We found that not only glycosylation of the three sites in the previously described RBD-containing fragments is dispensable for expression (except one that can be any) but it also does not affect binding to ACE2. Indeed all glycosylation sites are localized at the N-terminal portion of the RBD and are relatively close to each other not only in the sequence (residues 318, 330 and 357) but also in the 3D space (Fig. 6). We constructed a fragment (319–518), which contains only two glycosylation sites and still binds with an affinity undistinguishable from the fragments containing three glycosylation sites. Further mutations of all combinations of these sites revealed that only one of them is required for expression but none of them for binding. Therefore the S RBD contacts ACE2 by an area lacking carbohydrates, which is in agreement with the recently solved crystal structure of the S RBD (Li, F, Li, W, Farzan, M, and Harrison, S. C., submitted).
The entry of the SCV into cells can be inhibited by antibodies that bind the S glycoprotein and prevent its binding to ACE2. Such a monoclonal antibody that potently inhibits membrane fusion at nM concentrations was recently identified by screening phage display libraries [19]. This antibody competed with ACE2 for binding to the S glycoprotein suggesting that its mechanism of neutralization involves inhibition of the virus-receptor interaction. We have also identified several antibodies specific for the S RBD ([20] and Zhu and Dimitrov, in preparation). The mutants developed in this study could be useful for mapping the epitopes of the antibodies against the S RBD, most of which are likely to neutralize the virus by preventing binding to the receptor ACE2.
Most of the mutations (20) described in this study did not affect binding of the S RBD to ACE2. This finding suggests that the virus could easily mutate and escape antibodies that do not exhibit the same energy profile of binding to S as ACE2. However, further studies are required in the context of the whole oligomeric S protein to make more definite conclusions about possible mechanisms of immune evasion.
The results reported in this study could have implications for understanding the mechanisms of SCV entry, and for development of entry inhibitors, vaccine immunogens, and research tools. Future studies particularly the solution of the crystal structure of the S protein in free unbound form, and in complex with ACE2, as well as measurements of the energy profiles of binding to ACE2 and antibodies, could elucidate detailed mechanisms of the S RBD function that may help in the further development of clinically useful inhibitors and vaccines.
Methods
Plasmids and antibodies
Plasmid encoding the soluble form of ACE2, pCDNA3-ACE2-ecto, was kindly provided by M. Farzan from Harvard Medical School, Boston, Massachusetts. VTF7.3 is a kind gift from C. Broder, USUHS, Bethesda, MD. Expression vectors pSecTag2 series were purchased from Invitrogen (Carlsbad, California). The monoclonal anti-c-Myc epitope antibodies (unconjugated and conjugated to HRP) were obtained from Invitrogen (Carlsbad, CA).
Cloning of S fragments
Using the previously described S756 [6] plasmid as template, fragments S364–537 (5'-GATCGGATCCTCAACCTTT AAGTGC-3' and 5'-GATCGAATTCC AGTAC CAGTGAG-3'), S317–518 (5'-GATCGGATCCCCTAATATTACAAAC-3' and 5'-G ATCGAATTCGGTCAGTGG-3'), S317–471 (5'-GATCGGATCC CCTAATATTAC AAAC-3' and 5'-GATCGAATTCGAGCAGGTGGG-3'), S329–518 (5'-GATCGGA TCCTTCCC TTCTGTC-3' and 5'-GATCGAATTCG GTCAGTGG-3'), S329–458 (5'-GATC GGATCCTTCCCTTCTGTC-3' and 5'-GATCGAATTCGCACATTAGA TATGTC-3'), S319–518 (5'-GATCGGATCCA TTACAAACTTGTGTCC-3' and 5'-GATCGAATTCG GTCAGTGG-3'), S399–518 (5'-GATCGGATCCCCAGG ACAA ACTGG-3' and 5'-GA TCGAAT TCGGTCAGTGG-3'), and S317–493 (5'-GATCG GATCCCCTAATATTACA AAC-3' and 5'-GATCGAATTCAAGG TTGGTAGCC-3') were PCR amplified using the primers mentioned within the parentheses. The PCR amplified fragments were then directionally cloned into expression vector pSecTag 2B using the restriction enzymes Bam HI and Eco RI. The various mutations on S317–518 and S319–518 were generated using the QuickChange® XL Site Directed Mutagenesis kit (Stratagene, La Jolla, CA) following the manufacturer's protocol.
Protein expression
Various plasmids were transfected into 293 cells using the Polyfect transfection kit from Qiagen (Valencia, CA) following the manufacturer's protocol. Four hours after transfection, cells were infected with VTF7.3 recombinant vaccinia virus encoding the gene for the T7 polymerase. The soluble S fragments were obtained from the cell culture medium.
Western blotting
Loading buffer and DTT (final concentration 50 mM) were added to either S proteins concentrated from the culture supernatant using Ni-NTA agarose beads or directly to the supernatant, boiled and run on an SDS-PAGE. The monoclonal anti-c-Myc epitope antibody (Invitrogen, Carlsbad, CA) was diluted in TBST buffer and incubated with the membrane for 2 hours, washed and then incubated with the secondary antibody conjugated with HRP for 1 hour, washed four times, each time for 15 min, and then developed using the ECL reagent (Pierce, Rockford, IL).
Cell binding assay
Medium containing soluble S fragments was collected and cleared by centrifugation. Vero E6 cells (5 × 106) were incubated with 0.5 ml of cleared medium containing soluble S fragments and 2 μg of anti-c-Myc epitope antibody conjugated with HRP at 4°C for two hours. Cells were then washed three times with ice cold PBS and collected by centrifugation. The cell pellets were incubated with ABTS substrate from Roche (Indianapolis, IN) at RT for 10 min., the substrate was cleared by centrifugation, and OD405 was measured.
ELISA
For the detection of the S protein fragments, a sandwich ELISA was used in which the plate was coated with anti-His tag antibody. The S protein containing culture supernatants were added and detected with an anti-c-Myc epitope antibody. In the second ELISA, the S protein was bound to the C9-tagged ecto-domain of receptor ACE 2 that was captured on a plate coated with anti-C9 antibody (ID4). As in the previous ELISA, the S protein was detected with anti-c-myc epitope antibody. The second ELISA was used to score the binding of the various S protein fragments to the receptor ACE 2. In all experiments, the incubations with the c-myc epitope antibody were for 2 h at RT.
Sequence analysis of S RBD
Sequence similarity searches were performed using NCBI BLAST program [21] by selecting, separately, all non-redundant sequences (nr) and sequences derived from the 3-dimensional structure records from the Protein Data Bank (PDB). The BLAST analysis against nr database showed 19 SARS CoV-related sequences from different clones with identities of 97–99% from the top of the list as well as 7 different coronaviruses from other organisms which share only 20–35% sequence identities at the bottom. These sequences were collected and aligned with the sequence of SARS RBD fragment using ClustalW program [22] with default parameters. The multiple alignment sequence table was prepared by choosing the aligned sequences with optimal gaps and then a phylogram tree was constructed based on that alignment scores for the 7 different coronaviruses along with S RBD. Further, the BLAST against PDB database retrieved 5 hits and 4 of them have longer stretch of amino acids (PDB codes: 1KS5, 1K0H, 1NKG and 1QR0), which have detectable sequence similarities with different regions of SARS RBD.
Competing interests
The author(s) declare that they have no competing interests.
Acknowledgements
We thank M. Farzan for reagents, Stephen Harrison for supplying the co-ordinates of S RBD before publication and Advanced Biomedical Computing Center (ABCC), NCI-Frederick for the computing facilities.
==== Refs
Dimitrov DS Virus entry: molecular mechanisms and biomedical applications Nat Rev Microbiol 2004 2 109 122 15043007 10.1038/nrmicro817
Holmes KV SARS-associated coronavirus N Engl J Med 2003 348 1948 1951 12748314 10.1056/NEJMp030078
Lai MM Cavanagh D The molecular biology of coronaviruses Adv Virus Res 1997 48 1 100 9233431 10.1016/S0168-1702(96)01421-9
Ksiazek TG Erdman D Goldsmith CS Zaki SR Peret T Emery S A novel coronavirus associated with severe acute respiratory syndrome N Engl J Med 2003 348 1953 1966 12690092 10.1056/NEJMoa030781
Rota PA Oberste MS Monroe SS Nix WA Campagnoli R Icenogle JP Characterization of a novel coronavirus associated with severe acute respiratory syndrome Science 2003 300 1394 1399 12730500 10.1126/science.1085952
Xiao X Chakraborti S Dimitrov AS Gramatikoff K Dimitrov DS The SARS-CoV S glycoprotein: expression and functional characterization Biochem Biophys Res Commun 2003 312 1159 1164 14651994 10.1016/j.bbrc.2003.11.054
Li W Moore MJ Vasilieva N Sui J Wong SK Berne MA Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus Nature 2003 426 450 454 14647384 10.1038/nature02145
Wong SK Li W Moore MJ Choe H Farzan M A 193-amino acid fragment of the SARS coronavirus S protein efficiently binds angiotensin-converting enzyme 2 J Biol Chem 2004 279 3197 3201 14670965 10.1074/jbc.C300520200
Simmons G Reeves JD Rennekamp AJ Amberg SM Piefer AJ Bates P Characterization of severe acute respiratory syndrome-associated coronavirus (SARS-CoV) spike glycoprotein-mediated viral entry Proc Natl Acad Sci U S A 2004 101 4240 4245 15010527 10.1073/pnas.0306446101
Babcock GJ Esshaki DJ Thomas WD JrAmbrosino DM Amino acids 270 to 510 of the severe acute respiratory syndrome coronavirus spike protein are required for interaction with receptor J Virol 2004 78 4552 4560 15078936 10.1128/JVI.78.9.4552-4560.2004
Bisht H Roberts A Vogel L Bukreyev A Collins PL Murphy BR Severe acute respiratory syndrome coronavirus spike protein expressed by attenuated vaccinia virus protectively immunizes mice Proc Natl Acad Sci USA 2004 101 6641 6646 15096611 10.1073/pnas.0401939101
Bosch BJ van der ZR de Haan CA Rottier PJ The coronavirus spike protein is a class I virus fusion protein: structural and functional characterization of the fusion core complex J Virol 2003 77 8801 8811 12885899 10.1128/JVI.77.16.8801-8811.2003
Wang P Chen J Zheng A Nie Y Shi X Wang W Expression cloning of functional receptor used by SARS coronavirus Biochem Biophys Res Commun 2004 315 439 444 14766227 10.1016/j.bbrc.2004.01.076
Dimitrov DS The secret life of ACE2 as a receptor for the SARS virus Cell 2003 115 652 653 14675530 10.1016/S0092-8674(03)00976-0
Towler P Staker B Prasad SG Menon S Tang J Parsons T ACE2 X-ray structures reveal a large hinge-bending motion important for inhibitor binding and catalysis J Biol Chem 2004 279 17996 18007 14754895 10.1074/jbc.M311191200
Guy JL Jackson RM Acharya KR Sturrock ED Hooper NM Turner AJ Angiotensin-converting enzyme-2 (ACE2): comparative modeling of the active site, specificity requirements, and chloride dependence Biochemistry 2003 42 13185 13192 14609329 10.1021/bi035268s
Prabakaran P Xiao X Dimitrov DS A model of the ACE2 structure and function as a SARS-CoV receptor Biochem Biophys Res Commun 2004 314 235 241 14715271 10.1016/j.bbrc.2003.12.081
Li W Zhang C Sui J Kuhn JH Moore MJ Luo S Receptor and viral determinants of SARS-coronavirus adaptation to human ACE2 EMBO J 2005 24 1634 1643 15791205 10.1038/sj.emboj.7600640
Sui J Li W Murakami A Tamin A Matthews LJ Wong SK Potent neutralization of severe acute respiratory syndrome (SARS) coronavirus by a human mAb to S1 protein that blocks receptor association Proc Natl Acad Sci U S A 2004 101 2536 2541 14983044 10.1073/pnas.0307140101
Zhang MY Choudhry V Xiao X Dimitrov DS Human monoclonal antibodies to the S glycoprotein and related proteins as potential therapeutics for SARS Curr Opin Mol Ther 2005 7 151 156 15844623
McGinnis S Madden TL BLAST: at the core of a powerful and diverse set of sequence analysis tools Nucleic Acids Res 2004 32 W20 W25 15215342 10.1093/nar/gnh003
Chenna R Sugawara H Koike T Lopez R Gibson TJ Higgins DG Multiple sequence alignment with the Clustal series of programs Nucleic Acids Res 2003 31 3497 3500 12824352 10.1093/nar/gkg500
Lee B Richards FM The interpretation of protein structures: estimation of static accessibility J Mol Biol 1971 55 379 400 5551392 10.1016/0022-2836(71)90324-X
|
16122388
|
PMC1236967
|
CC BY
|
2021-01-04 16:38:58
|
no
|
Virol J. 2005 Aug 25; 2:73
|
utf-8
|
Virol J
| 2,005 |
10.1186/1743-422X-2-73
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-541611783410.1186/1477-7819-3-54ReviewGuidelines, guidelines and more guidelines: And we still do not know how to follow-up patients with breast cancer Heys Steven D [email protected] Shailesh [email protected] Andrew W [email protected] Tarun K [email protected] Section of Surgical Oncology, Department of Surgery, University of Aberdeen, Medical School, Foresterhill, Aberdeen, Scotland, AB25 2ZD, UK2005 23 8 2005 3 54 54 2 5 2005 23 8 2005 Copyright © 2005 Heys et al; licensee BioMed Central Ltd.2005Heys et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
A major challenge facing us is the provision of health care and appropriate allocation of available resources for the treatment of patients with breast cancer. This is of particular concern in the provision of follow-up care. With the increasing incidence of breast cancer and the improvements in survival which have resulted in up to 75% of patients surviving for more than five years, an increasing resource is required. However, there is controversy as to the most appropriate schedule for follow-up of these patients. This brief review has focused on the evidence-base and guidelines that currently exist in the United Kingdom for the follow-up of patients who have been treated for breast cancer.
Methods
A review of the current guidelines published in the United Kingdom for the follow-up of patients with breast cancer (National Institute for Clinical Excellence, Scottish Intercollegiate Guidelines Network, British Association of Surgical Oncology) and the randomised controlled trials evaluating the follow-up of patients with breast cancer was undertaken.
Results
The results have demonstrated the different follow-up protocols currently indicated in these guidelines within the same country. Furthermore, the lack of well designed, randomised controlled trials on which to base a follow-up protocol for patients with breast cancer is apparent.
Conclusion
The evidence-base on which these guidelines have been developed is lacking. It is apparent that well designed randomised controlled trials are needed urgently if we are to understand the most appropriate and effective ways of following up patients with breast cancer.
==== Body
Background
A major challenge in the provision of healthcare throughout the world in the 21st century is trying to ensure that the resources that are available meet the demand. Nowhere is the situation more acute than in the provision of care for patients with cancer. Whilst current statistics show that one in three people in the world will develop a malignant disease, this figure is the projected to increase dramatically during the next 10 years.
Of particular concern has been the continual rise in the incidence of breast cancer. Each year the incidence increases by approximately 2% and in the UK alone there are 45,000 new cases per annum [1]. However, whilst the incidence of breast cancer is increasing there have been many improvements and developments in surgery, radiotherapy, chemotherapy and hormone therapy for patients with breast cancer [2]. The resultant improvements in survival are well recognised, and in Scotland, for example, the five year survival has risen now to 75% [3]. These improvements in survival are most welcome but we do need to consider the utilisation and allocation of resources for the follow-up of these patients and their concerns as to their appropriate use.
Therefore, how should we follow-up patients with breast cancer who have undergone apparently curative therapy? Firstly, the natural history of the disease, i.e. the probability of local and/or distant recurrence of disease and the psychological morbidity must be considered in this regard. Secondly, the short-term and longer-term effects on the patient of the various treatments that they have been given (surgery, radiotherapy, chemotherapy, hormone therapy) must also be taken into account. For example after surgery, wound complications, postoperative pain, lymphoedema and disorders of body image occur. After chemotherapy and hormone therapy, there is risk of cardiac dysfunction, neurotoxicity, premature menopause, osteoporosis, osteoporotic fractures and psychosocial disturbances that must be taken into account. Furthermore, the possibility of a new primary cancer in the ipsilateral, or contralateral, breast is also important.
In terms of the disease itself, approximately 25% of patients will develop a systemic recurrence and die within five years. Importantly, 60% to 80% of all recurrences that occur are found during the first 3 years after treatment of the primary tumour in the breast [4,5]. Furthermore, in approximately three quarters of patients who develop disease recurrence, there are symptoms experienced by the patient or there are abnormalities on clinical examination to indicate recurrence of disease [4]. The risk of local disease recurrence in a breast which has been treated by breast conservation surgery, or the risk of a second primary breast cancer occurring in either breast, is approximately 0.5% to 1% per annum, every year, following completion of treatment [5].
Given this information, follow-up care for patients with breast cancer has been directed primarily towards detecting local and lymph node recurrence of disease and also with the aim of detecting metastatic disease using clinical examination and radiological and laboratory tests. However, three key questions should be considered when planning a follow-up programme for patients with breast cancer:
• How effective are regular hospital visits, clinical examination and laboratory investigations in detecting disease recurrence, and how effective is mammography in identifying local recurrence or second primary cancers in the breast?
• If disease recurrence (local, regional systemic) is identified can the patients' outcome in terms of survival be altered?
• If there is an effect on patients' outcome and survival, then what is the optimal schedule of investigations in order to achieve this?
What is the evidence?
A major limitation in trying to answer these questions is the quality of the evidence that is available. Whilst there are many retrospective and prospective observational studies of follow-up of breast cancer patients, these are all open to a variety of biases, which severely limit the interpretation of these observations. The only way in which we can answer the questions about follow-up is through well designed, adequately powered and well conducted clinical trials. At present there are few randomised controlled trials in the follow-up of patients with breast cancer that are available.
However, firstly can we affect the outcome of patients by detecting recurrent disease at an early stage by using an intensive follow-up schedule? In order to answer this question, two trials have examined more than 2,500 patients who were randomised to a follow-up schedule of clinical examination plus mammography or to a more intensive follow-up schedule of laboratory tests combined with radiological imaging. The protocols for these are shown in Tables 1 and 2[6-8].
Table 1 Randomised trial of intensive schedule versus standard schedule of follow-up of patients in the "GIVIO" trial
Intensive follow-up (n = 655) Standard follow-up (n = 665)
• Physical examination every 3 months for 2 years; then 6monthly for 3 years • Physical examination every 3 months for 2 years; then 6monthly for 3 years
• Serum biochemistry at each clinical examination • Mammography annually
• Chest x-ray every 6 months
• Isotope bone scan annually
• Liver ultrasound annually
• Mammography annually
(study detailed in JAMA 1994; 271: 1587–1592)
Table 2 Schedule for follow-up in the roselli del turco trial of intensive follow up
Intensive follow-up (n = 622) Standard follow-up (n = 621)
• Physical examination every 3 months for 2 years; then 6monthly for 3 years • Physical examination every 3 months for 2 years; then 6monthly for 3 years
• Chest x-ray every 6 months • Mammography annually
• Isotope bone scan every 6 months
• Mammography annually
(JAMA 1999; 281; 1586 and JAMA 1994; 271: 1593–1597)
When the results of these trials were pooled together and examined there was found to be no significant difference in the five-year disease-free survival or overall survival for either group of patients [9]. However, there was a difference in the detection of asymptomatic metastatic disease. In patients followed-up intensively 31% had asymptomatic metastases compared with only 21% in those being less intensively followed-up. Therefore, although an intensive follow-up schedule will detect metastatic disease earlier it does not impact on the patients' outcome with respect to disease-free and overall survival. In terms of quality of life, there were no differences between the patients in having intensive or less intensive follow-up schedules.
Furthermore, a recent systematic review and meta-analysis has focused on whether routine hospital visits were even effective in detecting loco-regional recurrences in patients who were asymptomatic following treatment for early breast cancer [10]. A total of 5,045 patients from 12 studies were analysed. It was found that in asymptomatic patients, only 40% of isolated loco-regional recurrences were diagnosed by routine visits and tests [10] but the difficulties with interpretation of the studies due to their poor quality was also clear from this analysis. However, the majority of recurrences were identified outside the patients planned routine follow-up schedule.
Another question that is now being asked, particularly in view of the lack of oncologists in many areas in the world, is what is the value of patients attending a "specialist" follow-up clinic? In a trial designed to address this question, after treatment for breast cancer 296 patients were randomised to follow-up, either by their general practitioner or by hospital specialist [11]. Although this was a small study, it was stated that there was no significant difference in the detection of metastases in the two groups of patients. However, there was a 60% increased detection in the group of patients being followed-up by the hospital specialists, although this was dismissed because it did not achieve statistical significance. Another important aspect of the findings from this study was that there was no difference in the patients' quality of life. Furthermore, patients in the general practitioner group were more satisfied with the continuity of care than patients' follow-up by hospital specialists. The limitations of this study, particularly in respect of the statistical power and short period of follow-up do limit the conclusions that can be drawn. Furthermore, one third of eligible patients declined to participate in this study. A further well designed and larger clinical study is necessary to address this issue.
An alternative approach to the follow-up was evaluated in a Swedish multicentre study where the role of nurse-led follow-up was examined [12]. A small group of 264 patients with early breast cancer were randomised to routine physician follow-up or follow-up by a nurse specialist. There was no difference in terms of patient satisfaction, anxiety, or depression and there were no differences between time to recurrence or death between the two groups of patients. The authors did note that the study was small and not powered to detect differences in recurrence and survival. On the basis of these encouraging results further studies would be required to confirm that this is an alternative way of follow-up for certain selected patients which may offer advantages in terms of continuity of care, patient education and allow a more appropriate utilisation of physician-time [12].
A key consideration is what do the patients themselves want? One small, randomised trial of 211 patients has addressed this issue [13]. Patients were randomised to have either a conventional follow-up clinic visit schedule (every 3 months for the first year, four months for the second year, six months up until 5 years after initial diagnosis and annually thereafter) or just to have mammography at routine intervals (initially every year for 5 years, then two yearly). Those patients who had undergone a mastectomy had a slightly different mammographic schedule with a mammogram one year after diagnosis and then every two years subsequently [13].
The results of this study revealed that approximately twice as many patients felt they would rather have a reduced schedule of follow-up rather than a more intensive one. Also the patients in the groups randomised to just mammographic follow-up were satisfied with this. However, a very important point to emerge was that this was not a universal finding amongst all the patients in the study. Importantly, those patients who were less than 50 years of age, who were at a stage between two and five years after diagnosis and those who had aggressive disease, were less likely to participate in this study [13].
Just how do we follow-up patients and what are the guidelines?
Given the data from this small number of randomised trials is it possible to decide upon, in an evidence-based fashion, the most appropriate follow-up for patients with breast cancer? The key facts to consider in attempting to do this and which have emerged from these randomised controlled trials are:
• 25% of patients develop recurrence and in 60% to 80% of them this will be in the three years following initial diagnosis,
• Hospital outpatient visits may detect 40% of locoregional recurrences in asymptomatic patients,
• A very intense schedule of hospital visits, laboratory and imaging tests does not affect disease-free or overall survival compared with a less intense schedule of clinical examination and mammography,
• Follow-up in hospital clinics compared with general practice follow-up does show a trend towards an increased detection of metastatic disease but quality of life was no different although the longer term effects on survival are unclear
• Patients' preference is for less intense follow-up with the exception of younger patients, those with more aggressive disease and those who are at a stage of between two to five years from their initial diagnosis.
With these considerations in mind, clinical practice guidelines which have been developed and defined as "systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances" [14], have been developed for the follow-up of patients with breast cancer. These guidelines have been developed on the basis of systematic reviews of the literature and in each guideline there is usually an explanation of the classification of levels of evidence (e.g. from meta-analyses, systematic reviews of RCTs, case-control studies, cohort studies, non-analytic studies and expert opinion), classification of grades of recommendation (e.g. based on the results from meta-analyses of RCT, or from a high quality RCTs, or from case control or cohort studies, or based on non-analytical studies or expert opinion) and also the guideline will state the time when it is due to be updated.
In recent years in the United Kingdom, a number of different guidelines have been produced in an attempt to ensure that patients with breast cancer have the most appropriate follow up. For example, the Association of Breast Surgery at the British Association of Surgical Oncology (BASO), the Scottish Intercollegiate Guidelines Network (SIGN), and the National Institute of Clinical Excellence, the Royal College of Radiologists and the Clinical Outcomes Group of the Department of Health have all tried to provide guidance as to the most appropriate way of follow-up [15-19]. The key points to emerge from these guidelines are shown in Table 3, and have their basis in the data outlined above. In addition to these guidelines from the UK, a variety of other organisations throughout the world have also produced their own guidelines, which consider follow-up of patients with breast cancer. It is beyond the scope of this article to consider these in detail but can be accessed elsewhere [20]. However, most recommend clinical examination every 3 to 6 monthly for 3 to 5 years, and then followed by annual clinical examinations. As regards mammography, the trials addressing this seems to recommend mammography six months after completion of radiotherapy and then at 1 or 2 yearly intervals thereafter.
Table 3 Guidelines issued in the united kingdom for the follow-up of patients with breast cancer
Organisation Recommendation
The Association of Breast Surgery at the British Association of Surgical Oncology • Patients on active treatment may be followed up until such treatment has been completed
• High risk patients may be followed up more closely with joint care by surgeons and oncologists according to local protocols
• Data about long term follow-up is essential in monitoring clinical outcomes
• Patients to be followed up for 5 years
• Routine mammography every 1 to 2 years for 10 years after diagnosis
Scottish Intercollegiate Guidelines Network (SIGN) • These guidelines state that there is insufficient clinical evidence to determine the optimal interval of clinical examination. They suggest that a "pragmatic schedule" should be adopted, for example, every 6 months for 2 years and then annually thereafter.
• For mammographic follow-up, in a breast which has been conserved, then this should be performed at least every 2 years and at intervals of not less than 1 year. For the contralateral breast mammography should be carried out every 1 – 2 years.
National Institute for Clinical Excellence (NICE) • Guidelines state that there should be a "limited" follow-up for 2 – 3 years and should be agreed by "local networks". This would not normally exceed 3 years unless patients were in clinical trials.
• The guidelines state that local networks should agree evidence-based policy for the frequency of mammographic follow up
The Royal College of Radiologists • Guidelines recommend that mammography is carried out at least every 2 years and not more than annually
The Clinical Outcomes Group, Department of Health • Recommends that mammography is carried out annually for 5 years and then every two years after that
It is always difficult for clinicians, particularly when several different guidelines exist in one county. However, it is easier to consider the areas common to these guidelines initially. All the guidelines agree that follow-up should be limited to clinical examination and mammographic surveillance with their being no recommendations for other laboratory or radiological imaging tests. But what is the interval as which these should be carried out and for how long? The difficulty is, of course, that there is no evidence on which to make these recommendations.
However, perhaps a clinical examination every six months for five years would be a reasonable recommendation but with the caveat of the lack of information available. Even more difficult to base on scientific evidence is the optimal interval for mammographic surveillance. Again we do not have the evidence but we must consider the randomised trials, what we understand about the biology of breast cancer and recurrence, and what we know from the various breast screening programmes that have been implemented internationally. Given this data an annual mammogram for five years and then every two years after breast conserving surgery seems a reasonable compromise. In patients who have had a mastectomy, then a surveillance mammogram every two years also seems to be a reasonable compromise for follow-up in these patients. However, we emphasise the lack of evidence for recommendations regarding mammographic surveillance.
This still leaves two important questions as to how long the patients should be followed-up for and by whom? Firstly, in terms of duration of follow-up, although we do not have the evidence, the National Institute of Clinical Excellence (NICE) recommends a limited period of follow-up of 2 to 3 years unless patients are entered into clinical trials. NICE also documents the financial savings that would accrue if such a policy were adopted [17]. However, there are many unanswered questions as to the appropriateness of this that needs to be answered. As regards who should follow-up the patient, one trial indicated that in selected patients this could be the general practitioner, but it is worth noting important exceptions as discussed above. Furthermore, the longer-term consequences of this approach remain unclear.
Another complicating factor with regard to follow-up is that several recently published trials have had a major impact on the usage of adjuvant hormone therapy, eg the indication that the armomatase inhibitors may be superior to tamoxifen, the value of further treatment with letrozole after 5 years of tamoxifen and the impact of changing patients after two or three years tamoxifen treatment to an aromatase inhibitor. Therefore, at the present time we are still not sure what the optimum for adjuvant hormonal therapy is and we also need to consider what should happen to the patients who are currently taking tamoxifen.
We also need to consider what will be the impact on bone mass with the increasing use of aromatase inhibitors and whether or not 'prophylactic' bisphosphonates may be required. Furthermore, intensive folow up does increase the rate of detection of asymptomatic metastases and whilst this did not impact on patients survival in the studies from the 20 years ago, with the current advances in systemic treatment (eg aromatase inhibitors, taxanes, trastuzumad etc) is it not possible that early treatment may now have survival advantages for these patients? One must not under estimate the need for continuing audit of results following the treatment for breast cancer and of the morbidity in these patients undergoing 'multi-modality therapy' which can have significant short and long term consequences. These are just some of the issues to be considered when following-up patients following treatment for breast cancer and therefore with increasing complexity of management it is likely that specialist input will be necessary still.
Conclusion
Despite the advances that have been made in the treatment of patients with breast cancer we are still unclear as to the optimal way in which patients should be follow-up once treatment has been completed. Despite the publication of many different guidelines with recommendations for follow-up it is clear that the evidence base on which these are founded is lacking at the present time. Furthermore, the evidence is from trials that were commenced twenty years ago and their relevance to the modern management of patients with breast cancer is now questionable.
It is now important that we consider the research priorities in follow-up of breast cancer patients, in particular with respect to stratifying patients according to their risk of disease recurrence, and the impacts of treatments on physical and psychological morbidity and quality of life. It is essential that well designed randomised controlled trials are undertaken if we are to understand the most appropriate and effective ways of following up patients with breast cancer.
Competing interests
SDH is a member of the national committee of the Association of Breast Surgery at the British Association of Surgical Oncology and in this regard was involved in the process of development of the Association of Surgery guidelines which are referred to in this paper
Authors' contributions
SDH initiated the manuscript and the literature review.
SDH, SC, AWH and TKS drafted the manuscript and all authors approved the final manuscript
Acknowledgements
Nil
==== Refs
Cancer Research, UK. Cancer facts and figures accessed on August 17, 2005
Early Breast Cancer Trialists' Collaborative Group Systemic treatment of early breast cancer by hormonal, cytotoxic, or immune therapy Lancet 1992 339 1 15 71–85 1345950
Scottish health statistics. Information. Breast cancer accessed on August 17, 2005
Huston TL Simmons RM Locally recurrent breast cancer after conservation therapy Am J Surg 2005 189 229 235 15720997 10.1016/j.amjsurg.2004.07.039
Cajucom CC Tsangaris TN Nemoto T Driscoll D Penetrante RB Holyoke ED Results of salvage mastectomy for local recurrence after breast-conserving surgery without radiation therapy Cancer 1993 71 1774 1817 8448741
Rosselli Del Turco M Palli D Cariddi A Ciatto S Pacini P Distante V Intensive diagnostic follow-up after treatment of primary breast cancer: A randomized trial. National Research Council Project on Breast Cancer follow-up JAMA 1994 271 1593 1597 7848404 10.1001/jama.271.20.1593
Palli D Russo A Saieva C Ciatto S Rosselli Del Turco M Distante V Pacini P Intensive diagnostic follow-up after treatment of primary breast cancer; 10-year update of a randomised trial JAMA 1999 281 1586 10235147 10.1001/jama.281.17.1586
Impact of follow-up testing on survival and health-related quality of life in breast cancer patients. A multi-center randomised controlled trials. The GIVIO Investigators JAMA 1994 271 1587 1592 8182811 10.1001/jama.271.20.1587
Rojas MP Telaro E Russo A Moschetti I Coe L Fossati R Palli D del Roselli Turco M Liberati A Follow-up strategies for women treated for early breast cancer (Review) Cochrane Database Syst Rev 2005 CD001768 15674884
de Bock GH Bonnema J van Der Hage J Kievit J van de Velde CJ Effectiveness of routine visits and routine tests in detecting isolated locoregional recurrences after treatment for early-stage invasive breast cancer. A meta-analysis and systematic review J Clin Oncol 2004 22 4010 4018 15459225 10.1200/JCO.2004.06.080
Grunfeld E Mant D Yudkin P Adewuyi-Dalton R Cole D Stewart J Fitzpatrick R Vessey M Routine follow-up of breast cancer in primary care: a randomised trial BMJ 1996 313 665 669 8811760
Koinberg IL Fridlund B Engholm GB Holmberg L Nurse-led follow-up on demand or by a physician after breast cancer surgery: a randomised study Eur Oncol Nursing Soc 2004 8 109 117 discussion 118–120 10.1016/j.ejon.2003.12.005
Gulliford T Opomu M Wilson E Hanham I Epstein R Popularity of less frequent follow-up for breast cancer in a randomised study: initial findings from the hotline study BMJ 1997 314 171 177 9022428
Field MJ Lohr KN Institute of Medicine Committee to advise the public health service on clinical practice guidelines. Clinical practice guidelines: directions for a new programme 1990 Washington DC: National Academy Press
The Association of Breast Surgery at BASO, Royal College of Surgeons of England Guidelines for the management of symptomatic breast disease Eur J Surg Oncol 2005 31 S1 S21
Scottish Intercollegiate Guidelines Network Breast cancer in women accessed on August 16, 2005
Guidance on Cancer Services Improving outcomes in breast cancer Manual update National Institute for Clinical Excellence 2002 accessed on August 16, 2005
The Royal College of Radiologists Guidelines on screening and symptomatic breast imaging, 2003 accessed on 15 August, 2005
Cancer sub-group of the Clinical Outcomes Group Guidance for purchasers: Improving outomes in breast cancer – the manual London, Department of Health 1996
Rojas MP Telaro E Russo A Moschetti I Coe L Fossati R Palli D del Roselli TM Liberati A Follow-up strategies for women treated for early breast cancer Cochrane Database Syst Rev CD001768 accessed on August 16, 2005
|
16117834
|
PMC1236968
|
CC BY
|
2021-01-04 16:39:05
|
no
|
World J Surg Oncol. 2005 Aug 23; 3:54
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-54
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-571613525110.1186/1477-7819-3-57ResearchNeoadjuvant chemotherapy versus primary surgery in advanced ovarian carcinoma Hegazy Mohamed AF [email protected] Refaat AF [email protected] Mohamed A [email protected] Ahmed E [email protected] Maged R [email protected] Mohamed [email protected] Amal AF [email protected] Surgical Oncology department, Mansoura University, Mansoura, Egypt2 Obstetrics and Gynecology department, Mansoura University, Mansoura, Egypt3 Department of Medicine, University of Pittsburgh, Pittsburgh, PA, USA2005 31 8 2005 3 57 57 5 4 2005 31 8 2005 Copyright © 2005 Hegazy et al; licensee BioMed Central Ltd.2005Hegazy et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Patients with advanced ovarian cancer should be treated by radical debulking surgery aiming at complete tumor resection. Unfortunately about 70% of the patients present with advanced disease, when optimal debulking can not be obtained, and therefore these patients gain little benefit from surgery. Neoadjuvant chemotherapy (NACT) has been proposed as a novel therapeutic approach in such cases. In this study, we report our results with primary surgery or neoadjuvant chemotherapy as treatment modalities in the specific indication of operable patients with advanced ovarian carcinoma (no medical contraindication to debulking surgery).
Patients and methods
A total of 59 patients with stage III or IV epithelial ovarian carcinomas were evaluated between 1998 and 2003. All patients were submitted to surgical exploration aiming to evaluate tumor resectability. Neoadjuvant chemotherapy was given (in 27 patients) where optimal cytoreduction was not feasible. Conversely primary debulking surgery was performed when we considered that optimal cytoreduction could be achieved by the standard surgery (32 patients).
Results
Optimal cytoreduction was higher in the NACT group (72.2%) than the conventional group (62.4%), though not statistically significant (P = 0.5). More important was the finding that parameters of surgical aggressiveness (blood loss rates, ICU stay and total hospital stay) were significantly lower in NACT group than the conventional group. The median overall survival time was 28 months in the conventional group and 25 months in NACT group with a P value of 0.5. The median disease free survival was 19 months in the conventional group and 21 months in NACT group (P = 0.4). In multivariate analysis, the pathologic type and degree of debulking were found to affect the disease free survival significantly. Overall survival was not affected by any of the study parameters.
Conclusion
Primary chemotherapy followed by interval debulking surgery in select group of patients doesn't appear to worsen the prognosis, but it permits a less aggressive surgery to be performed.
==== Body
Background
The diagnosis and management of ovarian cancer is one of the greatest challenges in oncology. Approximately, half of ovarian carcinoma patients die from the disease making it the leading cause of gynecologic cancer-related death in most industrialized countries [1].
Although our approach and knowledge of epithelial ovarian cancer has changed in the past 25 years, the overall survival has not been affected as approximately 65% to70% of all cases continue to be diagnosed with stage III or stage IV disease. Surgical reduction of tumor bulk has become the preferred first step in the management of advanced epithelial ovarian cancer [2]. Observations that the excision of large tumor masses could provide palliation and a modest extension of life have been recorded for more than 50 years. Enhancement of sensitivity to chemotherapy, yet unproven, could be the greatest benefit of tumor debulking [3].
Approximately 70% of patients present with advanced ovarian cancer, when optimal debulking can not be obtained, and therefore gain little benefit from surgery [4]. On the other hand, patients who are severely compromised medically carry an unwarranted risk to surgery. Neoadjuvant chemotherapy (NACT) has been proposed as a novel therapeutic approach to a variety of solid tumors when the disease is not amenable to surgical resection at the time of diagnosis or the patient is unfit for aggressive debulking surgery [5]. NACT has now been recognized as a useful modality for the treatment of various advanced cancers [6,7]. In cases with advanced ovarian carcinomas, platinum based chemotherapy regimens have been found to produce higher response rates and in some studies have produced a statistically significant survival advantages compared with drug regimens without platinum [8,9].
Thus the two treatment options available for treating patients with advanced ovarian tumor are either a primary surgical cytoreduction or to start with chemotherapy hoping for down staging the tumor and then go ahead with surgery.
In this study, we report our results with these two treatment modalities applied only to operable patients without medical contraindications to surgery. We assessed the patients for different variables, such as the ability to perform optimal debulking, rate of non-standard surgery (Excision of more than one organ), disease free survival and overall survival.
Patients and methods
This prospective trial included a total of 59 patients with stage III or IV epithelial ovarian carcinomas that were evaluated between 1998 and 2003. Patients who were selected for our study had advanced ovarian carcinoma and were free from severe concomitant medical illness that could preclude surgical interference (such as those with WHO performance status 2, or 3). Patient characteristics are summarized in Table 1. Patients were subjected to physical examination, serum level of CA 125 measurement, radiological studies, and histopathological confirmation of ovarian carcinoma. All patients were submitted to surgical exploration at the Surgical Oncology Unit, Mansoura University Hospitals. The purpose of this exploration was to evaluate tumor resectability; to perform primary debulking surgery when optimal cytoreduction seemed feasible and to treat primary unresectable tumors with neoadjuvant chemotherapy. Optimal debulking has been variously defined, however we adopted the Gynecologic Oncology Group definition which defines it as leaving residual disease of less than 1 cm [10]. This strategy was explained to the patients and informed consents were obtained. Surgical exploration was usually done laparoscopically (38 cases) unless it was contraindicated, when laparotomy was done (21 cases). Neoadjuvant chemotherapy was given (in 27 patients) when we considered that optimal cytoreduction was not feasible with the standard surgery, defined as 1) total abdominal hysterectomy with bilateral salpingoophorectomy, 2) appendectomy, 3) total infragastric omentectomy, 4) peritonectomy limited to the pelvis, paracolic gutters, anterolateral diaphragmatic area, and 5) pelvic, common iliac, and infrarenal paraaortic lymphadenectomy. Conversely primary debulking surgery was performed when we considered that optimal cytoreduction could be achieved by the standard surgery (32 patients). However, in a few cases non-standard surgery, meaning a single organ resection (e.g., small intestine, colon, spleen) in the way to achieve an optimal cytoreduction was adopted. All patients who underwent intestinal surgery were evaluated by a single surgeon (MH).
Table 1 Patient characteristics
Conventional(n = 32) NACT(n = 27) p
Age 53.6 ± 9.8 58.7 ± 4.6 NS
FIGO stage
IIIc 14 (43.8%) 11 (40.7%) NS
IV 18 (56.2%) 16 (59.3%) NS
Grade
II 14 (43.8%) 12 (44.4%) NS
III 18 (56.2%) 15 (55.6%) NS
Histological type:
Serous 9 (28.1%) 7 (25.9%) NS
Mucinous 13 (40.6%) 10 (37%) NS
Undifferentiated 10 (31.3%) 10 (37%) NS
Staging procedure
Laparoscopy 21 (65.6%) 17 (62.9%) NS
Laparotomy 11 (34.4%) 10 (37.1%) NS
Chemotherapy regimens were all platinum based and included cisplatinum 75 mg\m2 plus cyclophosphamide 600 mg\m2 (repeated every 3 weeks). This regimen was applied to all cases of the NACT group and was applied in all cases of the conventional group postoperatively.
The response to neoadjuvant chemotherapy was evaluated after the third cycle. Clinical response to chemotherapy was evaluated on clinical examination, serum CA 125 level, and computed tomography (CT) scan. Tumor response was classified according to the WHO criteria. [11]. Patients were then referred for second surgical cytoreduction when they presented no signs of progression during chemotherapy (18 cases). The surgicopathologic response to NACT was assessed at secondary surgery. Nine patients progressed under chemotherapy and were not surgically operated. The aggressiveness of surgical cytoreduction was evaluated in terms of the blood loss rates, and the length of intensive care unit and postoperative hospital stay. A report of the peroioperative and postoperative complications in both groups was recorded.
Survival curves since diagnosis (first surgical procedure) were calculated according to the Kaplan Meier method, and survival curves were compared by the log-rank test.
Results
Between April 1998 to January 2003, 59 patients presented with operable, locally advanced epithelial ovarian carcinoma. After surgical exploration, 32 patients seemed resectable and primary cytoreductive surgery was carried out, and 27 patients seemed unresectable and neoadjuvant chemotherapy was given to them. Among those patients, 9 progressed during chemotherapy and were not operated (Figure 1)
Figure 1 Treatment plan according to the results of initial exploration.
Patient characteristics are summarized in Table 1. When comparing the NACT patients with the conventionally treated patients as a group, the NACT group were statistically older (58.7 ± 4.6 years vs. 53.6 ± 4.6 years) than the conventional group (P = 0.04)
The staging procedure was laparoscopy in 38 patients and laparotomy in 21 patients. In the NACT group (n = 27), the mean interval between surgical staging and the start of chemotherapy was 13.7 days (range 5–33) days after laparoscopy and 18.9 days (range 7–41 days) after laparotomy (p = 0.055).
All patients then received platinum-based chemotherapy. The median number of neoadjuvant chemotherapy cycles was 3 (range 2–6). Only 2 patients showed partial response after 2 cycles and cytoreductive surgery was done. Twenty-five patients received 3 or more cycles. Among 27 patients of this series, 18 (66.6%) responded to NACT according to clinical examination, serum CA 125 level and abdominopelvic CT scan. All of them showed partial response. Conversely 9 patients showed no response (4 cases showed a stable disease and 5 cases showed progressive disease). Those 9 cases were not operated and referred to continue chemotherapy.
We aimed to compare the cases who were primarily operated (n = 32) and those who were operated after responding to neoadjuvant chemotherapy (n = 18) in terms of the degree of optimal debulking and the morbidity associated with surgical procedures.
In NACT group, optimal cytoreduction was achieved in 13 cases. Thus the optimal debulking rate was 48.1% among the overall number of patients in this group (n = 27) and 72.2% among those who were operated (n = 18). In the conventional group the optimal debulking rate was (62.4%). The difference was statistically insignificant (P = 0.5). However, one must remember that 9 patients progressed while on chemotherapy and were not operated (Table 2).
Table 2 Degree of debulking in the conventional group and the surgically operated patients in NACT group
Conventional Group (n = 32) NACT (n = 18)
Optimum cytoreduction 20(62.5%) 13(72.2%)
Suboptimal cytoreduction 12(37.5%0 5(27.8%)
In the conventional group, non-standard surgery was performed in 11 cases (34.4%), and in 4 cases (27.8%) of NACT group (Table 3). Resection and primary anastomosis of the small intestine occurred in 10 patients, partial cystectomy was done in 3 cases, colectomy was done in 4 cases, and splenectomy in 2 cases. Multiple organ resections (MOR) occurred in two cases in the conventional group and in one case in NACT group.
Table 3 The frequency of non-standard surgeries in both groups
Conventional Group (n = 32) NACT Group (n = 18) p
Number of patients 11 (34.4%) 4 (22.2%) NS
Organs resected
Small intestine 7 3
Colon 3 0
Bladder 2 1
Spleen 1 1
Patients in NACT group showed a significantly less blood loss rates (p = 0.02), less ICU stay (p = 0.03), and less total hospital stay (p = 0.05). There was no difference between perioperative morbidity and mortality in the two patient groups (Table 4). Complications were more in the cases that underwent intestinal surgery (3 cases of wound infection, 6 cases of fever, and 3 cases of DVT).
On assessment of the survival we compared the whole number of both groups i.e., we added the 9 patients who were not surgically operated in NACT group. The median overall survival time was 28 months in the conventional group and 25 month in the NACT group with an insignificant P value (P = 0.5) (Figure 2). The median disease free survival was 19 months in the conventional group and 22 months in the NACT group with an insignificant P value (P = 0.4) (Figure 3)
Table 4 Parameters of surgical morbidity in both groups
Conventional Group (n = 32) NACT Group (n = 18) p value
Duration of surgery
Mean 190 150 NS
range 70–350 90–270
Blood loss rates (cc)
Mean 735 420 0.02
Range 50–5000 50–3000
ICU stay (days)
Mean 4.4 1.7 0.03
Range 1–9 1–5
Hospital stay (days)
Mean 15.9 10.5 0.05
Range 6–49 4–31
Wound infection 2 2 NS
Fever > 38.5°C > 3 d 7 1 NS
Atelectasis 1 1 NS
Pleural effusion 2 0 NS
DVT 3 1 NS
Figure 2 Overall survival in NACT and conventional groups.
Figure 3 Disease free survival in NACT and conventional groups.
In multivariate analysis of both the conventional group and NACT group, the overall survival was not significantly affected by any of the study parameters (pathologic type, grade, stage, degree of optimal debulking). In the conventional group, the disease free survival was significantly affected with the degree of optimal cytoreduction only (P = 0.001) (Figure 4). In the NACT group the disease free survival was significantly affected by the tumor type (P = 0.02) and the degree of optimal debulking (P = 0.01).
Figure 4 Disease free survival in both groups according to the degree of debulking.
Discussion
The clinical basis of aggressive cytoreductive surgery in the initial management of ovarian cancer is the significantly improved survival gained by those patients in whom optimal cytoreductive surgery was accomplished [12,13]. The presence of residual disease after surgery is one of the most adverse prognostic factors for survival. Therefore, although the definition of optimal cytoreduction has been modified over the last two decades, it is generally agreed that every attempt should be made to surgically resect as much disease as safely possible [4].
The value of debulking after induction chemotherapy has been largely debated in the last decades. Recently several investigators introduced the concept of interval debulking surgery meaning a surgical procedure with debulking intent foreword and followed by cytoreductive chemotherapy [14,15]. Based on the GOG 152 data, interval debulking surgery does not seem to be indicated in patients who underwent primarily a maximal surgical effort by a gynecological oncologist [15].
In a population of patients with advanced ovarian carcinoma who deemed unresectable by surgical exploration, neoadjuvant chemotherapy helped to select patients for feasible and relatively less aggressive interval debulking. Patients who did not respond or progressed under chemotherapy were spared surgery [16].
An issue of importance is which criteria should be used to define the respectability of the tumor and consequently the selection of which patients might benefit from NACT approach. Imaging (computed tomography scan) based criteria have been developed by different authors [17,18]. Nelson et al [17] showed that the predictive value of a positive test (CT scan demonstrating non respectability) was only 67%. Bristow et al developed a predictive index that was able to correctly predict surgical outcome (optimal < 1 cm versus suboptimal residual disease status) [19]. The specificity or the ability to identify patients undergoing optimal debulking was 80%. The authors agree with Ansquer et al [8] and Vergote et al [14] in that a laparoscopy and in certain situations exploratory laparotomy provides certain advantages as a selection tool. It allows for making a histological diagnosis and objectively documents the extent of the disease. At the same time it identifies patients who can be optimally debulked, thus not denying the possible benefit of such a procedure. The issue of port site implantation in this patient group can probably be addressed by proper technique (Closure of the peritoneum and excision of trocar port site at the definitive surgery) and immediate (< 1 week) start of chemotherapy [19]. In this study we performed laparoscopy as a selection tool in most patients unless it was contraindicated when a laparotomy was done. The limits of standard debulking surgery that were found at exploration were extensive bowel involvement, large involvement of the peritoneum located in the upper abdomen particularly in the dorsal diaphragmatic area, and liver metastasis. These cases were referred for neoadjuvant chemotherapy. Initiation of chemotherapy was significantly delayed in the laparotomy group than the laparoscopy group. No case presented with port site recurrence in the laparoscopy patients.
Chemotherapy was platinum based. The number of preoperative cycles ranged from 2–6. Most patients were explored after 3 cycles. It is noteworthy to mention that three patients who were explored after 5–6 cycles tended more frequently to present no microscopic disease. Indeed the optimal number of chemotherapy cycles to be given before planned surgery is still a major, unresolved issue. In previous published studies the number of preoperative chemotherapy cycles ranged from 2–10 [20-22]. It seems that the chance of achieving an optimal debulking increases in responding patients with the numbers of cycles before surgery [23]. This potential advantage has to be balanced against the risk of emergence/ selection of drug resistant cell clones and cumulative drug toxicity associated with the increased number of chemotherapy cycles [20].
Eighteen patients of NACT group had a clinical response. Optimal cytoreduction rate was 48.1% among the overall number of patients in this group (n = 27) and 72.2% among those who were operated (n = 18). This correlates with previous reports of Jacob et al [12] who reported optimal cytoreduction in 77 % of patients and Surwit et al [24] who reported 55% of cytoreduction less than 1 cm.
We believe that the benefit of neoadjuvant chemotherapy does not lie in its ability to obtain larger percentage of optimal cytoreduction because the increased and the more widespread use of newer technologies as ultrasonic aspirator, argon beam coagulator and ultra radical surgical procedures could increase the fraction of patients who are optimally debulked upfront but at the likely cost of increase morbidity [24].
The value of neoadjuvant chemotherapy is to obtain optimum cytoreduction by means of less aggressive surgery. In our study, debulking surgery in NACT group was less aggressive than in the conventional group with less blood loss rates, shorter intensive care stay and shorter postoperative hospitalization. These finding are consistent with the data of Schwartz et al [23] who reported that the aggressiveness of debulking surgery seems to be decreased after neoadjuvant chemotherapy. There was no significant difference between perioperative morbidity and mortality in the two patient groups.
In our study, the median overall survival time yields no significant difference in both groups. Onnis et al [25] described 88 patients treated with NACT compared with 248 patients treated with upfront surgery followed by chemotherapy. The overall survival was not improved. In an analysis by Surwit et al, the median survival of 29 patients who underwent primary chemotherapy was 22 months, which the author said was similar to that of patients who undergo primary surgery [24]. Schwartz et al [21] reported on 59 patients treated with NACT of whom 41 were eventually operated compared with a control group of conventionally treated patients, the patients receiving NACT were significantly older and had a poor performance status but still obtained a similar survival. Vergote et al [14] reported that NACT resulted in survival rates in selected patients with advanced ovarian cancer that were comparable to those associated with primary cytoreductive surgery.
Conversely, Kuhn et al [26], Rose et al (GOG 152) [15], and Muggia et al (GOG 158) [27] reported prolonged survival times and significantly better median survival in NACT group than the conventionally treated patients. This controversy might be attributed to different patient characteristics and different treatment modalities.
In our study, there was no significant difference in the median disease free survival between both groups. Our results are similar to those of Kayikcioglu et al [28] who reported a disease free survival of 16.03 (0–84 months, median= 12 months) in 158 patients with advanced ovarian carcinoma treated by conventional surgery. In 145 patients who received NACT, he reported a disease free survival of 13.9 ± 10.12 months (0–48 median 13.9 months).
Conclusion
Primary cytoreductive surgery is still the gold standard in the treatment of ovarian carcinoma [29]. Neoadjuvant chemotherapy for advanced unresectable ovarian carcinoma lead to the selection of a group of patients sensitive to chemotherapy, in whom secondary cytoreductive surgery can be achieved in a less aggressive manner. Also neoadjuvant chemotherapy prevents mutilating surgery in patient with a very poor prognosis either because of progressive disease or because of primary chemoresisetence. These findings must be confirmed by a larger prospective study. A large randomized trial evaluating the efficacy and morbidity of primary surgery versus neoadjuvant chemotherapy followed by interval debulking surgery is ongoing.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MAFH: obtained full data about the patients, applied the study design, performed surgical interference to the patients, searched literature and drafted the manuscript,
RAFH: shared in the study design, performed statistical work and helped to draft the manuscript and edited the final version.
MAE: obtained patient consent and shared in surgical interference to the patients.
AES,: shared in surgical interference to the patients and in collecting their data.
MRE: shared in surgical interference to the patients and in collecting their data
ME shared in surgical interference to the patients and in collecting their data,
AAFH,: gave the patients chemotherapy protocol and followed them up.
Funding
This work was supported by an internal funding from the Oncology Center, Mansoura University, Mansoura, Egypt.
Acknowledgements
The authors thank Prof Dr. Mohamed El Maadawy and Prof Dr. Adel Denewar for helpful comments.
==== Refs
Randall TC Rubin SC Cytoreductive surgery for ovarian cancer Surg Clin North Am 2001 81 871 876 11551131 10.1016/S0039-6109(05)70171-7
NIH consensus development panel Ovarian cancer screening, treatment and follow up JAMA 1995 273 491 497 7837369 10.1001/jama.273.6.491
Griffith CT Parker LM Lee S Finkler NJ The effect of residual mass size on response to chemotherapy after surgical cytoreduction for advanced ovarian cancer: long term results Int J Gynacol Cancer 2002 12 323 331 10.1046/j.1525-1438.2002.01150.x
Eisenkop SM Spiretos NM Montag TW Nalick RH Wang H The impact subspecialty training on the management of advanced ovarian cancer Gynacol Oncol 1992 46 203 209 10.1016/0090-8258(92)90107-T
Frei E Clinical cancer research, an embattled species Cancer 1982 50 1979 1992 7127245
Van der Woude HJ Bloem JL Verstrate KL Osteosarcoma and Ewing sarcoma after neoadjuvant chemotherapy: value of dynamic MR imaging in detecting viable tumor before surgery Am J Roentgenol 1995 165 593 598 7645476
Bellaton E Bertozzi AI Behar C Chastagner P Brisse H Sainte-Rose C Doz F Desjardins L Neoadjuvant chemotherapy for extensive Unilateral retinoblastoma Br J Ophalmol 2003 87 327 329 10.1136/bjo.87.3.327
Ansquer Y Leblanc E Clough K Morice P Dauplat J Mathevet P Lhomme C Scherer C Tigaud JD Benchaib M Fourme E Castaigne D Querleu D Dargent D Neoadjuvant chemotherapy for unresectable ovarian carcinoma Cancer 2001 91 2329 2334 11413522 10.1002/1097-0142(20010615)91:12<2329::AID-CNCR1265>3.0.CO;2-U
Deraco M Raspagliesi F Kusamura S Management of peritoneal surface component of ovarian cancer Surg Oncol Clin N Am 2003 12 561 583 14567018 10.1016/S1055-3207(03)00027-9
Gallo A Frigerio L Neoadjuvant chemotherapy and surgical considerations in ovarian cancer Curr Opin Obstet Gynacol 2003 1 25 31 10.1097/00001703-200302000-00004
Miller AB Hoogstraten B Staquet M Winkler A Reporting results of cancer treatment Cancer 1981 47 207 214 7459811
Jacob JH Gershenson DM Morris M Copeland LJ Burke TW Wharton JT Neoadjuvant chemotherapy and interval debulking for advanced epithelial ovarian cancer Gynecol Oncol 1991 42 146 150 1894174 10.1016/0090-8258(91)90335-3
Hoskins WJ Surgical staging and cytoreductive surgery of epithelial ovarian Cancer Cancer 1993 71 1534 1540 8431891
Vergote I de Wever I Tjalma W Van Gramberen M Decloedt J Van Dam P Interval debulking surgery: an alternative for primary surgical debulking? Semin Surg Oncol 2000 19 49 53 10883024 10.1002/1098-2388(200007/08)19:1<49::AID-SSU8>3.0.CO;2-Z
Rose PG Nerenstone S Brady M Clarke-Pearson D Olt G Rubin SC Moore DH Phase III randomized study of interval secondary cytoreduction in patients with advanced stage ovarian carcinoma with suboptimal residual disease. A Gynecology Oncology Group study Proc ASCO 2002 21 201a [Abstract 802] [Last accessed August 27, 2005]
Cannistra SA Cancer of the ovary N Engl J Med 1993 329 1550 1559 8155119 10.1056/NEJM199311183292108
Nelson BE Rosenfield AT Schwartz PE Preoperative abdominopelvic Computed tomography prediction of optimal cytoreduction in epithelial ovarian carcinoma J Clin Oncol 1993 11 166 172 8418230
Meyer JI Kennedy AW Friedman R Ayoub A Zepp RC Ovarian carcinoma: Value of CT in predicting success of debulking surgery Am J Roentgenol 1995 165 875 878 7676985
Bristow RE Duska LR Lambrou NC Fishman EK O'Neill MJ Trimble EL Montz FJ A model for predicting surgical outcome in patients with advanced ovarian carcinoma using computed tomography Cancer 2000 89 1532 1540 11013368 10.1002/1097-0142(20001001)89:7<1532::AID-CNCR17>3.0.CO;2-A
Baekelandt M The potential role of neoadjuvant chemotherapy in advanced ovarian caner Int J Gynecol Cancer 2003 13 163 168 14656274 10.1111/j.1525-1438.2003.13354.x
Schwartz PE Rutherford TJ Chambers JT Kohorn EI Thiel RP Neoadjuvant chemotherapy for advanced ovarian cancer: long term survival Gynecol Oncol 1999 72 93 99 9889037 10.1006/gyno.1998.5236
Chambers JT Chambers SK Voynick IM Schwartz PE Neoadjuvant chemotherapy in a stage X ovarian carcinoma Gynecol Oncol 1990 37 327 331 2351315 10.1016/0090-8258(90)90361-N
Schwartz PE Neoadjuvent chemotherapy for the management of ovarian cancer Best Pract Res Clin Obstet Gynaecol 2002 16 585 596 12413936 10.1053/beog.2002.0304
Surwit E Childers J Atlas I Nour M Hatch K Hallum A Alberts D Neoadjuvant chemotherapy for advanced ovarian cancer Int J Gynaecol Cancer 1996 6 356 361 10.1046/j.1525-1438.1996.06050356.x
Onnis A Marchetti M Padovan P Castellan L Neoadjuvant chemotherapy in advanced ovarian cancer Eur J Gyncol Oncol 1996 17 393 396
Kuhn W Rutke S Spathe K Schmalfeldt B Florack G von Hundelshausen B Pachyn D Ulm K Graeff H Neoadjuvant chemotherapy followed by tumor debulking prolongs survival for patients with poor prognosis in International Federation of Gynecology and Obstetrics Stage IIIC ovarian carcinoma Cancer 2001 92 2585 2591 11745193 10.1002/1097-0142(20011115)92:10<2585::AID-CNCR1611>3.0.CO;2-#
Muggia FM Braly PS Brady MF Sutton G Niemann TH Lentz SL Alvarez RD Kucera PR Small JM Phase III randomized study of cisplatin versus paclitaxel versus cisplatin and paclitaxel in patients with suboptimal stage III or IV ovarian cancer: a gynecologic oncology group study J Clin Oncol 2000 18 106 115 10623700
Kayikçiog Lu F Köse MF Boran N Çalikan E Tulunay G Neoadjuvant chemotherapy or primary surgery in advanced epithelial ovarian carcinoma Int J Gynecol Cancer 2001 11 466 470 11906550 10.1046/j.1525-1438.2001.01064.x
Rose PG Nerenstone S Brady MF Clarke-Pearson D Olt G Rubin SC Moore DH Small JM Gynecologic Oncology Group Secondary surgical cytoreduction for advanced ovarian carcinoma N Eng J Med 2004 351 2489 2497 10.1056/NEJMoa041125
|
16135251
|
PMC1236969
|
CC BY
|
2021-01-04 16:39:05
|
no
|
World J Surg Oncol. 2005 Aug 31; 3:57
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-57
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-591614656710.1186/1477-7819-3-59Case ReportMucinous cystadenoma of the pancreas with predominant stroma creating a solid tumor Ae Lee Won [email protected] Department of Pathology, College of Medicine Dankook University, Cheonan, Republic of Korea2005 7 9 2005 3 59 59 23 2 2005 7 9 2005 Copyright © 2005 Ae Lee; licensee BioMed Central Ltd.2005Ae Lee; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Mucinous cystic neoplasm (MCN) of the pancreas is basically cystic epithelial neoplasm, unilocular or multilocular, occurring almost exclusively in women.
Case presentation
A 51-year-old female presented with a pancreatic mass incidentally found on the abdominal computed tomography. She underwent distal pancreatectomy. The sectioned surface of the pancreas revealed a circumscribed, whitish gray ovoid firm mass with some cystic spaces. Microscopically, glandular or small cystic structures were scattered in the predominant stroma creating a solid appearance. The subepithelial stromal component was composed of cytologically bland looking spindle cells, which resembled ovarian stroma. The stromal cells were reactive to CD 34, vimentin, progesterone receptor and calretinin. The microscopy was consistent with mucinous cystadenoma of the pancreas.
Conclusion
This case of mucinous cystadenoma of the pancreas showed very interesting pathology: It was solid rather than cystic, and accompanied by abundant benign transitional epithelia, which was a very unusual and novel finding in the mucinous cystic neoplasm of the pancreas.
==== Body
Background
Mucinous cystic neoplasm (MCN) of the pancreas is a cystic neoplasm, unilocular or multilocular, occurring almost exclusively in women. The overwhelming majority of cases occur in the body-tail of the pancreas. MCNs show two distinct components: an inner mucinous epithelial layer and an outer dense cellular ovarian-type stromal layer [1-5].
The case reported here is an exceptional case of MCN of the pancreas with predominant stromal component creating a solid tumor, and abundant transitional cell differentiation.
Case presentation
A 51-year-old female presented with a pancreatic mass found incidentally on the abdominal computed tomography for routine health examination. The mass was located in the body of the pancreas and was ill-defined with faint inhomogeneous low density at both the arterial and the venous phases of computerized tomography (CT) scan (Figure 1). Endoscopic retrograde cholangiopancreaticography showed mild indentation and slight irregularity of the neck portion of the main pancreatic duct suggesting extrinsic compression of the main pancreatic duct. The patient underwent distal pancreatectomy.
Figure 1 Contrast enhanced abdominal computed tomography showing an ill defined mass revealing faint inhomogeneous low density in the body of the pancreas.
The sectioned surface of the resected pancreas revealed a non-encapsulated, partially lobulate, whitish gray ovoid firm mass with 2.5 cm in its greatest dimension. The mass was predominantly solid with some cystic spaces, which contained mucinous fluid (Figure 2A). The mass was not communicated with pancreatic ductal system. Microscopically, the tumor was circumscribed with focal entrapped normal pancreatic acini in the periphery of the tumor. Glandular or small cystic structures were scattered in the predominant stroma, which created the tumor's solid appearance (Figure 2B). The glands or cysts were lined by a single layer of tall columnar epithelial cells, which revealed basally located nuclei and abundant intracellular mucin that was positively stained on periodic acid Schiff and alcian blue. The subepithelial stroma was cellular and composed of bland looking spindle cells mimicking ovarian stroma (Figure 3A). The columnar epithelia contained occasional goblet cells and endocrine cells and revealed abundant pseudopyloric metaplasia (Figure 3D). Some cysts were lined by bland looking polygonal or ovoid stratified cells suggestive of benign transitional epithelia (Figure 3F). The stroma adjacent to transitional epithelia was hypocellular and densely hyalinized compared to that adjacent to the columnar epithelia. As a whole the stroma was composed of cytologically bland looking spindle cells, which had variable cellularity, areas of extensive stromal hyalinization, no cytologic atypia and no mitoses.
Figure 2 A. Macroscopic photograph of the tumor. Sectioned surface reveals a non-encapsulated, circumscribed, gray-white, firm, solid mass with partially lobulate margin and irregular small cysts. B. Low power view of microscopic features of the tumor. Cystic or glandular epithelial components are set in abundant dense stroma (hematoxylin and eosin, × 40)
Figure 3 Microscopic photograph of the tumor. A. Cystic structure is lined by benign columnar mucinous epithelial cells (arrow). Subepithelial stroma is composed of bland looking spindle cells (asterisk) mimicking ovarian stroma (hematoxylin and eosin, × 200). B. Stromal cells are immunoreactive against calretinin (× 200). C. Stromal cells are reactive against progesterone receptor (× 400). D. Columnar epithelia are accompanied by intestinal metaplasia with scattered goblet cells (right) as well as pseudopyloric metaplasia (left) (hematoxylin and eosin, × 200). E. Scattered endocrine cells (arrows) are highlighted by reactivity to chromogranin A (× 400). F. Cysts are lined by benign transitional-type epithelium, beneath which the stroma is hyalinized (hematoxylin and eosin, × 200). G. Transitional cells are immunoreactive to high molecular weight cytokeratin (× 400).
Immunohistochemical studies were performed on formalin-fixed, paraffin-embedded tissue sections by the avidin-biotin peroxidase complex method. The primary antibodies used were as follow as; Pancytokeratin (1:50, AE and AE3, Zymed, San Francisco, USA), high molecular weight cytokeratin (1:100, 34betaE12, DAKO, Denmark), cytokeratin 7 (1:50, OV-TL 12/30, DAKO), cytokeratin 20 (1:40, Ks20.8, DAKO), calretinin (1:50, polyclonal, DAKO), estrogen receptor (1:40, 6F11, Novocastra, Newcastle, UK), progesterone receptor (1:80, 1A6, Novocastra), chromogranin A (1:200, LK2H10 and PHE5, NeoMarker, Fremint, CA, USA), CD34 (1:40, QBEnd 10, DAKO), vimentin (1:50, V-9, Biogenex, san Ramon, CA), α-smooth muscle actin (1:50, α-sm-1, Novocastra, Newcastle, UK), desmin (1:100, D33, DAKO), S-100 protein (1:150, B32.1, Biomeda, Foster city, CA), CD99 (1:100, HO36-1.1, NeoMarker, Fremint, CA), CD99 (1:100, HO36-1.1, Neomarker), c-kit (1:300, 104D2, DAKO). Both mucinous and transitional epithelia were diffusely immunoreactive to pancytokeratin and cytokeratin 7, but not reactive to cytokeratin 20 except for goblet cells. Transitional epithelia were diffusely reactive to high molecular weight cytokeratin (Figure 3G), whereas mucinous epithelia were not reactive to it. Scattered endocrine cells were highlighted by positive reactivity to chromogranin A (Figure 3E). The stromal cells were diffusely positive for CD 34 and vimentin, and focally positive for calretinin (Figure 3B) and progesterone receptor (Figure 3C), while negative for estrogen receptor, alpha-smooth muscle actin, desmin, c-kit, CD99, S100 protein and cytokeratin.
Discussion
MCNs of the pancreas are defined by cystic epithelial neoplasms composed of columnar mucin-producing epithelium. This tumor occurs almost exclusively in women and shows no communication with the pancreatic duct system. According to the grade of dysplasia, tumors may be classified as adenoma, borderline and non-invasive or invasive carcinoma [1-5].
The present tumor was composed of mucin producing epithelium with an ovarian type stroma. The stroma was predominant overgrowing epithelial element and creating a solid tumor. From this point of view, the main differential diagnoses of the present case include MCN with sarcomatous stroma, benign mesenchymal tumors and solid-pseudopapillary neoplasm. MCN with sarcomatous stroma is a rare variant of MCNs associated with a malignant sarcomatous stroma. This sarcomatous stroma is extremely hypercellular, contains mitotic figures, and shows marked atypia and pleomorphism of stromal cells [6]. Unlike the sarcomatous stroma, the stroma of the present case showed neither significant cellular atypia nor mitoses. Primary mesenchymal tumors of the pancreas are extraordinarily rare. Examples of benign pancreatic mesenchymal tumors composed of bland looking spindle cells include inflammatory myofibroblastic tumor [7], extragastrointestinal stromal tumor [8] and solitary fibrous tumor [5]. The possibility of primary mesenchymal tumors of the pancreas can be excluded by reason of the following points. Although the predominant stromal component overgrew the epithelial element in the present case, the epithelial element was distributed in the periphery as well as the center of the tumor, suggesting that the epithelial component is a true tumor element, not non-neoplastic tissue entrapped in the tumor. Moreover, inflammatory myofibroblastic tumor can be excluded from the viewpoint of no significant mixture of chronic inflammatory cells in the present case. Recently, the pancreatic counterpart of gastrointestinal stromal tumor was described and was based on its c-kit positivity [8]. Although in the present case the spindle cells within stroma were reactive to CD34, they were not reactive to c-kit. To the best of my knowledge, only a case of solitary fibrous tumor of the pancreas has been reported in English literature [5]. Although the CD34 positivity of stromal cells in the present case mimicked solitary fibrous tumor, the characteristic histology of solitary fibrous tumor, which correspond to patternless growth of short fascicles, a short storiform arrangement of the spindle or ovoid cells and vascularization with slit-like space was not observed in the present case. In my investigation of CD34 reactivity for normal ovarian tissue, the normal ovarian stroma was also documented to be reactive to CD34 (unpublished data). Solid- pseudopapillary neoplasm is somewhat similar to the present case from the viewpoint of mixed solid and cystic features but different by reason that solid-pseudopapillary neoplasm is composed of monomorphic polyhedral cells forming solid and pseudopapillary structures [9].
The epithelial component of MCN is composed of columnar cells which can also reveal pseudopyloric, gastric foveolar, small and large intestinal, and squamous differentiation, as is also observed in ovarian MCN [1-4]. In the present case, abundant pseudopyloric and intestinal metaplasia as well as transitional differentiation was observed. Transitional epithelia were distinguished by high molecular weight cytokeratin positivity. The mixed mucinous cystadenoma and benign Brenner tumor were described in the ovary [10-12]. However, the accompaniment of transitional epithelium was not reported in MCN of the pancreas. The stroma adjacent to the transitional epithelium was denser and more hyalinized than that adjacent to the columnar epithelium. These histologic findings suggest the possibility of a pancreatic counterpart of mixed mucinous cystadenoma and benign Brenner tumor of the ovary.
The stromal component of MCN is composed of ovarian-type stroma which express vimentin and in a high proportion, progesterone receptor, estrogen receptor, calretinin, and alpha inhibin [11,13-15]. Calretinin has been shown to recognize testicular Leydig cells and hilar ovarian cells. In the present case, stromal cells were reactive for progesterone and calretinin. The possible derivation of the stromal component of MCNs from the ovarian primordium is supported by morphology, tendency to undergo luteinization, presence of hilar-like cells, and immunophenotypic sex cord-stromal differentiation [11,13-15]. It has been hypothesized that ectopic ovarian stroma incorporated during embryogenesis in the pancreas may release hormones and growth factors causing nearby epithelium to proliferate and form cystic tumors [11,13-15].
Conclusion
This case is a pancreatic tumor showing very interesting and unusual pathology. According to the current WHO classification of pancreas, this case belongs to the mucinous cystadenoma. But this case was solid rather than cystic, and showed abundant benign transitional epithelia, which are very unusual and a novel finding that has not been described in MCNs of the pancreas. This histopathology is very similar to the mixed mucinous cystadenoma and benign Brenner tumor of the ovary [10,12].
Competing interests
The author(s) declares that she has no competing interests.
Authors' contributions
WAL performed the pathologic examination, researched the relevant literature and prepared the manuscript.
Funding source
This research was conducted by the research funds of Dankook University in 2004.
Acknowledgements
Permission of the patient was obtained for publication of her case records.
==== Refs
Suzuki Y Atomi Y Sugiyama M Isaji S Inui K Kimura W Sunamura M Furukawa T Yanagisawa A Ariyama J Takada T Watanabe H Suda K Cystic neoplasm of the pancreas: a Japanese multiinstitutional study of intraductal papillary mucinous tumor and mucinous cystic tumor Pancreas 2004 28 241 246 15084964 10.1097/00006676-200404000-00005
Thompson LD Becker RC Przygodzki RM Adair CF Heffess CS Mucinous cystic neoplasm (mucinous cystadenocarcinoma of low-grade malignant potential) of the pancreas: a clinicopathologic study of 130 cases Am J Surg Pathol 1999 23 1 16 9888699 10.1097/00000478-199901000-00001
Wilentz RE Albores-Saavedra J Hruban RH Mucinous cystic neoplasms of the pancreas Semin Diagn Pathol 2000 17 31 42 10721805
Zamboni G Scarpa A Bogina G Iacono C Bassi C Talamini G Sessa F Capella C Solcia E Rickaert F Mariuzzi GM Kloppel G Mucinous cystic tumors of the pancreas: clinicopathological features, prognosis, and relationship to other mucinous cystic tumors Am J Surg Pathol 1999 23 410 422 10199470 10.1097/00000478-199904000-00005
Kloppel G Luttges J WHO-classification 2000: exocrine pancreatic tumors Verh Dtsch Ges Pathol 2001 85 219 228 11894402
van den Berg W Tascilar M Offerhaus GJ Albores-Saavedra J Wenig BM Hruban RH Gabrielson E Pancreatic mucinous cystic neoplasms with sarcomatous stroma: molecular evidence for monoclonal origin with subsequent divergence of the epithelial and sarcomatous components Mod Pathol 2000 13 86 91 10658914 10.1038/modpathol.3880013
Yamamoto H Watanabe K Nagata M Tasaki K Honda I Watanabe S Soda H Takenouti T Inflammatory myofibroblastic tumor (IMT) of the pancreas J Hepatobiliary Pancreat Surg 2002 9 116 119 12021906 10.1007/s005340200013
Yamaura K Kato K Miyazawa M Haba Y Muramatsu A Miyata K Koide N Stromal tumor of the pancreas with expression of c-kit protein: report of a case J Gastroenterol Hepatol 2004 19 467 470 15012791 10.1111/j.1440-1746.2003.02891.x
Klimstra DS Wenig BM Heffess CS Solid-pseudopapillary tumor of the pancreas: a typically cystic carcinoma of low malignant potential Semin Diagn Pathol 2000 17 66 80 10721808
Murphy GF Welch WR Urcuyo R Brenner tumor and mucinous cystadenoma of borderline malignancy in a patient with Turner's syndrome Obstet Gynecol 1979 54 660 663 503400
Roth LM Gersell DJ Ulbright TM Ovarian Brenner tumors and transitional cell carcinoma: recent developments Int J Gynecol Pathol 1993 12 128 133 8463036
Waxman M Pure and mixed Brenner tumors of the ovary: clinicopathologic and histogenetic observations Cancer 1979 43 1830 1839 445369
Fukushima N Mukai K 'Ovarian-type' stroma of pancreatic mucinous cystic tumor expresses smooth muscle phenotype Pathol Int 1997 47 806 808 9413043
Lapertosa G [Histogenetic considerations on mucinous cystomas of the ovary based on histochemical and immunohistochemical findings] Pathologica 1989 81 381 401 2696922
Yeh MM Tang LH Wang S Robert ME Zheng W Jain D Inhibin expression in ovarian-type stroma in mucinous cystic neoplasms of the pancreas Appl Immunohistochem Mol Morphol 2004 12 148 152 15354741
|
16146567
|
PMC1236970
|
CC BY
|
2021-01-04 16:39:05
|
no
|
World J Surg Oncol. 2005 Sep 7; 3:59
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-59
|
oa_comm
|
==== Front
Emerg Infect Dis
Emerging Infect. Dis
EID
Emerging Infectious Diseases
1080-6040
1080-6059
Centers for Disease Control and Prevention
15663854
04-0486
10.3201/eid1012.040486
Research
Research
Differential Virulence of West Nile Strains for American Crows
Differential West Nile Virulence for American Crows
Brault Aaron C. *†
Langevin Stanley A. *†
Bowen Richard A. ‡
Panella Nicholas A. *
Biggerstaff Brad J. *
Miller Barry R. *
Komar Nicholas *
* Centers for Disease Control and Prevention, Fort Collins, Colorado, USA:
† University of California, Davis, California, USA;
‡ Colorado State University, Fort Collins, Colorado, USA
Address for correspondence: Aaron C. Brault, Center for Vectorborne Diseases, Department of Pathology, Microbiology and Immunology, School of Veterinary Medicine, University of California, Davis, CA 95616, USA; fax: 530-752-3349; email: [email protected]
12 2004
10 12 21612168
Increased viremia and deaths in American Crows inoculated with a North American West Nile viral genotype indicate that viral genetic determinants enhance avian pathogenicity and increase transmission potential of WNV.
Crow deaths were observed after West Nile virus (WNV) was introduced into North America, and this phenomenon has subsequently been used to monitor the spread of the virus. To investigate potential differences in the crow virulence of different WNV strains, American Crows were inoculated with Old World strains of WNV from Kenya and Australia (Kunjin) and a North American (NY99) WNV genotype. Infection of crows with NY99 genotype resulted in high serum viremia levels and death; the Kenyan and Kunjin genotypes elicited low viremia levels and minimal deaths but resulted in the generation of neutralizing antibodies capable of providing 100% protection from infection with the NY99 strain. These results suggest that genetic alterations in NY99 WNV are responsible for the crow-virulent phenotype and that increased replication of this strain in crows could spread WNV in North America.
Keywords:
West Nile virus
American Crows
strains
mortality
virulence
Kunjin
research
==== Body
West Nile virus (WNV, Flaviviridae: Flavivirus) is maintained in nature by transmission between mosquitoes and birds and has an extensive geographic range, including Europe, Africa, the Middle East, southern Asia, and Australia (1). In 1999, WNV was identified in North America (2) and has become the leading cause of arboviral encephalitis in humans and horses (3), as well as having been implicated in deaths of members of at least 198 bird species (4). Corvids, including the American Crow (Corvus brachyrhynchos), appear to be most susceptible (5,6), and corvid deaths have subsequently been used as a sentinel to track the spread of the virus (7).
Experimental injection of American Crows with the North American genotype of WNV (NY99 strain) has confirmed its highly pathogenic phenotype. Mean peak viremia titers in American Crows exceed 9 log10 PFU/mL in sera, with 100% deaths within 6 days postinfection (dpi) (5). With the exception of bird deaths in Israel (8), where a strain 99.8% similar to the NY99 genotype has circulated since 1997 (9), no bird deaths have been reported during numerous well-characterized WNV epidemics in North Africa (10), Europe (11–13), Russia (14), and South Africa (15). A closely related virus that circulates in Australia (Kunjin [KUN]) has never been associated with outbreaks of human or animal diseases, including bird diseases, nor have bird deaths been reported from enzootic transmission foci in Africa, where a virus that shares 96.5% nucleotide identity with the NY99 strain has previously been isolated (16,17). Possible explanations for the lack of reporting of bird deaths before 1998 include the following: failure to identify bird deaths in other regions, a higher susceptibility to WNV-induced disease among North American birds, or the fact that the North American WNV strain possesses increased avian virulence determinants. Additionally, the possible immunologic cross-protection of birds with lesser virulent strains could be a factor that has limited the identification of bird deaths outside the Middle East. Immunologically naïve bird populations in North America could be at an increased risk of acquiring severe disease.
The close genetic relatedness of the North American WNV genotype with the bird-pathogenic Israeli WNV strain suggests differential avian pathogenicity among WNV strains (9). To evaluate whether WNV-associated deaths in American Crows was due to infection by a more virulent genotype, we injected American Crows with NY99, a closely related WNV strain from Kenya (KEN) and a more distantly related WNV strain from Australia (KUN) and monitored viremia titers and illness. In addition, birds that survived challenge with the KEN or KUN viruses were challenged with a lethal dose of the NY99 strain to assess development of a cross-protective immunologic response.
Materials and Methods
Viral Strains and Birds Used
The lowest passage WNV available were used for crow virulence studies to avoid incorporating confounding cell-culture–related genetic substitutions. The NY99 isolate used was originally isolated from an American Crow brain (strain NY99-4132) and was subsequently passaged once in Vero cells before being used for these studies. The Kenya-3829 (KEN) isolate was made from a pool of male Culex univittatus mosquitoes (16) and passaged twice in Vero cells. The Kunjin (KUN-6453) isolate was made from Cx. annulirostris mosquitoes and was passaged once in Vero cells and once in hamster kidney cells (Table 1). After-hatch-year American Crows were obtained by using net traps with the assistance of the Kansas Department of Wildlife Resources. The crows were banded and transported to Fort Collins, Colorado, where they were housed in the Colorado State University Animal Disease Laboratory in groups of two in 1-m3 cages. Crows were fed a combination of ground corn and dried cat food and dog food.
Table 1 West Nile viral strains used for virulence studies in American Crows
Virus Strain Source Passage historya Location Genetic lineageb
NY99 NY99-4132 American Crow (brain) V1 USA I
KEN KEN-3829 Culex univittatus V2 Kenya I
Kunjin (KUN) KUN-6453 Cx. annulirostris V1, BHK1 Australia I
aViruses were propagated in Vero (V) or baby hamster kidney (BHK) cells. Numbers following passage source represent the number of viral passages.
bGenetic lineages as reported previously (9).
Detection of Preexisting Flaviviral Antibodies
To confirm that crows had not previously been exposed to WNV or another endemic flavivirus, St. Louis encephalitis virus (SLEV), crows were bled before injection and serum-tested by plaque reduction neutralization assays (PRNTs) with WNV and SLEV viruses. Serum was diluted 1:5, heat inactivated at 56°C for 30 min, and incubated with an equal volume of virus (SLEV; strain TBH-28) and WNV (strain NY99-4132) to a final concentration of 100 PFU/0.1 mL. Samples were incubated at 37°C for 1 h, and 0.1 mL of each was added to a confluent monolayer of Vero cells in 6-well plates (Costar Inc., Cambridge, MA). After incubation for 1 h, cell monolayers were overlaid with 0.5% agarose; a second overlay containing 0.005% neutral red was added 48 h later. Plates were read 1–2 days after addition of the second overlay. A 90% reduction in PFU, as compared to the serum-negative control, was used as the determinant of neutralization. Detection of any neutralizing activity to either SLEV or WNV within the serum of any crow precluded use for experimental inoculation.
Virus Injection
Viral stocks were diluted to 3.2 log10 PFU/0.1 mL in minimal essential media (MEM) containing no fetal bovine sera (FBS). One hundred microliters of the diluted stocks was subcutaneously injected on the breast region of eight American Crows in four infection groups. Crows were injected with 1) NY99, 2) KEN, 3) KUN WNVs, or 4) with a media-only injection that served as a virus-negative control. In addition, a higher dose inoculum of 3.8 log10 PFU/0.1 mL was prepared for injection of a fifth group of crows with KEN WNV. All crows were examined for signs of disease twice daily for 14 days after injection and bled once daily from 1 to 7 dpi for characterization of viremia. Blood samples were collected from the jugular or brachial vein by using a 26-gauge needle; 0.2 mL of blood was added to 0.9 mL of MEM supplemented with 20% FBS to obtain approximately a 10–1 serum dilution. Coagulation was allowed to take place at room temperature for 30 min, at which point samples were placed on ice and spun at 3,700 x g for 10 min to pellet clotted cells. The supernatants from these samples were frozen at –80°C until samples were titrated for infectious units.
Assaying for Infectious Virus
Infectious virus was assayed by plaque formation on monolayers of Vero cells. Briefly, serial 10-fold dilutions of serum were added to Vero cells that were overlaid as described previously for PRNTs. PFU were enumerated at 3 dpi and multiplied by the dilution factor to determine viral titers per mL serum. Initial 1:10 dilution of serum as well as the use of 200 µL of the lowest dilution, resulted in a limit of viral PFU detection of 1.7 log10 PFU/mL serum. Inocula for all three viruses were back-titrated by plaque assay in order to confirm the uniformity of the doses administered.
Determination of Cross-Protection
Blood (0.6 mL) was drawn at 14 dpi to determine the levels of WNV-specific antibodies and cross-neutralization by using a 2-way β PRNT with homologous and heterologous WNV strains. Briefly, twofold dilutions of bird serum samples were incubated at 56°C for 30 min and mixed with either NY99, KEN, or KUN viruses. Samples were allowed to incubate for 1 h at 37°C, at which point the samples were injected onto Vero cells and overlaid as previously described for PRNT. Plaques were counted, and neutralization was reported as a 90% reduction in plaque formation as compared to the results for the serum-negative control.
Crows that survived through 14 dpi were subsequently challenged with 3.2 log10 PFU of NY99 virus from the same seed that was used for the initial infection of the experimental NY99 infection group. Crows were bled daily through 7 dpi and were held through 11 dpi, at which point an additional 0.6 mL of blood was drawn to assess modulations in neutralizing activity after secondary challenge. Serum samples from the seven daily bleedings were diluted 1:10 in MEM diluent, spun, immediately assayed for the presence of infectious virus on Vero cells, and then stored at –80°C. Samples demonstrating virus were thawed and titrated on Vero cells as described above. Additionally, serum drawn at the end of the time course was assayed for antibody by PRNT.
Statistical Analyses
Statistical analyses were performed on peak viremia level, duration of viremia, day of viremia onset, and day of death. All analyses with the exception of day of death were performed by analyses of variance (ANOVA). Multiple comparisons, i.e., confidence intervals (CI) for the difference of means, were performed by using Tukey's highest significant difference (HSD) adjustment for comparisons of means. Because only two virus groups had birds that died, the day-of-death comparisons were analyzed by using a Student t test with Welch's modification for unequal variances. Proportions of illness and death were compared with the Fisher exact test.
Results
Flaviviral antibodies were not detected in any of the preinoculation serum samples assayed by PRNT. Therefore, all captured American Crows were used for experimental inoculation. Peak viremia titers ranging from 6.7 log10 to 10.7 log10 PFU/mL serum (mean peak viremia titers = 9.2 log10 PFU/mL serum, 95% CI 8.2 log10 PFU/mL serum–10.2 log10 PFU/mL serum) developed in all crows injected with the NY99 WNV genotype (Figure 1). Onset of viremia occurred within 24 h for three of the eight crows injected with NY99 and was present in all eight birds within 48 h postinjection (mean onset of viremia = 1.8 dpi, 95% CI 1.4 dpi–2.1 dpi) (Table 2). Mean onset of viremia and mean peak viremia titers differed significantly among the virus groups (mean onset, F = 31.6, df = 3,22, p < 0.001; mean peak viremia, F = 74.9, df = 2,21, p < 0.001). In contrast to the NY99-infected crows, detectable viremia (>1.7 log10 PFU/mL sera) developed in two crows infected with the KEN WNV. The onset of viremia in these two birds was delayed until 3 dpi and 4 dpi, and the mean peak viremia level was lower than that of the NY99 infection group (7.5 log10 PFU/mL) (difference of mean onset of viremia = 1.8 dpi, 95% CI 0.4 – 3.1). When the inoculum dose was increased to 3.8 log10 PFU for the KEN strain, viremia developed in all eight of the crows, with peak titers ranging from 4.2 to 6.1 log10 PFU/mL serum (mean = 4.9 log10 PFU/mL serum, 95% CI 4.3–5.4 log10 PFU/mL serum). The onset of viremia was delayed in the higher dose KEN group compared to the NY99 infection group (mean = 4.5 dpi, 95% CI 3.9–5.1 dpi; difference of mean onset of viremia = 2.8 dpi, 95% CI 1.9–3.6 dpi). In all eight crows inoculated with 3.2 log10 PFU of KUN virus, peak viremia titers were 2.7–4.9 log10 PFU/mL serum (mean = 4.2 log10 PFU/mL, 95% CI 3.5–4.8 log10 PFU/mL serum). Onset of viremia relative to the NY99-infected crows was slightly delayed, with a mean onset at 2.4 dpi (95% CI 1.9–2.8 dpi) (difference of mean onset of viremia = 0.6 dpi, 95% CI 0.2–1.5 dpi) (Figure 1). Viremia developed in KUN-infected crows, lasting from 1 to 5 days with a mean duration of 3 days (95% CI 1.9–4.1 days). This finding differs qualitatively from the NY99- and KEN-infected birds, which sustained viremia for at least 4 days; viremia levels ceased only when the bleeding time course was halted or at time of death; the differences between the viremia durations for the KUN-infected crows and the NY99 and KEN groups were not statistically significant when adjustments were made for multiple comparisons.
Figure 1 Viremia profiles for West Nile virus (WNV)–infected American Crows after injection of 1,500 PFU of KUN or KEN/NY99 WNV. Viral titers were determined by plaque formation on Vero cells and represented as geometric means. A detection limit of >1.7 log10 PFU/mL crow serum was determined. Bars represent standard deviations (SD) of the mean. hd, high dose.
Table 2 Clinical profile of American Crows infected with WNV strains NY99 (strain NY99-4132), KEN (strain KEN-3829), and KUN (strain KUN-6453)a
Virus group Mortality:
no. died/N (%) Morbidity: no. ill/N (%) Mean day of death ± SD Mean peak viremiaa ± SD (mean duration ± SD) (n) Mean day of peak viremiaa ± SD
NY99 8/8 (100) 8/8 (100) 5.1 ± 0.6 9.2 ± 1.2 (4.2 ± 0.7) (8) 4.3 ± 0.9
KEN 1/8 (12.5) 2/8 (25) 9 ± NA 7.5 ± 0 (4.5 ± 0.7) (2) 5.0 ± 1.4
KEN-hdb 2/8 (25) 3/8 (38) 10.5 ± 2.1 4.8 ± 0.6 (3.1 ± 0.8) (8) 5.5 ± 0.9
KUN 0/8 0/8 NA 4.2 ± 0.8 (1.8 ± 0.5) (8) 3.3 ± 0.7
Control 0/8 0/8 NA NA NA
aViral titers were expressed as the log10 PFU/mL of crow sera as determined by plaque assay on Vero cells.
bhd, high-dose group (6,000 [3.8 log10] PFU); NA, not applicable.
All crows in the NY99 group died by dpi 6 (Figure 2). Signs of illness (unresponsiveness, anorexia, weight loss), signs of encephalitis (shaking, convulsion, ataxia), or both developed in all NY99-infected crows. In addition, hemorrhage from oral and cloacal cavities was evident in five (62.5%) of the eight crows in the NY99 group. One crow died of infection with NY99 at 4 dpi, five at 5 dpi, and the remaining two at 6 dpi (Figure 1). Only one crow (12.5%) died of infection with the KEN virus with the 3.2 log10 PFU injection. When the dose was increased to 3.8 log10 PFU, 2 (25%) crows did not survive the infection. Regardless of the dose administered, the crows infected with the KEN virus demonstrated a reduced mortality rate (p < 0.001), compared to that of the NY99 virus. Virus was isolated from the brains of the small subset of crows that died of infection with the KEN strain (data not shown). In addition to the three deaths from the KEN WNV genotype, an additional two crows showed signs of illness, yet survived through 14 dpi (Table 2). No illness or death was identified within the KUN infection group, yielding a significant difference from the NY99 infection group (p < 0.001), but the KUN group was not statistically differentiated from the KEN WNV infection groups (p = 0.53).
Figure 2 Survivorship of eight American Crows, each injected with 3.2 log10 PFU of NY99, KEN, or KUN virus. An additional eight crows were injected with a high dose (hd) of the KEN virus (3.8 log10 PFU). Crows were monitored daily for signs of disease through 14 dpi. No deaths were found within the control group (data not shown).
None of the eight crows previously challenged with KUN virus had detectable illness after secondary challenge with 3.2 log10 PFU of NY99 virus (Figure 3), which clearly indicates a cross-protective immune response against NY99; the lower 95% confidence limit on cross-protection probability was 0.63. In fact, viremia was not detected in any of the eight crows rechallenged with the NY99 WNV on any of the 7 dpi (Figure 4). PRNTs demonstrated a homologous neutralization response in all eight of the crows for KUN virus (Table 3). Heterologous titers against NY99 virus were equal to or only twofold lower than those against KUN virus.
Figure 3 Survivorship of American Crows previously immunized with West Nile virus (WNV)-KUN or WNV-KEN viruses after injection with 1,500 PFU of NY99 WNV. hd, high dose.
Figure 4 Viremia production of American Crows previously immunized with West Nile virus (WNV)-KUN or WNV-KEN viruses after injection with 3.2 log10 PFU of NY99 WNV. No detectable levels of viremia (>1.7 log10 PFU/mL crow serum) developed in the KUN virus–immunized crows (0/8). Data points for the naïve (unexposed to WNV) crows challenged with the NY99 virus represent the mean of three samples chosen randomly. Bars represent standard deviations (SD) of the mean. hd, high dose; PRNT, plaque reduction neutralization assay.
Table 3 Cross-neutralization immune response of American Crows at 14 days postinfection with either KEN or KUN viruses
Sample no. Inoculation NY99 KEN KUN Difference
Crow 8 KEN 640a 640b NT 0
Crow 1 KUN 80 NTc 160 2-fold
Crow 2 KUN 80 NT 80 0
Crow 3 KUN 160 NT 320 2-fold
Crow 4 KUN 40 NT 80 2-fold
Crow 5 KUN 40 NT 40 0
Crow 6 KUN 20 NT 20 0
Crow 7 KUN 80 NT 80 0
Crow 8 KUN 80 NT 160 2-fold
aValues represent the greatest reciprocal dilution in which >90% plaque inhibition was achieved as compared to sera-negative control cultures.
bHomologous titers are depicted in bold print.
cNT not tested; KEN, West Nile virus strain from Kenya; KUN (Kunjin), West Nile virus strain from Australia.
Only one of the seven crows from the lower dose (3.2 log10 PFU) KEN WNV inoculation group survived rechallenge with the NY99 strain (Figure 3). Sera drawn before the NY99 rechallenge from all crows within this group demonstrated that an immune response had developed in one crow (the single survivor). This crow demonstrated illness after the original KEN WNV challenge and was one of the two crows that had detectable viremia levels and subsequently exhibited a homologous protective antibody titer that was indistinguishable from its heterologous titer against the NY99 virus (1:640) (Table 4). The six KEN-infected survivors that did not become viremic from the original KEN viral challenge were devoid of detectable neutralizing antibody titers and had unmodified infections after the NY99 challenge. The viremia profile and clinical outcome (Figure 3 and Figure 4) were indistinguishable from infection of naïve birds: five crows died on 5 dpi and an additional crow died on 6 dpi. The single surviving crow that had demonstrated a 1:640 heterologous PRNT titer against NY99 WNV did not manifest a NY99 viremia level and had an unmodified 1:640 PRNT titer after the NY99 challenge. All crows from the group that received the higher dose of KEN generated KEN viremia titers and exhibited homologous PRNT titers (1:1,280–2,560) that were indistinguishable (less than fourfold difference) from heterologous titers against the NY99 virus. Neither clinical disease nor NY99 viremia levels were identified in these crows after secondary challenge with the NY99 virus, but neutralizing antibody titers increased up to 16-fold. The rise in PRNT titer was most likely the result of secondary infection or exposure; however, no control American Crows (to which a secondary challenge was not administered) were assayed for elevated PRNT titers at 24 dpi.
Table 4 Cross-neutralization immune response of American Crows 24 days postinfection (dpi) with either KEN or KUN virusesa
Sample no. Inoculation NY99 KEN KUN Difference
Crow 8 KEN 640b 640c NT 0
Crow 1 KUN 160 NTd 320 2-fold
Crow 2 KUN 320 NT 320 0
Crow 3 KUN 160 NT 160 0
Crow-4 KUN 160 NT 320 2-fold
Crow 5 KUN 160 NT 160 0
Crow 6 KUN 320 NT 320 0
Crow-7 KUN 640 NT 640 0
Crow 8 KUN 640 NT 640 0
aFollowing secondary NY99 challenge at 14 dpi.
bValues represent the greatest reciprocal dilution in which >90% plaque inhibition was achieved as compared to sera-negative control cultures.
cHomologous titers are depicted in bold print.
dNT not tested; KEN, West Nile virus strain from Kenya; KUN (Kunjin), West Nile virus strain from Australia.
Sequence analyses of the coding differences between the NY99 and KEN viruses (Table 5) were performed on a NY99 virus (that had undergone an additional 2 Vero cell passages) to assess the possibility that limited cell-culture propagation could have resulted in attenuating genetic substitutions found between the KEN and NY99 genotype. These analyses did not demonstrate any genetic modification at any of the KEN or NY99 variable sites, further indicating that the genotype is stable for up to at least 3 passages and that the attenuated phenotype of the KEN or KUN viruses was unlikely to be the result of an additional tissue culture passage.
Table 5 Amino acid differences between the NY99 and KEN West Nile virus strainsa,b
Viral gene Amino acid position NY99 KEN
Capsida 3 Leu Asn
Capsid 8 Val Ala
Envelope 126 Ile Thr
Envelope 159 Val Ile
NS1 70 Ala Ser
NS2a 52 Thr Ala
NS2b 103 Val Ala
NS3 249 Pro Thr
NS3 356 Thr Ile
NS4a 85 Ala Val
NS4b 249 Glu Asp
aSource (17).
bKEN, West Nile virus strain from Kenya; Leu, leucine; Val, valine; Ile, isoleucine; Ala, alanine; Thr, threonine; Pro, proline; Glu, glutamine; Asn, asparagine; Ser, serine; Asp, aspartic acid.
cVariable structural amino acid residues have been designated by bold text.
Discussion
Viremia levels observed in these studies confirm previous observations that American Crows have the potential to serve as amplification hosts for the NY99 genotype of WNV but suggest that corvids may not be important hosts for alternative WNV genotypes because of substantially reduced viremia titers that would not favor efficient virus transmission. Furthermore, these results demonstrate that viral-encoded determinants of avian pathology that are absent from KEN and KUN viruses exist in the NY99 virus. The viremia levels observed in crows inoculated with the KEN or KUN viruses were significantly lower than and delayed in their onset compared to those seen after inoculation with the NY99 strain. These data demonstrate that the differential pathogenic phenotypes of the WNV strains are the result of viral genetic differences that encode particular virulence determinants. Despite the finding that mouse virulence of the NY99 and KUN WNV strains (18) correlates well with the virulence phenotype identified in crow experiments here, experimental infection of mice with the KEN WNV strain did not demonstrate an attenuated phenotype (D.W.C. Beasley and A.D.T. Barrett, pers. comm.). This observation indicates that differential pathogenic mechanisms could modulate virulence in disparate vertebrate hosts.
Elevated viremia level could be a predominant factor for severe clinical outcome. KUN and KEN WNV-infected crows in which clinical signs did not develop did not manifest peak viremia titers >6 log10 PFU/mL; however, peripheral titers exceeded this level for the three crows in which neurologic symptoms and death occurred. Additionally, viremia levels of all crows injected with NY99 surpassed this level, which suggests that once a peripheral circulatory threshold titer is achieved, virus is capable of accessing the nervous system through a nonspecific mechanism. Intracerebral injection of mice with WNV strains differing in neuroinvasive capacity has demonstrated uniform lethality, indicating that the ability to enter the nervous system and not neurovirulence, is instrumental for virulence of WNV strains (18,19). If this phenomenon is true for WNV strains in crows, then the mechanism by which the crow-virulent genotype achieves extremely high peripheral titers must be elucidated. Viruses capable of replicating to higher titers could result from a unique access to cell types that facilitate high-titer replication through more efficient receptor-envelope interactions, viral replicase differences that increase replication efficiency within host cells, decreased sensitivity to host innate immunologic responses, or by altering the physiological host responses such as fever.
Immunologic status of a host can play an important role in limiting disease expression. WNV that are capable of inducing substantial levels of viremia and neuroinvasion of immunodeficient mice do not necessarily cause viremia or enter the neural tissues of mice with competent immune systems (19). Studies have demonstrated that previous infection with heterologous flaviviruses reduces the incidence of encephalitis and can provide protection from fatal WNV challenge in a hamster model for WNV pathogenesis (20,21). In contrast, a neutralization study performed with WNV strains of different lineages demonstrated that neutralizing antibodies against an Indian WNV strain provided poor protection against a South African WNV strain (22). Our results demonstrated that prior immunization with KUN virus can provide protection from lethal NY99 challenge in crows. Crows in which a detectable level of viremia did not develop from the initial KEN viral challenge exhibited viremia levels and death rates indistinguishable from NY99-infected naïve crows. Crows injected with the higher doses, which led to productive infections with the KEN virus, produced neutralizing-antibody titers that were protective against lethal NY99 challenge. The cross-neutralization of WNV strains suggests that areas in which WNV virus is endemic could be much less susceptible to invasion by the crow-virulent NY99 genotype.
The effect that endemic flaviviruses such as SLEV has on the genetic stability of WNV in North America remains unclear; however, the fact that WNV and SLEV are distinguishable serologically through PRNT (23) and that WNV activity within the United States has occurred sympatrically within SLEV transmission foci (3) suggest that SLEV seroprevalence in birds has little impact on WNV transmission. Previous studies have demonstrated in a flaviviral pathogenesis hamster model that previous exposure to SLEV can significantly reduce WNV viral titers (21). Future experiments are warranted to determine if such protection is afforded in avian species.
Experimental inoculation with an Egyptian WNV strain has demonstrated deaths in sparrows and crows (24), providing evidence that bird deaths could result from natural infection with alternative WNV genotypes. Despite this fact, no bird deaths were reported during a well-described Egyptian epidemic involving the same viral strain used to experimentally inoculate these birds (10). Our results demonstrated that low numbers of deaths can occur from infection with alternative WNV strains, but the NY99 WNV genotype is significantly more virulent for American Crows. This result, coupled with the finding that similar pathogenicity was identified between the NY99 and KEN WNV in house sparrows (25), indicates the dual role of viral pathogenic phenotype and host susceptibility for the expression of virulence in a particular bird species. Differential susceptibility of mouse strains for WNV infection has been identified and correlated with immunologic gene expression (26). Future experimental inoculation of Old World corvids with differential WNV genotypes would be useful to assess the role that host susceptibility has on the emergence of WNV genotypes in different geographic regions.
The mutations that encode the determinants for differential crow virulence are currently unknown. In crows inoculated with a recombinant virus containing WNV structural genes and nonstructural (NS) genes of yellow fever virus (YFV), viremia did not develop (27). The fact that the parental YFV-17D vaccine strain did not replicate to detectable levels in chickens (28) indicates that flaviviral NS gene regions could modulate viral replication in birds. Analysis of the complete genomes of the NY99 and KEN WNV has identified a maximum of 11 amino acids (Table 5) and 22 nucleotides from the 3´NCR that could mediate this phenotype (17). Seven (64%) of the 11 amino acid differences between these viruses resided with the NS gene region. The close genetic identity between the KEN and NY99 WNV genotypes makes this an optimal system for the systematic identification of genetic elements that encode viral pathogenic determinants. Studies are under way to identify the specific viral genetic determinants of crow virulence through the use of infectious cDNAs generated from both the NY99 and KEN WNV genotypes.
Acknowledgments
We thank Robert B. Tesh and David Beasley for providing the low-passage Kunjin isolate used in this study, Max Tehee and Paul Gordy for technical assistance, Charles Cope and Tom Janousek for invaluable assistance with crow trapping, and Ann M. Powers for critical review of the manuscript.
A.C.B. was supported by the American Society for Microbiology as a National Center for Infectious Diseases postdoctoral fellow. Trapping of American Crows was performed under U.S. Fish and Wildlife Scientific Collecting Permit number MB-032526. Experimental inoculations of crows were performed under Centers for Disease Control and Prevention IACUC protocol numbers 02-26-012-MSA and 03-13-013-MSA.
Dr. Brault is an assistant molecular arbovirologist in the Center for Vectorborne Diseases and assistant professor of pathology, microbiology and immunology in the School of Veterinary Medicine, University of California, Davis. His main research interests include the identification of viral molecular determinants of pathogenesis and vector infectivity.
Suggested citation for this article: Brault AC, Langevin SA, Bowen RA, Panella NA, Biggerstaff BJ, Miller BR, et al. Differential virulence of West Nile strains for American Crows. Emerg Infect Dis [serial on the Internet]. 2004 Dec [date cited]. http://dx.doi.org/10.3201/eid1012.040486
==== Refs
References
1. Hayes CG West Nile fever. In: Monath TP, editor. The arboviruses: epidemiology and ecology, vol. V. Boca Raton (FL): CRC Press; 1989 p. 59–88.
2. Komar N West Nile viral encephalitis. Sci Tech Rev. 2000;19 :166–76 11189714
3. O'Leary DR, Marfin AA, Montgomery SP, Kipp AM, Lehman JA, Biggerstaff BJ, The epidemic of West Nile virus in the United States, 2002. Vector Borne Zoonotic Dis. 2004;4 :61–70 10.1089/153036604773083004 15018774
4. Komar N West Nile virus: epidemiology and ecology in North America. Adv Virus Res. 2003;61 :185–234 10.1016/S0065-3527(03)61005-5 14714433
5. Komar N, Langevin S, Hinten S, Nemeth N, Edwards E, Hettler D, Experimental infection of North American birds with the New York 1999 strain of West Nile virus. Emerg Infect Dis. 2003;9 :311–22 12643825
6. McLean RG, Ubico SR, Docherty DE, Hansen WR, Sileo L, McNamara TS West Nile virus transmission and ecology in birds. Ann N Y Acad Sci. 2001;951 :54–7 10.1111/j.1749-6632.2001.tb02684.x 11797804
7. Eidson M, Komar N, Sorhage F, Nelson R, Talbot T, Mostashari F, Crow deaths as a sentinel surveillance system for West Nile virus in the northeastern United States, 1999. Emerg Infect Dis. 2001;7 :615–20 10.3201/eid0704.010402 11585521
8. Bin H, Grossman Z, Pokamunski S, Malkinson M, Weiss L, Duvdevani P, West Nile fever in Israel in 1999–2000: from geese to humans. Ann N Y Acad Sci. 2001;951 :127–42 10.1111/j.1749-6632.2001.tb02691.x 11797770
9. Lanciotti RS, Roehrig JT, Deubel V, Smith J, Parker M, Steele K, Origin of the West Nile virus responsible for an outbreak of encephalitis in the northeastern United States. Science. 1999;286 :2333–7 10.1126/science.286.5448.2333 10600742
10. Taylor R, Work T, Rizk F A study of the ecology of West Nile virus in Egypt. Am J Trop Med Hyg. 1956;5 :579–620 13354882
11. Murgue B, Murri S, Zientara S, Durand B, Durand JP, Zeller H West Nile outbreak in horses in southern France, 2000: the return after 35 years. Emerg Infect Dis. 2001;7 :692–6 10.3201/eid0704.010417 11585534
12. Savage HM, Ceianu C, Nicolescu G, Karabatsos N, Lanciotti R, Vladimirescu AL, Entomologic and avian investigations of an epidemic of West Nile fever in Romania in 1996, with serologic and molecular characterization of a virus isolate from mosquitoes. Am J Trop Med Hyg. 1999;61 :600–11 10548295
13. Tsai TF, Popovici F, Cernescu C, Campbell GL, Nedelcu NI West Nile encephalitis epidemic in southeastern Romania. Lancet. 1998;352 :767–71 10.1016/S0140-6736(98)03538-7 9737281
14. Platonov AE, Shipulin GA, Shipulina OY, Tyutyunnik EN, Frolochkina TI, Lanciotti RS, Outbreak of West Nile virus infection, Volgograd region, Russia, 1999. Emerg Infect Dis. 2001;7 :128–32 10.3201/eid0701.010118 11266303
15. McIntosh BM, McGillivray GM, Dickinson DB, Taljaard JJ Ecological studies on Sindbis and West Nile viruses in South Africa. IV. Infection in a wild avian population. S Afr J Med Sci. 1968;33 :105–12 5731827
16. Miller BR, Nasci RS, Godsey MS, Savage HM, Lutwama JJ, Lanciotti RS, First field evidence for natural vertical transmission of West Nile virus in Culex univittatus complex mosquitoes from Rift Valley province, Kenya. Am J Trop Med Hyg. 2000;62 :240–6 10813479
17. Charrel RN, Brault AC, Gallian P, Lemasson JJ, Murgue B, Murri S, Evolutionary relationship between Old World West Nile virus strains. Evidence for viral gene flow between Africa, the Middle East, and Europe. Virology. 2003;315 :381–8 10.1016/S0042-6822(03)00536-1 14585341
18. Beasley DW, Li L, Suderman MT, Barrett AD Mouse neuroinvasive phenotype of West Nile virus strains varies depending upon virus genotype. Virology. 2002;296 :17–23 10.1006/viro.2002.1372 12036314
19. Halevy M, Akov Y, Ben-Nathan D, Kobiler D, Lachmi B, Lustig S Loss of active neuroinvasiveness in attenuated strains of West Nile virus: pathogenicity in immunocompetent and SCID mice. Arch Virol. 1994;137 :355–70 10.1007/BF01309481 7944955
20. Price WH, Thind IS Protection against West Nile virus induced by a previous injection with dengue virus. Am J Epidemiol. 1971;94 :596–607 5133787
21. Tesh RB, Travassos Da Rosa AP, Guzman H, Araujo TP, Xiao SY Immunization with heterologous flaviviruses protective against fatal West Nile encephalitis. Emerg Infect Dis. 2002;8 :245–51 10.3201/eid0803.010238 11927020
22. Blackburn NK, Thompson DL, Jupp PG Antigenic relationship of West Nile strains by titre ratios calculated from cross-neutralization test results. Epidemiol Infect. 1987;99 :551–7 10.1017/S0950268800068059 3678408
23. Calisher CH, Karabatsos N, Dalrymple JM, Shope RE, Porterfield JS, Westaway EG, Antigenic relationships between flaviviruses as determined by cross-neutralization tests with polyclonal antisera. J Gen Virol. 1989;70 :37–43 10.1099/0022-1317-70-1-37 2543738
24. Work T, Hurlbut H, Taylor R Indigenous wild birds of the Nile Delta as potential West Nile circulating reservoirs virus. Am J Trop Med Hyg. 1955;4 :872–88 13259011
25. Langevin SA, Brault A, Panella NA, Bowen R, Komar N West Nile virus strains vary in virulence for house sparrows (Passer domesticus). Am J Trop Med Hyg. In press
26. Darnell MB, Koprowski H, Lagerspetz K Genetically determined resistance to infection with group B arboviruses. I. Distribution of the resistance gene among various mouse populations and characteristics of gene expression in vivo. J Infect Dis. 1974;129 :240–7 10.1093/infdis/129.3.240 4816304
27. Langevin SA, Arroyo J, Monath TP, Komar N Host-range restriction of chimeric yellow fever-West Nile vaccine in fish crows (Corvus ossifragus). Am J Trop Med Hyg. 2003;69 :78–80 12932102
28. Langevin SA, Bunning M, Davis B, Komar N Experimental infection of chickens as candidate sentinels for West Nile virus. Emerg Infect Dis. 2001;7 :726–9 10.3201/eid0704.010422 11585538
|
15663854
|
PMC1237116
|
CC BY
|
2022-01-31 23:19:09
|
no
|
Emerg Infect Dis. 2004 Dec; 10(12):2161-2168
|
utf-8
|
Emerg Infect Dis
| 2,004 |
10.3201/eid1012.040486
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 1620101010.1371/journal.ppat.001000105-PLPA-RA-0007R2plpa-01-01-01Research ArticleImmunologyInfectious DiseasesMicrobiologyPediatricsEubacteriaMus (Mouse)The Role of Innate Immune Responses in the Outcome of Interspecies Competition for Colonization of Mucosal Surfaces Innate Immunity in Microbial CompetitionLysenko Elena S Ratner Adam J Nelson Aaron L Weiser Jeffrey N *Departments of Microbiology and Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania, United States of AmericaPortnoy Daniel A EditorUniversity of California at Berkeley, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 22 7 2005 1 1 e115 3 2005 10 5 2005 Copyright: © 2005 Lysenko et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Since mucosal surfaces may be simultaneously colonized by multiple species, the success of an organism may be determined by its ability to compete with co-inhabitants of its niche. To explore the contribution of host factors to polymicrobial competition, a murine model was used to study the initiation of colonization by Haemophilus influenzae and Streptococcus pneumoniae. Both bacterial species, which occupy a similar microenvironment within the nasopharynx, persisted during colonization when given individually. Co-colonization, however, resulted in rapid clearance of S. pneumoniae from the upper respiratory tract, associated with increased recruitment of neutrophils into paranasal spaces. Systemic depletion of either neutrophil-like cells or complement was sufficient to eliminate this competitive effect, indicating that clearance was likely due to enhanced opsonophagocytic killing. The hypothesis that modulation of opsonophagocytic activity was responsible for host-mediated competition was tested using in vitro killing assays with elicited neutrophil-like cells. Components of H. influenzae (but not S. pneumoniae) stimulated complement-dependent phagocytic killing of S. pneumoniae. Thus, the recruitment and activation of neutrophils through selective microbial pattern recognition may underlie the H. influenzae–induced clearance of S. pneumoniae. This study demonstrates how innate immune responses may mediate competitive interactions between species and dictate the composition of the colonizing flora.
Synopsis
Bacterial infection commonly begins with organisms that colonize and proliferate on mucosal surfaces. These microenvironments may be occupied by multiple microbial species, suggesting that successful colonizers are distinguished by their capacity to prevail over their competitors. This study examines interactions between two bacterial species that both colonize and infect the human upper respiratory tract. In a mouse model, strains of both Haemophilus influenzae and Streptococcus pneumoniae efficiently colonize the nasal mucosa when tested individually. In contrast, following co-inoculation, H. influenzae rapidly and completely outcompetes S. pneumoniae. This competitive effect is dependent on the local responses from the host in the form of a specific type of white blood cell (neutrophil) that acts to engulf and kill microorganisms that have been labeled by proteins that bind to microbial surfaces (complement). The results of this study show that recognition of microbial products from one species may activate inflammatory responses that promote the clearance of another competing species. This study also demonstrates how manipulations such as antibiotics or vaccines, which are meant to diminish the presence of a single pathogen, may inadvertently alter the competitive interactions of complex microbial communities.
Citation:Lysenko ES, Ratner AJ, Nelson AL, Weiser JN (2005) The role of innate immune responses in the outcome of interspecies competition for colonization of mucosal surfaces. PLoS Pathog 1(1): e1.
==== Body
Introduction
For many microorganisms, including some that have the potential to behave as pathogens, their primary interaction with a host is through stable colonization of mucosal surfaces [1]. The composition of the flora that inhabits these sites is, in general, highly specific to a particular host species, suggesting that host factors must be an important determinant in the selection of colonizing microbes [1]. In few instances is the molecular basis of the host–microbial interaction that leads to this highly specific relationship understood.
Host–microbial relationships are commonly studied using experimental systems that examine single microbial species. Yet the mucosal surfaces where these organisms reside are often colonized with diverse populations comprised of different species. Successful occupants of such environments must have mechanisms that allow for their persistence by excluding potential competitors in a process referred to as microbial interference [2]. However, specific host or microbial mechanisms that promote or inhibit competitive interactions have not been characterized in vivo [3]. A more thorough understanding of such mechanisms is warranted, since the balance of competitive factors is increasingly being altered by the use of selective antimicrobials or vaccines that target a limited array of colonizing species or strains. In these situations, members of the microflora that might otherwise be suppressed may become predominant and, in some cases, the source of infection.
In prior studies, we have investigated competitive interactions between microbes using two distantly related prokaryotes: the gram-positive Streptococcus pneumoniae and the gram-negative Haemophilus influenzae [4–6]. Both species reside primarily on the mucosal surface of the human nasopharynx and under certain conditions are capable of causing a similar spectrum of disease [7,8]. S. pneumoniae and H. influenzae are among the most prevalent bacterial pathogens causing otitis media in children and community-acquired pneumonia or chronic bronchitis in adults [9–11]. Their shared niche may be a consequence of common mechanisms to promote colonization such as the expression of the cell-surface adhesin phosphorylcholine and a secreted protease that inactivates human immunoglobulin A1 [12–14]. The prevalence of asymptomatic carriage for both species may exceed 50% in some populations, especially infants [15,16]. This suggests that co-colonization might be common, and, therefore, that these species may have evolved specific mechanisms for targeting one another. Several such mechanisms, all of which predict that S. pneumoniae should prevail, have been studied in vitro. These include the rapid killing of H. influenzae by the toxic effects of hydrogen peroxide, which is generated at potentially bactericidal concentrations by the aerobic metabolism of S. pneumoniae [4]. In addition, a cell-surface neuraminidase expressed by S. pneumoniae is capable of removing sialic acid, a structure that decorates the surface of the lipopolysaccharide (LPS) of H. influenzae and contributes to its survival during infection [5].
The purpose of this study was to examine the interaction between these two species in vivo during experimental colonization. We demonstrate that these species compete in a murine model of carriage in a manner opposite to that predicted by in vitro investigation. Our findings demonstrate that species-specific stimulation of innate immune responses may be a determining factor in the outcome of polymicrobial interactions during colonization.
Results
Competition during Nasopharyngeal Colonization
In the course of establishing a mouse model for the colonization of the upper respiratory mucosa by H. influenzae, we found that BALB/c mice with severe combined immunodeficiency (SCID) were susceptible to chronic colonization of the upper respiratory tract with an encapsulated type b isolate (strain Hi636) (Figure 1A) [17]. Because SCID mice can also be colonized by encapsulated S. pneumoniae (strain Sp1121), this observation allowed for testing the effects of co-colonization (Figure 1B) [18]. Immunofluorescent staining of tissue sections from co-colonized mice showed that Hi636 and Sp1121 may co-localize in dense clusters along the epithelial surface and in mucoid material within the lumen of adjacent paranasal air spaces of the anterior nasopharynx (Figure 2). The presence of both species in the same microenvironment of co-colonized animals provided the rationale to analyze their competition either through direct bacterial–bacterial interaction or indirectly through the induction of local host responses.
Figure 1 Colonization of BALB/c SCID Mice by H. influenzae Strain Hi636 and S. pneumoniae Strain Sp1121
The mean density of H. influenzae strain Hi636 (A) and S. pneumoniae strain Sp1121 (B) in upper respiratory tract lavage ± standard deviation was determined on the day indicated following intranasal inoculation with 1 × 107 CFU of either one (open symbols) or both species (solid symbols). * p < 0.01 compared to co-colonization. Dashed line denotes the lower limit of detection.
Figure 2 Immunofluorescence Showing Co-Localization of H. influenzae and S. pneumoniae in the Murine Nasopharynx
BALB/c SCID mice were co-colonized with Hi636 and Sp1121, and at 24 h post-inoculation adjacent 5-μm frozen parasagittal tissue sections through the lateral nasal spaces of the same animals were stained with anti-capsular polysaccharide serum specific to type b H. influenzae (A), type 23F S. pneumoniae (B), or secondary antibody control with no primary antibody (C). Magnification, 400×.
Quantitative cultures of upper respiratory tract lavage fluid showed that over a 2-wk period there was stable colonization by Hi636 that was unaffected by co-colonization with Sp1121 (see Figure 1A). In contrast, colonization by Sp1121 was significantly reduced by day 1 post-inoculation in mice simultaneously challenged with Hi636 in the contralateral naris (p < 0.01; Figure 1B). By day 14 post-inoculation, there was no detectible Sp1121 in cultures obtained from dual inoculated animals. A similar competitive effect of H. influenzae on S. pneumoniae was observed even when S. pneumoniae colonization was pre-established by inoculation 24 h prior to intranasal challenge with H. influenzae (data not shown). Since the inhibition of Hi636 on Sp1121 colonization contrasted with the previously demonstrated bactericidal effect of S. pneumoniae on H. influenzae during co-culture in vitro, we pursued the hypothesis that a host response to the combined presence of both species mediates competition in vivo [4,19].
The presence of competition in SCID animals indicated that the mechanism responsible for the inhibitory effect of Hi636 on Sp1121 colonization is independent of adaptive immunity. Subsequently, we demonstrated that immunocompetent C3H/HeOuJ mice were susceptible to nasopharyngeal colonization (>24 h at a density of >104 CFU/ml of upper respiratory tract lavage) by Hi636 (Figure 3). In C3H/HeOuJ mice, a competitive effect of Hi636 on Sp1121 similar to that observed in SCID mice was also seen by 24 h post-inoculation (p < 0.01).
Figure 3 Competition between Species during Co-Colonization of Immunocompetent Mice
The density of H. influenzae strain Hi636 (Hi) and S. pneumoniae strain Sp1121 (Sp) in upper respiratory tract lavage was determined at 24 h after intranasal inoculation of C3H/HeOuJ mice. Box-and-whiskers plot indicates high and low values, median, and interquartile ranges; n ≥ 10 in each group. Co-inoculated species shown in parentheses. The lower limit of detection for bacteria in lavage culture was 102 CFU/ml.
Neutrophils and Complement Are Required for Competitive Interactions
Histopathological examination of nasopharyngeal sections showed a minimal cellular inflammatory response within the epithelium or subepithelium of mock or singly colonized SCID or immunocompetent mice (Figure 4A and 4C), as previously described [20]. Dual colonized animals, however, showed a marked influx of cells, with a predominance of neutrophils, confined to the lumen of the lateral nasal air spaces, indicative of an acute, localized, suppurative rhinitis (Figure 4B and 4D). The influx of these cells correlated with an increased concentration of macrophage inhibitory protein 2 (MIP-2), one of the murine CXC chemokines that can recruit neutrophils, in upper respiratory tract lavage fluid from co-colonized SCID and C3H mice (p < 0.03 compared to mock colonized; Figure 5).
Figure 4 H&E-Stained Parasagittal Sections Showing the Luminal Space between Two Adjacent Nasal Turbinates
Representative sections are shown for BALB/c SCID mice mock colonized (A) or co-colonized with H. influenzae strain Hi636 and S. pneumoniae strain Sp1121 (B); C3H/HeOuJ mice mock colonized (C) or co-colonized with Hi636 and Sp1121 (D). Arrows indicate cells infiltrating into the lumen of nasal spaces in co-colonized mice. Under higher magnification these cells have the morphologic appearance of neutrophils. Magnification, 400× (insert magnification, 1,000×).
Figure 5 The Effect of Co-Colonization on the Concentration of MIP-2 in Upper Respiratory Tract Lavage Fluid
MIP-2 levels normalized to total protein content were determined in mock colonized (open bars) and dual colonized (solid bars) mice including BALB/c SCID mice, C3H/HeOuJ mice, and C3H/HeOuJ mice treated with RB6-8C5 or rat IgG control as indicated. Values are geometric means ± standard deviation. *p < 0.01 compared to other co-colonized groups of C3H mice. Dual colonized SCID and C3H mice showed significantly higher levels of MIP-2 compared to mock colonized controls (p < 0.03).
To examine whether the influx of neutrophils contributed to the competitive interaction between species, C3H/HeOuJ mice were pretreated with RB6-8C5, a rat monoclonal antibody (mAb) to murine Ly-6G expressed on neutrophil-like cells, prior to bacterial challenge. Treatment with RB6-8C5 but not control antibody resulted in complete loss of the inhibitory effect of H. influenzae on S. pneumoniae (Figure 6A). The effect of RB6-8C5 correlated with the depletion of neutrophils from peripheral blood and in the lumen of the lateral nasal spaces in tissue sections of co-colonized mice (Figure 6B). This absence of an influx of neutrophil-like cells in RB6-8C5-treated mice occurred despite significantly higher concentrations of MIP-2 compared to other co-colonized groups (p < 0.01) and was suggestive of a loss of feedback inhibition on expression of this chemokine (see Figure 5).
Figure 6 The Effect of Depletion of Neutrophil-Like Cells on Competition between Species during Co-Colonization
(A) The density of H. influenzae strain Hi636 (Hi) and S. pneumoniae strain Sp1121 (Sp) in upper respiratory tract lavage was determined at 24 h after intranasal inoculation in C3H/HeOuJ mice pretreated with RB6-8C5 to deplete neutrophil-like cells or pretreated with rat IgG as a control. Box-and-whiskers plot indicates high and low values, median, and interquartile ranges; n ≥ 5 in each group. Co-inoculated species shown in parentheses. The lower limit of detection for bacteria in lavage culture was 102 CFU/ml.
(B) Representative parasagittal sections of the lateral nasal tissues adjacent to the turbinates of co-colonized C3H/HeOuJ mice: (i) untreated control (H&E-stained), (ii) pretreated to deplete neutrophil-like cells (H&E-stained), or (iii) untreated control (stained with mAb Ly-6G to detect neutrophil-like cells). Magnification, 400×.
A similar loss of the inhibitory effect of H. influenzae on S. pneumoniae colonization was seen in C3H/HeOuJ mice pretreated with cobra venom factor to deplete complement (Figure 7). Together these observations suggested that loss of pneumococci from the mucosal surface resulted from opsonization by complement, followed by phagocytic clearance by Ly-6G-positive neutrophil-like cells.
Figure 7 The Effect of Complement Depletion on Competition between Species during Co-Colonization
The density of H. influenzae strain Hi636 (Hi) and S. pneumoniae strain Sp1121 (Sp) in upper respiratory tract lavage was determined at 24 h after intranasal inoculation in C3H/HeOuJ mice pretreated with cobra venom factor (CoVF) to deplete complement or PBS control. Box-and-whiskers plot indicates high and low values, median, and interquartile ranges; n ≥ 5 in each group. Co-inoculated species shown in parentheses. The lower limit of detection for bacteria in lavage culture was 102 CFU/ml.
Stimulation of Neutrophil-Mediated Killing by Innate Recognition of H. influenzae
To test whether the inflammatory cells recruited in response to co-colonization were sufficient to account for interspecies competition, neutrophil-enriched peritoneal exudate cells (PECs) were analyzed in ex vivo killing assays using murine complement. Neutrophil-enriched PECs elicited by administration of heat-inactivated Hi636 (104–106 bacteria) in casein showed a dose-dependent increase in their ability to kill S. pneumoniae compared to controls consisting of casein alone (p < 0.001) (Figure 8A). Hi636 stimulation of neutrophil-enriched PECs, however, had no effect on survival of H. influenzae (Figure 8B). Increased pneumococcal killing by H. influenzae–stimulated neutrophil-enriched PECs required active complement (Figure 8A) but was independent of the presence of specific antibody, since serum obtained from SCID mice provided a sufficient source of complement for killing assays. Increased pneumococcal killing by H. influenzae–stimulated neutrophil-enriched PECs also correlated with an increase in the proportion of Ly-6G-positive PECs co-expressing the activation marker CD11b/CD18, which has recently been shown to be important in defense against pneumococcal infection (Figure 8C) [21]. Heat-inactivated Hi675, a non-typeable H. influenzae isolate, showed a similar capacity to elicit activated neutrophil-like cells and stimulate killing of S. pneumoniae (Figure 8A). In contrast to observations with heat-inactivated H. influenzae, intraperitoneal administration of equivalent doses of heat-inactivated Sp1121 neither stimulated activation of neutrophil-enriched PECs nor killing of either bacterial species (Figure 8A and 8B, and data not shown). Moreover, co-administration of heat-inactivated Sp1121 together with heat-inactivated Hi636 did not appear to add to levels of pneumococcal killing by neutrophil-enriched PECs compared to administration of heat-inactivated Hi636 alone (data not shown). Prior treatment of animals with RB6-8C5 mAb eliminated H. influenzae–induced killing of S. pneumoniae associated with the loss of elicited neutrophil-like cells, as also demonstrated during in vivo competition experiments (Figure 8A). Together these results suggest that the innate immune response to components of H. influenzae was sufficient to stimulate increased opsonophagocytic clearance of S. pneumoniae by neutrophil-like cells during co-colonization.
Figure 8 Stimulation of Neutrophil-Like Cells and Opsonophagocytic Killing by Components of H. influenzae
Killing of S. pneumoniae strain Sp1121 (A) or H. influenzae strain Hi636 (B) by neutrophil-enriched PECs was determined over a 45-min incubation with active complement. The effect on killing by neutrophil-enriched PECs of i.p. administration of heat-inactivated whole bacteria at the dose indicated is shown relative to controls consisting of casein administration alone. No stimulation of killing was observed in controls using inactivated complement (−C) or cells from animals pretreated with mAb RB6-5C8. Values represent three or more independent determinations in duplicate ± standard error of the mean of the percent killing relative to controls using neutrophil-enriched PECs elicited without the addition of heat-inactivated bacteria. *p < 0.001 compared to groups without heat-inactivated H. influenzae and active complement.
(C) A representative FACS analysis of neutrophil-enriched PECs showing the effect of i.p. administration of casein alone (i) or with 106 heat-inactivated Hi636 (ii) on the proportion of Ly-6G positive cells co-expressing the activation marker CD11b/CD18 (boxed).
Discussion
This study demonstrates that the composition of the colonizing flora may be affected by competition between multiple microbial species through the innate host responses they induce. Either of the two species analyzed in this report persisted on the mucosal surface of mice when given individually, but in combination one species quickly and consistently became predominant. This competitive relationship between species, moreover, was the result of the host's response to co-colonization and was not predicted by in vitro investigation of direct bacterial–bacterial interactions [4].
In the example of microbial interference described in this study, clearance of S. pneumoniae required both complement and the recruitment of neutrophil-like cells to the mucosal surface. A central role of these host factors was not unexpected since complement-mediated opsonization followed by ingestion and killing by professional phagocytes such as neutrophils or macrophages is a major host defense against this encapsulated gram-positive pathogen [22]. A further consideration in defining the contribution of these components of innate immunity is that inflammatory responses to polymicrobial colonization may be markedly different from those to a single type of organism. In an earlier report, we described the synergistic proinflammatory responses of respiratory epithelial cells in vitro and in the nasal mucosa in vivo when exposed to H. influenzae and S. pneumoniae in combination [20]. Discrete signals from each species contribute to levels of the neutrophil-recruiting chemokine MIP-2 (or IL-8 in human epithelial cells) significantly greater that seen with either organism alone. This synergistic response of the epithelium correlates with an influx of neutrophil-like cells into the nasal spaces. In the current study, which examines the outcome of these inflammatory responses, it was not practical to obtain adequate numbers of neutrophils directly from the nasal spaces to address whether these were sufficient for the clearance of S. pneumoniae. When tested ex vivo, murine neutrophils enriched from PECs did not kill S. pneumoniae Sp1121 efficiently. This suggested that the recruitment of neutrophils may not be sufficient to account for the competitive effect described here. Rather, stimulation of neutrophil-like cells with bacterial components of H. influenzae was required for efficient clearance of S. pneumoniae. Killing in these assays was dependent on active complement, consistent with its role as an opsonin promoting phagocytosis. If similar events occur in the local environment of the nasopharynx, innate immune responses, consisting of complement and enhanced neutrophil recruitment and activation through recognition of microbial products, may underlie the host's role in clearance of colonizing bacteria from the mucosal surface.
Activation and enhanced opsonophagocytosis of neutrophil-like cells ex vivo was found in response to products of H. influenzae but not equivalent doses of products from S. pneumoniae. The selective innate responses of inflammatory cells such as neutrophils to products from one microbe but not another, therefore, may provide a mechanism whereby one species induces clearance of a competitor. The LPS of H. influenzae and multiple other cellular components, including peptidoglycan, lipoproteins, phosphorylcholine, and an incompletely characterized soluble cytoplasmic fraction, have been implicated in promoting inflammatory responses [23–27]. Purified LPS from other species has been shown to trigger an increase in the migration, life span, and activity of neutrophils [28–30]. The molecular nature of the signal(s) from H. influenzae mediating the recruitment and activity of neutrophils is the topic of ongoing investigation. Results to date do not suggest that purified LPS of H. influenzae is sufficient to stimulate killing of S. pneumoniae by neutrophil-enriched PECs (data not shown). Although components of S. pneumoniae, including cell wall fragments and lipoteichoic acid and its toxin (pneumolysin), have been shown to be potent inducers of inflammation, these appeared to be at least 100-fold less active on a per cell basis in generating the neutrophil responses described here [31–34]. H. influenzae products, furthermore, stimulated killing of another species (S. pneumoniae), but had no effect in opsonophagocytic killing assays on the same species and strain from which these products were derived. Thus, our findings demonstrate that one species may compete with another through selective induction of host responses and may benefit from the differences in its susceptibility to the antimicrobial host factors it induces.
This study shows the importance of specific innate immune responses in dictating the initial success of a species in becoming established within a competitive niche such as the mucosal surface of the nasopharynx. Selective microbial pattern recognition, as demonstrated here for phagocytic activity, may act in the setting of a complex milieu of organisms that differ in their ability to trigger these host-specific responses. This process ultimately selects for the persistence of those species best able to evade the local host clearance factors induced by polymicrobial stimulaton of the innate immune system. The role of innate immunity in colonization described here is distinct from its more extensively studied role in infection.
An additional consideration is that the clearance of S. pneumoniae in co-colonized SCID mice demonstrates that the effects of complement and neutrophil-like cells were independent of adaptive immunity and the presence of antibody. Antibody-independent clearance was also demonstrated by in vitro assays in which we observed efficient killing in the presence of serum lacking anti-phagocytic antibodies. Antibody-independent opsonophagoctic killing of S. pneumoniae, as previously recognized in the classic studies of Wood et al., may be important in protection during the critical period prior to the acquisition of specific anti-capsular antibody [35]. Activation of phagocytic cells, however, has not been a feature of standardized opsonophagocytic killing assays for S. pneumoniae [36].
Colonization of mucosal surfaces is often the first step in the development of disease for many important pathogens. This study demonstrates that the presence of one species may impact the ability of another to persist in the same microenvironment on a mucosal surface. The focus of this report is on bacteria that commonly colonize and potentially infect the respiratory tract of humans. There may be clinical relevance to our observations that in a model of dual colonization H. influenzae was able to induce responses that caused the complete elimination of S. pneumoniae, a leading opportunistic pathogen. In regard to colonization, numerous surveys have described carriage rates for H. influenzae and S. pneumoniae, although, to our knowledge, none appear to have examined the effect of colonization by one species on the density of the other in a quantitative manner. In regard to disease involving the respiratory tract, however, some reports suggest that antagonism between these two species may occur in the natural host [37,38]. Most H. influenzae disease is currently caused by non-typeable strains, which were not tested in in vivo experiments, because of their less efficient colonization of either SCID or immunocompetent mice compared to the type b strain used in our study (data not shown). Nonetheless, an isolate of non-typeable H. influenzae was shown to be equally effective in stimulating neutrophil-mediated killing.
The competitive interactions described in this report may also be applicable to other combinations of microbes where there is evidence of antagonism in vivo. For example, a previously unrecognized competitive interaction between S. pneumoniae and Staphylococcus aureus could explain recent reports that children who receive the pneumococcal conjugate vaccine have lower rates of vaccine-type S. pneumoniae carriage, but higher rates of Sta. aureus nasal colonization as well as otitis media caused by Sta. aureus [39–41]. In this regard, the composition of the normal flora is generally regarded as a critical factor in protection from potentially more virulent opportunistic organisms. Our study provides an initial mechanistic understanding of how manipulation of the colonizing flora could have unexpected consequences on competitors. Since an expanding number of medical interventions impact the composition of the microflora, it would seem prudent to more fully appreciate the scope of competitive interactions on mucosal surfaces.
Our findings also demonstrate that the success of an organism in initiating carriage may depend on its ability to resist innate clearance mechanisms of mucosal surfaces generated in the setting of polymicrobial stimulation. Since characteristics that enhance evasion of innate immunity are often critical determinants of microbial pathogenicity, competition between species may promote the selection for virulence among species such as S. pneumoniae and H. influenzae that must first establish a niche on heavily colonized surfaces.
Materials and Methods
Bacterial strains and culture conditions.
H. influenzae and S. pneumoniae strains were grown as described elsewhere [42]. Strains used in vivo were selected because of their ability to efficiently colonize the murine nasopharynx and included Hi636 (a type b capsule-expressing, spontaneously streptomycin-resistant mutant of H. influenzae strain Eagan that was genetically modified to constitutively express phosphorylcholine), and Sp1121 (a type 23F capsule-expressing S. pneumoniae isolate from the human nasopharynx) [43,44]. Hi675 is a spontaneously streptomycin-resistant mutant of a non-typeable H. influenzae clinical isolate (A860516) provided from the collection of Dr. Loek van Alphen.
Mouse model of nasopharyngeal colonization.
Six-week-old C.B-17/lcrCrlBR (BALB/c SCID, Charles River Laboratories, Wilmington, Massachusetts, United States) or C3H/HeOuJ (toll-like receptor 4 sufficient, Jackson Laboratory, Bar Harbor, Maine, United States) mice were housed in accordance with institutional animal care and use committee protocols. Mice were used in a previously described model of nasopharyngeal colonization with S. pneumoniae [18]. Briefly, groups of at least five mice per condition were inoculated intranasally with 10 μl containing 1 × 107 CFU of PBS-washed, mid-log phase H. influenzae,
S. pneumoniae, or both applied separately to each naris. Unless specified otherwise, 24 h post-inoculation the animal was sacrificed, the trachea cannulated, and 200 μl of PBS instilled. Lavage fluid was collected from the nares for determination of viable counts of bacteria in serial dilutions plated on selective medium containing antibiotics to inhibit the growth of contaminants (streptomycin, 100 μg/ml, to select for H. influenzae
Hi636 and neomycin, 20 μg/ml, to select for S. pneumoniae Sp1121).
Neutrophil and complement depletion.
mAb RB6-8C5, a rat anti-mouse IgG2b directed against Ly-6G on the surface of murine myeloid (and limited subpopulations of lymphoid) lineage cells , was purified from ascites of nude mice given the RB6-8C5 hybridoma [45,46]. To deplete neutrophils, 150 μg of mAb/animal was administered by intraperitoneal (i.p.) injection 24 h prior to intranasal challenge with bacteria. This dose was shown in pilot experiments to result in peripheral blood neutropenia (<50 granulocytes/μl) for a period of at least 48 h. Controls were given the equivalent i.p. dose of total rat IgG (Sigma, St. Louis, Missouri, United States).
Hypocomplementemia was induced by i.p. injection of 25 μg/animal of cobra venom factor (Quidel, San Diego, California, United States) in PBS 18 h prior to bacterial challenge. This procedure was previously shown to reduce levels of immunodetectible C3 to less than 3% of normal and result in a period of hypocomplementemia of greater than 48 h [47].
Histology and immunofluorescence.
After the collection of lavage fluid, heads were fixed by serial overnight incubations in 4% paraformaldehyde, Decal (Decal Chemicals, Congers, New York, United States), and 70% ethanol. Paraffin-imbedded tissue was sectioned and stained with hematoxilin and eosin (H&E). For immunofluorescence, some samples were frozen using Tissue-Tek O.C.T. embedding medium (Sakura Finetek, Torrance, California, United States) in a Tissue-Tek Cryomold, and 5-μm-thick sections were cut, air dried, and fixed in acetone at 4 °C. Sections were rehydrated in PBS and incubated for 30 min in blocking solution of 5% normal goat serum in PBS. After washing in PBS, sections were incubated for 60 min at room temperature with primary antibody consisting of polyclonal rabbit anti–type b H. influenzae (DIFCO, Detroit, Michigan, United States) or anti-type 23F S. pneumoniae (Statens Serum Institut, Copenhagen, Denmark) diluted 1:1,000 in blocking solution. After further washing in PBS, secondary antibody (Texas Red–labeled goat anti-rabbit IgG, ICN Diagnostics, Orangeburg, New York, United States) was added in blocking solution for 60 min at room temperature and detected by fluorescence microscopy.
To label neutrophils, fixed, paraffin-embedded sections were rehydrated through a series of xylene and ethanol washes. Slides were then microwaved in 10 mM citric acid buffer (pH 6.0) and quenched for endogonous peroxidases with 3% hydrogen peroxide. Endogenous biotin was blocked with an Avidin-Biotin blocking kit (Vector Laboratories, Burlingame, California, United States) followed by a peptide blocking reagent (Coulter Immunotech, Hialeah, Florida, United States). Anti-mouse Ly-6G primary antibody was diluted to 1:1,000 in PBS containing 0.1% BSA and 0.2% Triton X-100 for overnight incubation at 4 °C. A biotinylated anti-rat secondary antibody was added for 30 min at 37 °C followed by avidin-horseradish peroxidase ABC reagent (Vector Laboratories). Signal was developed using DAB kit (Vector Laboratories).
Measurement of MIP-2 concentration.
Upper respiratory tract lavage fluid was assayed for the concentration of MIP-2 by ELISA in duplicate according to the manufacturer's instructions (PharMingen, San Diego, California, United States). These values were normalized to the total protein of these samples (micro BCA protein assay, Pierce Biotechnology, Rockford, Illinois, United States).
Isolation and characterization of murine neutrophils.
Neutrophil-enriched PECs were isolated from C3H/HeOuJ mice as previously described [48]. Briefly, phagocytes were obtained by lavage of the peritoneal cavity (8 ml/animal with PBS containing 0.02 M EDTA) of mice treated 24 h and again 2 h prior to cell harvest by i.p. administration of 10% casein in PBS (1 ml per dose). Cells collected from the peritoneal cavity cells were enriched for neutrophils using separation in a Ficoll density gradient centrifugation using Mono-Poly Resolving Medium according to the manufacturer's instructions (MP Biomedicals, Irvine, California, United States). This neutrophil-enriched fraction was collected and washed with 5 ml of Hank's buffer without Ca++ or Mg++ (Invitrogen, Carlsbad, California, United States) plus 0.1% gelatin. An aliquot of these cells was characterized using FACS for staining of granulocytes with anti-mouse Ly-6G mAb (BD Biosciences, San Jose, California, United States) [49]. Expression of the cell-surface activation marker CD11b/CD18 (BD Biosciences) was quantified for cells co-staining for expression of Ly-6G [50]. Where indicated, heat-inactivated H. influenzae (Hi636 or Hi675) and S. pneumoniae (Sp1121) were co-administered with the casein solution. PBS-washed, mid-log phase bacteria were heat-inactivated by treatment at 65 °C for 30 min and shown to be non-viable.
Phagocytic killing assays.
Neutrophil-enriched PECs were counted by trypan blue staining and adjusted to a density of 7 × 106 cells/ml. Killing during a 45-min incubation at 37 °C with rotation was assessed by combining 102 PBS-washed, mid-log phase bacteria (in 10 μl) with complement source (in 20 μl), 105 mouse phagocytes (in 40 μl), and Hank's buffer with Ca++ and Mg++ (Gibco, San Diego, California, United States) plus 1% gelatin (130 μl). Earlier time points and fewer effector cells relative to the number of target cells were shown in pilot experiments to result in less killing. The complement source consisted of serum from either SCID or X-linked immunodeficient mice (CBA/CaHN-Btkxid/J) previously shown to lack opsonophagocytic antibody to thymus-independent type 2 antigens including pneumococcal capsular polysaccharide and phosphorylcholine [51]. After stopping the reaction by incubation at 4 °C, viable counts were determined in serial dilutions. Percent killing was determined relative to the same experimental condition without i.p. administration of bacterial products (casein alone). No loss of bacterial viability was seen in controls using heat-inactivated complement (56 °C for 30 min). Additional controls consisting of heat-inactivated Hi636 administered without casein gave similar levels of killing, confirming that killing was stimulated by bacterial products rather than by casein.
Statistical analysis.
Statistical comparisons of colonization among groups were made by the Kruskal-Wallis test with Dunn's post-test (Prism 4, GraphPad Software, San Diego, California, United States). In vitro killing assays were compared by ANOVA.
We thank Drs. N. Boyko and J. Cebra for input into animal experiments. This work was supported by grants from the US Public Health Service to JNW (AI44231, AI05467, and AI38446) and The Morphology Core of the Center for the Molecular Studies of Liver and Digestive Disease (P30 DK50306). AJR was supported by a Pediatric Infectious Disease Society–St. Jude Fellowship.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. ESL, AJR, and JNW conceived and designed the experiments. ESL and ALN performed the experiments. ESL, AJR, ALN, and JNW analyzed the data. AJR and JNW wrote the paper.
Abbreviations
H&Ehematoxylin and eosin
i.p.intraperitoneal
LPSlipopolysaccharide
mAbmonoclonal antibody
MIP-2macrophage inhibitory protein 2
PECperitoneal exudate cell
SCIDsevere combined immunodeficiency
==== Refs
References
Nataro J Cohen P Mobley H Weiser J 2005 Colonization of mucosal surfaces Washington (DC) ASM Press 456 p.
Brogden K Guthmiller J Taylor C 2005 Human polymicrobial infections Lancet 365 253 255 15652608
Ratner A 2005 Competitive and cooperative interactions in the respiratory microflora Nataro J Cohen P Mobley H Weiser J Colonization of mucosal surfaces Washington (DC) ASM Press 87 96
Pericone C Overweg K Hermanns PW Weiser JN 2000 Inhibitory and bactericidal effects of hydrogen peroxide production by Streptococcus pneumoniae on other inhabitants of the upper respiratory tract Infect Immun 68 3990 3997 10858213
Shakhnovich E King S Weiser J 2002 Neuraminidase expressed by Streptococcus pneumoniae desialylates the lipopolysaccharide of Neisseria meningitidis and Haemophilus influenzae: A paradigm for interbacterial competition among pathogens of the human respiratory tract Infect Immun 70 7161 7164 12438402
Weiser J Bae D Fasching C Scamurra R Ratner A 2003 Antibody-enhanced pneumococcal adherence requires IgA1 protease Proc Natl Acad Sci U S A 100 415 420
Weiser J 2004 Mechanisms of carriage Tuomanen E The pneumococcus 169 182
Murphy T 2003 Respiratory infections caused by non-typeable Haemophilus influenzae
Curr Opin Infect Dis 16 129 134 12734445
Klein JO 1997 Role of nontypeable Haemophilus influenzae in pediatric respiratory tract infections Pediatr Infect Dis 16 S5 S8
Sethi S Murphy T 2001 Bacterial infection in chronic obstructive pulmonary disease in 2000: A state-of-the-art review Clin Microbiol Rev 14 336 363 11292642
Musher D Spindel S 1996 Community-acquired pneumonia Curr Clin Top Infect Dis 16 102 124 8714250
Swords W Buscher B Ver Steeg Ii K Preston A Nichols W 2000 Non-typeable Haemophilus influenzae adhere to and invade human bronchial epithelial cells via an interaction of lipooligosaccharide with the PAF receptor Mol Microbiol 37 13 27 10931302
Cundell DR Gerard NP Gerard C Idanpaan-Heikkila I Tuomanen EI 1995
Streptococcus pneumoniae anchor to activated human cells by the receptor for platelet-activating factor Nature 377 435 438 7566121
Kilian M Reinholdt J Lomholt H Poulsen K Frandsen E 1996 Biological significance of IgA1 proteases in bacterial colonization and pathogenesis: Critical evaluation of experimental evidence APMIS 104 321 338 8703438
De Lencastre H Kristinsson K Brito-Avo A Sanches I Sa-Leao R 1999 Carriage of respiratory tract pathogens and molecular epidemiology of Streptococcus pneumoniae colonization in healthy children attending day care centers in Lisbon, Portugal Microb Drug Resist 5 19 29 10332718
Masuda K Masuda R Nishi J Tokuda K Yoshinaga M 2002 Incidences of nasopharyngeal colonization of respiratory bacterial pathogens in Japanese children attending day-care centers Pediatr Int 44 376 380 12139560
Boyko N Kobuley D Weiser J Cebra J 2004 Are B1 cells selectively and specifically stimulable by antigen based on the BCR? Clin Inv Med 27 S4
McCool T Weiser J 2004 Limited role of antibody in clearance of Streptococcus pneumoniae in a murine model of colonization Infect Immun 72 5807 5813 15385481
Shuto T Xu H Wang B Han J Kai H 2001 Activation of NK-kappa B by nontypeable Haemophilus influenzae is mediated by toll-like receptor 2-TAK1-dependent NIK-IKK alpha/beta-I kappa B alpha and MKK3/6-p38 MAP kinase signaling pathways in epithelial cells Proc Natl Acad Sci U S A 98 8774 8779 11438700
Ratner A Lysenko E Paul M Weiser J 2005 Synergistic proinflammatory responses induced by polymicrobial colonization of epithelial surfaces Proc Natl Acad Sci U S A 102 3429 3434 15728393
Rijneveld A de Vos A Florquin S Verbeek J van der Poll T 2005 CD11b limits bacterial outgrowth and dissemination during murine pneumococcal pneumonia J Infect Dis 191 1755 1760 15838804
White B 1979 The biology of the pneumococcus Cambridge Harvard University Press 799 p.
Wang B Cleary P Xu H Li J 2003 Up-regulation of interleukin-8 by novel small cytoplasmic molecules of nontypeable Haemophilus influenzae via p38 and extracellular signal-regulated kinase pathways Infect Immun 71 5523 5530 14500470
Burroughs M Rozdzinski E Geelen S Tuomanen E 1993 A structure-activity relationship for induction of meningeal inflammation by muramyl peptides J Clin Invest 92 297 302 8325996
Leake E Holmes K Lim D DeMaria T 1994 Peptidoglycan isolated from nontypeable Haemophilus influenzae induces experimental otitis media in the chinchilla J Infect Dis 170 1532 1538 7995993
Swords W Ketterer M Shao J Campbell C Weiser J 2001 Binding of the non-typeable Haemophilus influenzae lipooligosaccharide to the PAF receptor initiates host cell signalling Cell Microbiol 3 525 536 11488814
Clemans DL Bauer RJ Hanson JA Hobbs MV St Geme JW 3rd 2000 Induction of proinflammatory cytokines from human respiratory epithelial cells after stimulation by nontypeable Haemophilus influenzae
Infect Immun 68 4430 4440 10899840
Remer K Brcic M Jungi T 2003 Toll-like receptor-4 is involved in eliciting an LPS-induced oxidative burst in neutrophils Immunol Lett 85 75 80 12505201
Fan J Malik A 2003 Toll-like receptor-4 (TLR-4) signaling augments chemokine-induced neutrophil migration by modulating cell surface expression of chemokine receptors Nat Med 9 315 321 12592402
Sabroe I Prince L Jones E Horsburgh M Foster S 2003 Selective roles for Toll-like receptor (TLR)2 and TLR4 in the regulation of neutrophil activation and life span J Immunol 170 5268 5275 12734376
Tuomanen E Liu H Hengstler B Zak O Tomasz A 1985 The induction of meningeal inflammation by components of the pneumococcal cell wall J Infect Dis 151 859 868 3989321
Schroder N Morath S Alexander C Hamann L Hartung T 2003 Lipoteichoic acid (LTA) of Streptococcus pneumoniae and Staphylococcus aureus activates immune cells via Toll-like receptor (TLR)-2, lipopolysaccharide-binding protein (LBP), and CD14, whereas TLR-4 and MD-2 are not involved J Biol Chem 278 15587 15594 12594207
Cockeran R Theron A Steel H Matlola N Mitchell T 2001 Proinflammatory interactions of pneumolysin with human neutrophils J Infect Dis 183 604 611 11170986
Malley R Henneke P Morse S Cieslewicz M Lipsitch M 2003 Recognition of pneumolysin by Toll-like receptor 4 confers resistance to pneumococcal infection Proc Natl Acad Sci U S A 100 1966 1971 12569171
Wood WBJ Smith MR Watson B 1946 Studies on the mechanism of recovery in pneumococcal pneumonia. IV. The mechanism of phagocytosis in the absence of antibody J Exp Med 84 387 401 19871577
Romero-Steiner S Libutti D Pais LB Dykes J Anderson P 1997 Standardization of an opsonophagocytic assay for the measurement of functional antibody activity against Streptococcus pneumoniae using differentiated HL-60 cells Clin Diag Lab Immunol 4 415 422
May RJ 1954 Pathogenic bacteria in chronic bronchitis Lancet 2 839 842
Luotonen J 1982
Streptococcus pneumoniae and Haemophilus influenzae in nasal cultures during acute otitis media Acta Otolaryngol 93 295 299 6977986
Bogaert D de Groot R Hermans P 2004
Streptococcus pneumoniae colonisation: The key to pneumococcal disease Lancet Infect Dis 4 144 154 14998500
Veenhoven R Bogaert D Uiterwaal C Brouwer C Kiezebrink H 2003 Effect of conjugate pneumococcal vaccine followed by polysaccharide pneumococcal vaccine on recurrent acute otitis media: A randomised study Lancet 361 2189 2195 12842372
Regev-Yochay G Dagan R Raz M Carmeli Y Shainberg B 2004 Association between carriage of Streptococcus pneumoniae and Staphylococcus aureus in children JAMA 292 716 720 15304469
Gould J Weiser J 2002 The inhibitory effect of C-reactive protein on bacterial phosphorylcholine-platelet activating factor receptor mediated adherence is blocked by surfactant J Infect Dis 186 361 371 12134232
Lysenko E Richards J Cox A Stewart A Martin A 2000 The position of phosphorylcholine on the lipopolysaccharide of Haemophilus influenzae affects binding and sensitivity to C-reactive protein mediated killing Mol Microbiol 35 234 245 10632893
McCool TL Cate TR Moy G Weiser JN 2002 The immune response to pneumococcal proteins during experimental human carriage J Exp Med 195 359 365 11828011
Hestdal K Ruscetti F Ihle J Jacobsen S Dubois C 1991 Characterization and regulation of RB6-8C5 antigen expression on murine bone marrow cells J Immunol 147 22 28 1711076
Bryder D Sasaki Y Borge O Jacobsen SE 2004 Deceptive multilineage reconstitution analysis of mice transplanted with hemopoietic stem cells, and implications for assessment of stem cell numbers and lineage potentials J Immunol 172 1548 1552 14734733
Szalai A Digerness S Agrawal A Kearney J Bucy R 2000 The Arthus reaction in rodents: Species-specific requirement of complement J Immunol 164 463 468 10605043
Luo Y Dorf M 1997 Isolation of mouse neutrophils Current protocols in immunology New York Wiley 3.20.21 3.20.26
Fleming T Fleming M Malek T 1993 Selective expression of Ly-6G on myeloid lineage cells in mouse bone marrow. RB6-8C5 mAb to granulocyte-differentiation antigen (Gr-1) detects members of the Ly-6 family J Immunol 151 2399 2408 8360469
van Eeden S Klut M Walker B Hogg J 1999 The use of flow cytometry to measure neutrophil function J Immunol Methods 232 23 43 10618507
Rawlings D Saffran D Tsukada S Largaespada D Grimaldi J 1993 Mutation of unique region of Bruton's tyrosine kinase in immunodeficient XID mice Science 261 358 8332901
|
16201010
|
PMC1238736
|
CC BY
|
2021-01-05 12:11:27
|
no
|
PLoS Pathog. 2005 Sep 22; 1(1):e1
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010001
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 10.1371/journal.ppat.001000205-PLPA-RA-0016R2plpa-01-01-04Research ArticleCell BiologyInfectious DiseasesMicrobiologyEubacteriaInteraction between Polyketide Synthase and Transporter Suggests Coupled Synthesis and Export of Virulence Lipid in M. tuberculosis
Coupled Polyketide Synthesis and TransportJain Madhulika Cox Jeffery S *Department of Microbiology and Immunology, University of California, San Francisco, California, United States of AmericaPortnoy Daniel EditorUniversity of California at Berkeley, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 1 e231 3 2005 3 6 2005 Copyright: © 2005 Jain and Cox.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Virulent mycobacteria utilize surface-exposed polyketides to interact with host cells, but the mechanism by which these hydrophobic molecules are transported across the cell envelope to the surface of the bacteria is poorly understood. Phthiocerol dimycocerosate (PDIM), a surface-exposed polyketide lipid necessary for Mycobacterium tuberculosis virulence, is the product of several polyketide synthases including PpsE. Transport of PDIM requires MmpL7, a member of the MmpL family of RND permeases. Here we show that a domain of MmpL7 biochemically interacts with PpsE, the first report of an interaction between a biosynthetic enzyme and its cognate transporter. Overexpression of the interaction domain of MmpL7 acts as a dominant negative to PDIM synthesis by poisoning the interaction between synthase and transporter. This suggests that MmpL7 acts in complex with the synthesis machinery to efficiently transport PDIM across the cell membrane. Coordination of synthesis and transport may not only be a feature of MmpL-mediated transport in M. tuberculosis, but may also represent a general mechanism of polyketide export in many different microorganisms.
Synopsis
Pathogenic microbes have developed sophisticated strategies to evade the human immune system and establish infection. Mycobacterium tuberculosis, the causative agent of tuberculosis, exports a wide array of lipid virulence factors to the cell surface and into host cells, where they can interact with the host immune system. A strategy to combat M. tuberculosis infections may be to interfere with the bacterium's ability to make and secrete these lipids. In the authors' efforts to understand this process, they have found that synthesis and export of a key lipid virulence factor are coupled. They propose that the synthesis and transport proteins form a complex that promotes efficient lipid export. The coordination of lipid synthesis and export, analogous to co-translational translocation of secreted proteins, may be a general mechanism employed by many different microorganisms to actively transport hydrophobic molecules from cells.
Citation:Jain M, Cox JS (2005) Interaction between polyketide synthase and transporter suggests coupled synthesis and export of virulence lipid in M. tuberculosis. PLoS Pathog 1(1): e2.
==== Body
Introduction
Mycobacterium tuberculosis, the causative agent of tuberculosis, has an extraordinary ability to resist the bactericidal mechanisms of the human immune system [1]. The integrity of the complex mycobacterial cell wall and its associated lipids has long been thought to be important for virulence [2–6]. This complex lipid metabolism is reflected in the enormous capacity of M. tuberculosis for polyketide lipid synthesis: there are at least 24 different polyketide synthases and numerous lipid modification enzymes encoded in its genome [7]. Recent evidence has shown that surface-exposed polyketides, associated with the outer leaflet of the cell wall, are central to the pathogenesis of M. tuberculosis [8–11]. Genes involved in the synthesis and transport of these polyketides are required for bacterial growth and virulence in mice. Furthermore, these lipids have been shown to provide a direct physical barrier to host-induced damage [12], as well as to modulate the immune response to infection by altering cytokine profiles in macrophages [13–15].
Many of the surface-exposed polyketides in M. tuberculosis require active transport from their site of synthesis in the cytoplasmic membrane to the cell surface. Unfortunately, little is known about the mechanism of polyketide secretion. MmpL7, a member of the MmpL (mycobacterial membrane protein large) family of proteins in M. tuberculosis, was the first MmpL shown to be required for the transport of a specific polyketide, phthiocerol dimycocerosate (PDIM) to the cell surface [8]. A number of genes involved in PDIM biogenesis have also been identified [16–18], and the biosynthetic pathway is shown in a schematic in Figure 1A. The PpsA–E gene products extend straight chain fatty acids to phthiocerol, and the mycocerosic acid synthase (Mas) enzyme catalyzes the addition of methyl branches to synthesize mycocerosic acids [16,17,19]. The enzymes FadD26 and FadD28 are thought to be AMP ligases that activate straight chain fatty acids for transfer to the Pps and Mas enzymes [20]. Finally, both MmpL7 and DrrC are required for transport of PDIM to the cell surface [8,10]. MmpL8, another MmpL family member, has been shown to be required for the biogenesis and transport of sulfolipid-1 (SL-1) [9,11]. Given the apparent specificity in MmpL-mediated transport, it is likely that the other eleven MmpL proteins encoded by the M. tuberculosis genome transport specific lipids that may be involved in bacterial virulence.
Figure 1 PDIM Synthesis and Export Pathway and Topology of MmpL7
(A) Schematic of the known steps in the PDIM synthesis and transport pathway. PpsA–E and Mas are polyketide synthases that extend fatty acids to phthiocerol and mycocerosic acids, respectively [16,17]. These are then esterified to produce PDIM. MmpL7 and DrrC are required for the transport of PDIM to the cell surface [8,10]. R is =O (keto) or –OCH3 (methoxy).
(B) Predicted membrane topology of MmpL7 indicating the two non-TM domains 1 and 2.
MmpLs belong to a broader class of proteins termed RND (resistance, nodulation, and cell division) permeases [21]. MmpL7, like most RND transporters, is predicted to contain twelve transmembrane (TM) domains with two non-TM domains between TM #1 and #2 (domain 1) and between TM #7 and #8 (domain 2) (Figure 1B). The structure of AcrB, a well studied RND transporter involved in multidrug efflux in Escherichia coli, has been determined, and its non-TM domains have been localized to the periplasm [22,23]. AcrB interacts with an outer membrane protein, TolC, to expel a variety of drugs from the cytoplasmic membrane across the entire cell wall [24,25]. Since MmpLs are similar in predicted topology to RND permeases, and are required for the transport of cytoplasmic polyketide substrates, we reasoned that they may also interact with other proteins that are involved in the transport of lipids to the cell surface. To better understand the transport of PDIM to the outer leaflet of the cell wall, we probed for interactors of MmpL7. Surprisingly, we found an interaction between MmpL7 domain 2 and the PDIM synthetic enzyme PpsE that is required for the final step of phthiocerol synthesis. Because PpsE acts in the cytoplasm, where PDIM is synthesized, the interaction suggests that MmpL7 domain 2 is accessible to the cytoplasm. We propose that MmpL7 acts as a scaffold to recruit PDIM synthetic machinery, forming a complex that coordinates lipid synthesis and transport.
Results
Identification of MmpL7–PpsE Interaction
To determine the mechanism of PDIM transport to the surface of M. tuberculosis bacilli, we first sought to identify components that interact with the two large non-TM domains of MmpL7 using a yeast two-hybrid approach (Figure 1B) [26]. We created two “bait” constructs to express either an MmpL7 domain 1 or an MmpL7 domain 2 fusion with the LexA DNA-binding protein in yeast. Although MmpL7 domain 1 activated transcription of both the LEU2 and lacZ yeast reporters on its own, MmpL7 domain 2 expression did not auto-activate the reporters (Figure 2A and 2B) and thus was suitable for screening an M. tuberculosis genomic “prey” library for interactors. We identified positive library clones by selecting for growth on medium lacking leucine followed by blue–white screening on plates containing X-gal. Surprisingly, our screen identified an interaction between MmpL7 domain 2 and a 373-amino-acid fragment of PpsE, the enzyme required for the final step of phthiocerol synthesis (Figure 2).
Figure 2 MmpL7 Domain 2 Interacts with the PDIM Synthesis Enzyme PpsE
(A) Yeast two-hybrid reporter strains harboring the indicated bait and prey plasmids were streaked onto solid media with or without leucine. Growth on leucine-negative plates indicates a positive interaction. MmpL7d2, MmpL7 domain 2; MmpL8d2, MmpL8 domain 2.
(B) The same strains as in (A) were transferred onto X-gal-containing indicator plates (inset), and reporter activity was quantified from liquid cultures using β-galactosidase assays.
(C) Linear representation of full-length PpsE protein (1,488 amino acids) with the MmpL7 interaction region denoted. Catalytic domains of PpsE are also shown. ACP, acyl carrier protein; AT, acyl transferase; CE, condensing enzyme; KS, ketoacyl synthase.
A number of observations suggest that the interaction between MmpL7 domain 2 and PpsE is meaningful. First, PpsE was one of the strongest interactors identified in the screen as judged by blue color on indicator plates as well as quantitative β-galactosidase assays (>100-fold increase over controls; Figure 2B). Second, the interaction between MmpL7 domain 2 and PpsE is specific because PpsE did not interact with either LexA alone or with domain 2 of the homologous transporter MmpL8 (Figure 2A and 2B). Western blot analysis using anti-LexA antibodies indicated that the MmpL8 domain 2 and MmpL7 domain 2 baits were expressed at equivalent levels (data not shown). Finally, as PpsE is required for the synthesis of PDIM and MmpL7 is required for the transport of PDIM [8], the interaction between PpsE and MmpL7 suggested the intriguing possibility that PDIM synthesis and transport are coupled.
To confirm the interaction between MmpL7 domain 2 and PpsE in vitro and test the interaction's role in vivo, we first sought to generate a mutant MmpL7 domain 2 that does not interact with PpsE. We used a reverse yeast two-hybrid approach [27] to identify single amino acid changes in MmpL7 domain 2 that disrupt its interaction with PpsE. We engineered our original MmpL7 domain 2 bait vector to include a C-terminal GFP fusion to easily avoid mutations that resulted in protein truncations. We randomly mutagenized MmpL7 domain 2 using error-prone PCR, introduced these constructs into yeast cells bearing the PpsE prey plasmid, and screened for colonies that were both white on X-gal indicator plates and GFP positive. Colonies with the desired phenotype were isolated and the bait plasmids were recovered, retested, and sequenced. Interestingly, of the 12 plasmids that contained mutations that led to single amino acid changes in MmpL7 domain 2, nine clustered in a 50-amino-acid region of the protein (Figure 3A). Eight of the mutations led to substitutions with proline or glycine, which are more likely to be structurally disruptive and possibly lead to global unfolding of the protein. The four remaining mutants, W571R, Y594H, Y594C, and I611S (amino acid numbers correspond to positions in full-length MmpL7) were reconstituted into the original bait vector lacking GFP and assayed for interaction both on X-gal plates and by liquid β-galactosidase assays (Figure 3B). Western blot analysis confirmed that these changes did not lead to differences in protein levels (data not shown). Since the I611S mutation is the most conservative mutation and leads to a large reduction in the interaction with PpsE, this mutant was chosen for subsequent analysis.
Figure 3 Identification of Residues in MmpL7 Domain 2 Required for PpsE Interaction
(A) Twelve MmpL7 domain 2 mutants defective for PpsE binding were isolated in a reverse two-hybrid screen. These amino acid substitutions are displayed on a linear map of MmpL7 domain 2, with changes to amino acids other than proline or glycine in bold. Amino acid numbers correspond to positions in full-length MmpL7. TM domains 7 and 8 are denoted by hatched bars.
(B) Yeast strains expressing the PpsE prey construct and various MmpL7 domain 2 bait plasmids were transferred onto X-gal indicator plates (inset), and reporter activity was quantified from liquid cultures by monitoring β-galactosidase activity.
(C) Beads containing equal amounts of MmpL7 domain 2 and the I611S mutant were incubated with protein extracts containing myc-tagged PpsE and washed. Bound proteins were eluted and separated by SDS-PAGE, and PpsE was visualized by Western blot using anti-myc antibodies. GST-coated beads served as a negative control, and 1% of the protein extract added to the pulldown was loaded as a positive control (“input”).
MmpL7 Domain 2 Interacts with PpsE In Vitro
To independently test the interaction between MmpL7 domain 2 and PpsE identified in the two-hybrid screen, we performed in vitro GST pulldown experiments (Figure 3C). Glutathione agarose beads coated with MmpL7 domain 2 GST fusion protein were incubated with lysates from M. smegmatis cells expressing full-length, myc-tagged PpsE. PpsE interacted with beads coated with MmpL7 domain 2 but not with beads containing only GST protein (Figure 3C, lanes 2 and 3). Furthermore, the I611S mutation led to drastically reduced binding with PpsE (lane 4). Coomassie staining of proteins eluted from the beads demonstrated that equal amounts of MmpL7 domain 2 and the I611S mutant protein were present during the pulldown (data not shown). Importantly, while the I611S mutation had a large effect on PpsE binding, we consistently observed low-level residual binding both by GST pulldown and yeast two-hybrid assays. Taken together, our results demonstrate that MmpL7 domain 2 can specifically interact with PpsE in vitro.
MmpL7 Domain 2 Inhibits PDIM Synthesis In Vivo
We next sought to determine whether the interaction between MmpL7 and PpsE occurs in vivo. Since MmpL7 is an integral membrane protein and PpsE is most likely present in the cytoplasm, we reasoned that MmpL7 domain 2 must be accessible to the cytoplasm in order to interact with PpsE. Overexpression of MmpL7 domain 2 in the cytoplasm would then act as a dominant negative by titrating PpsE away from endogenous full-length MmpL7.
To this end we expressed MmpL7 domain 2 in the cytoplasm of wild-type cells, under the control of the constitutive groEL2 promoter. To assay PDIM synthesis and transport, we labeled cells with 14C-propionate (which is incorporated into PDIM) and extracted surface-exposed lipids from the cell wall. In this procedure, radioactively labeled cells were resuspended in hexanes and then harvested by centrifugation to yield a supernatant fraction that contained surface-exposed lipids, and a pellet fraction that contained the remaining total lipids associated with the cell. Lipids from both fractions were extracted and separated by thin layer chromatography (TLC) in order to resolve PDIM (Figure 4A). As expected, in wild-type cells PDIM was present in both the pellet and supernatant fractions (Figure 4A, lanes 1 and 2), demonstrating that PDIM is both synthesized and transported. In a PDIM synthesis mutant, fadD28
−, PDIM was absent from both fractions (lanes 3 and 4), whereas in the PDIM transport mutant, mmpL7
−, PDIM was present only in the pellet fraction (lanes 5 and 6). Interestingly, overexpression of MmpL7 domain 2 in wild-type cells resulted in more than 98% inhibition of PDIM synthesis (lanes 7 and 8). Although overexpression of MmpL7 domain 2 did not cause any growth defects (data not shown), it was possible that it had a nonspecific effect on lipid synthesis. To confirm that overexpression of MmpL7 domain 2 specifically inhibits PDIM synthesis, we monitored SL-1, which like PDIM also incorporates propionic acid, in the same extracts and observed no differences in SL-1 synthesis or transport (Figure 4A).
Figure 4 MmpL7 Domain 2 Acts as a Dominant Negative Inhibitor of PDIM Synthesis In Vivo
(A) The indicated strains carrying either no plasmid (−), control vector (vector), or a plasmid with MmpL7 domain 2 under the control of the constitutive groEL2 promoter (MmpL7d2) were labeled with 14C-propionate. Surface-exposed lipids (S) were extracted by resuspension in hexanes, and cell pellets (P) were harvested by centrifugation. Lipids from both fractions were extracted and separated by TLC under solvent conditions to separate PDIM (upper panel, keto and methoxy forms) and SL-1 (lower panel).
(B) Top: lipids were extracted as in (A) from pellets of wild-type cells carrying either no plasmid (−), the MmpL7 domain 2 expression plasmid (d2), or the MmpL7 domain 2 expression plasmid with the I611S mutation (d2-I611S). Bottom: Western blot analysis was performed to confirm equivalent expression of wild-type MmpL7 domain 2 and the I611S mutant by using antibodies against the hemagglutinin epitope tag.
(C) Complementation of an mmpL7− M. tuberculosis strain with the wild-type (mmpL7wt) or the I611S mutant mmpL7 (mmpL7I611S). Surface-exposed lipids (S) and lipids associated with the remaining cell pellet (P) were extracted and separated by TLC to resolve PDIM as in (A).
The effect on PDIM synthesis was surprising, as mmpL7
− cells are still able to synthesize PDIM; thus, a simple prediction would have been that a dominant negative form of MmpL7 would inhibit PDIM transport not synthesis. Although the mechanism of the dominant negative effect of domain 2 is unclear, the specific effect of MmpL7 domain 2 on PDIM synthesis, in addition to its specific interaction with PpsE, suggests that MmpL7 domain 2 interacts with PpsE in vivo.
To test whether the dominant negative effect of MmpL7 domain 2 on PDIM synthesis is indeed due to its interaction with PpsE, we overexpressed the I611S mutant form that fails to interact with PpsE. The I611S mutation was generated in the MmpL7 domain 2 overexpression construct and introduced into wild-type M. tuberculosis cells. Although MmpL7 domain 2 I611S was expressed at levels similar to those of the wild-type version, it was unable to inhibit PDIM synthesis to the same extent (Figure 4B). PDIM synthesis in this strain was approximately 50% of wild-type, consistent with the severely reduced, but not completely abolished, interaction of the I611S mutant MmpL7 domain 2 with PpsE. This result strongly suggests that MmpL7 domain 2 inhibits PDIM synthesis via direct inhibition of PpsE.
We reasoned that if MmpL7 interacts with PpsE in vivo to coordinately synthesize and transport PDIM, then an MmpL7 mutant that does not interact with PpsE may be defective for PDIM transport. To test this we introduced the I611S change in the context of full-length MmpL7 and expressed the wild-type and mutant forms of MmpL7 in an mmpL7
− strain. We found that both forms were able to complement the PDIM transport defect (Figure 4C, lanes 8 and 10). Therefore, although the I611S mutation led to decreased interaction between MmpL7 domain 2 and PpsE in earlier experiments, in the context of full-length protein this mutation alone was not sufficient to decrease MmpL7 activity in vivo. Although the reason is unclear, this could be because MmpL7 has multiple interactions with PpsE, and perhaps with other members of the PDIM synthesis and transport machinery, that compensate for this mutation.
Dominant Negative Effect of MmpL7 Domain 2 Requires MmpL7
A simple mechanism to account for the dominant negative effect of MmpL7 domain 2 expression is that the protein interacts directly with PpsE in the cytoplasm in such a way as to render the synthase inactive. Alternatively, because other RND family members act as trimers [22,23], it is possible that MmpL7 domain 2 may inhibit PpsE while in a complex with full-length MmpL7. To distinguish between these two possibilities, we tested whether the dominant negative effect of MmpL7 domain 2 requires the presence of full-length MmpL7. Interestingly, expression of MmpL7 domain 2 in mmpL7− transposon mutant cells failed to inhibit PDIM synthesis (Figure 5A, lane 4). Because the transposon mutant may express fragments of MmpL7, we also created an M. tuberculosis strain in which the full mmpL7 gene was removed (Figure S1). Like the transposon mutant, the ΔmmpL7 strain was unable to transport PDIM (Figure S1C) and was insensitive to the dominant negative effect of MmpL7 domain 2 expression (Figure 5A, lanes 5 and 6). Expression of MmpL7 domain 2 was similar in all three strains (Figure 5B, lanes 2–4). This finding demonstrates that the activity of cytoplasmically expressed MmpL7 domain 2 requires the presence of wild-type MmpL7. Taken together, our results suggest that MmpL7 domain 2 enters into a complex with endogenous MmpL7 that interacts with PpsE, trapping the synthase in an inactive state, thus inhibiting PDIM synthesis.
Figure 5 Dominant Negative Effect of MmpL7 Domain 2 Requires the Presence of Full-Length MmpL7
(A) Wild-type cells, an mmpL7 transposon mutant (mmpL7 −), and a complete mmpL7 knockout (ΔmmpL7) carrying either no plasmid (−) or the MmpL7 domain 2 expression construct (+). Labeled lipids were extracted from pellets as described in Figure 4 and separated by TLC to resolve PDIM.
(B) Western blot of MmpL7 domain 2 showing that it is expressed at equivalent levels in the different M. tuberculosis strains.
Discussion
PDIM, like other polyketide lipids, is a key molecule in the pathogenesis of M. tuberculosis. In this study, we have identified a novel interaction between MmpL7, a protein required for PDIM transport, and PpsE, an enzyme required for PDIM synthesis. Overexpression of the interaction domain of MmpL7 causes a drastic defect in PDIM synthesis, suggesting that this domain interacts with PpsE in vivo and inhibits its activity. To our knowledge, this is the first report of an interaction between a synthetic enzyme and its cognate transporter. We propose that MmpL7 interacts with the PDIM synthetic machinery to form a complex that coordinately synthesizes and transports PDIM across the cell membrane (Figure 6).
Figure 6 Model of PDIM Synthesis and Transport
MmpL7 interacts with PpsE, a subunit of the Pps enzyme required for PDIM synthesis. We propose that MmpL7 acts as a scaffold to recruit PDIM synthesis machinery, including Pps and perhaps Mas, leading to coordinate synthesis and transport of PDIM across the cytoplasmic membrane (CM). Whether MmpL7, or other factors, are required for delivery of PDIM through the peptidoglycan (PG) and mycolyl-arabinogalactan (mAG) layers is unclear.
Interestingly, the dominant negative effect of domain 2 on PDIM synthesis is dependent upon the presence of full-length MmpL7. This strongly suggests that domain 2 incorporates into a complex with endogenous MmpL7 and exerts its effect only in this context. Like AcrB, an RND transporter in E. coli [22,23], MmpL7 may normally act as a trimer or a higher order oligomer, and a hybrid complex of full-length MmpL7 with domain 2 may trap PpsE in an inactive state.
Since MmpL7 is dispensable for PDIM production it is curious that expression of domain 2 inhibits PDIM synthesis. We propose a simple model to reconcile this paradox. In a stepwise process of efficiently coordinating PDIM synthesis and export, MmpL7 may possess both inhibitory and activating activity on PDIM synthesis. For example, domain 2 may act to inhibit PpsE activity until the entire PDIM synthesis–transport complex is assembled, at which point the inhibition is relieved. Thus, MmpL7 domain 2 may exert its dominant negative effect by stabilizing the inhibitory state of this complex. In the absence of MmpL7, however, there is no inhibition or activation of PDIM synthesis, therefore PDIM synthesis is unaffected. This model would also explain why the I611S mutation, when reconstituted into full-length MmpL7, has no apparent effect. If the isoleucine residue is important for inhibition of PpsE, then the I611S mutation would not necessarily lead to a defect in PDIM synthesis or transport.
Since MmpL7 and AcrB share the defining features of RND transporters, it is tempting to draw parallels between the two proteins. Both contain 12 putative TM helices with non-TM loops between TM #1 and #2, and TM #7 and #8. In the crystal structure of AcrB, the non-TM domains are predicted to be periplasmic [22,23] and the TM domains form a central cavity that is accessible to the cytoplasm. TM prediction algorithms (TMPred, TMHMM) suggest that the non-TM domains of MmpL7 are also periplasmic, although no experimental data exist to validate this prediction. Since the interaction domain of MmpL7 lies between TM #7 and #8, in order to interact with PpsE, it must be accessible to the cytoplasm. There are a number of ways in which this could occur. First, like the glutamate transporter EAAT1 [28], the interaction domain of MmpL7 may be reentrant through the membrane and thus interact with PpsE. Alternatively, the PpsE protein may access the extracellular portion of MmpL7 via a central pore created by the MmpL7 TM domains. Finally, since MmpLs and AcrB are distantly related members of the RND permease family, the structure of MmpL7 may differ from AcrB, and the orientation of MmpL7 in the membrane may be such that domain 2 is located in the cytoplasm. Indeed, there are examples of evolutionarily related transporters with opposite membrane topology [29].
Since there is specificity in MmpL-mediated transport, it is attractive to speculate that this specificity may be in part due to the interaction with the cognate transporter. There is evidence in both E. coli and Pseudomonas aeruginosa that when the non-TM regions of two different RND permeases with different drug efflux specificities are swapped, the respective drug specificities are also switched [30,31]. We constructed analogous chimeras between MmpL7 and MmpL8; although these hybrids were expressed, they were nonfunctional (data not shown). Despite the negative result, this suggests that portions other than domains 1 and 2 are required for MmpL function.
Given the results presented here, we propose that MmpL proteins act not only as transporters but also as scaffolds to couple polyketide synthesis and secretion. This model may also provide a framework to explain the role of two other RND family transporters in polyketide synthesis. In M. tuberculosis, mmpL8− mutants are defective for SL-1 synthesis and accumulate a partially lipidated precursor SL1278 [9]. Originally, we proposed that MmpL8 may transport SL1278 across the cell membrane, where subsequent enzymatic steps would convert it to mature SL-1. However, in light of the interaction between MmpL7 and PpsE, it is now tempting to speculate that MmpL8 may similarly recruit a biosynthetic enzyme required to complete the synthesis of SL-1 prior to transport. Likewise, an RND transporter in Streptomyces coelicolor, ActII-ORF3, is also involved in the biogenesis of a polyketide, γ-actinorhodin [32]. Therefore, the coupling of polyketide synthesis and transport via interactions between synthases and cognate transporters may represent a general mechanism utilized by RND family members to efficiently export complex polyketides. This paradigm is reminiscent of protein secretion where newly synthesized polypeptides are co-translationally translocated across the membrane [33]. Coupling of synthesis and transport may be energetically favorable while promoting specificity and directionality in transport processes.
Materials and Methods
Strains and plasmids.
M. tuberculosis cells (Erdman strain) were cultured in 7H9 medium supplemented with 10% OADC, 0.5% glycerol, and 0.1% Tween-80, or on 7H10 solid agar medium with the same supplements except for Tween-80 [8]. Kanamycin (20 μg ml−1) and hygromycin (50 μg ml−1) were used where necessary. All strains and plasmids used in this study are described in Table S1.
Construction of M. tuberculosis knockout strain.
The ΔmmpL7 (MJM39) mutant strain was created by homologous recombination using specialized transducing phage phMJ1 [34]. This deletion replaced all 2,763 bp of mmpL7 with a hygromycin resistance cassette, and correct replacement of the gene was confirmed by Southern blot analysis (Figure S1).
Yeast two-hybrid assays.
The yeast two-hybrid screen was performed as described [26], and expression of bait proteins in yeast was confirmed by Western blotting using antibodies against LexA (kind gift of R. Brent). Bait constructs using MmpL7 domain 2 and MmpL8 domain 2 were created by PCR amplification and insertion into pEG202. The prey library was constructed using random Erdman genomic DNA fragments inserted into pjsc401. Positive interactors were selected on media lacking leucine and then screened by blue–white screening on agar plates containing X-gal (5-bromo-4-chloro-3-indolyl β-D-galactoside). β-galactosidase activity was assayed in yeast cells permeabilized with chloroform and sodium dodecyl sulfate as previously described [35]. For reverse two-hybrid assays, error-prone PCR was performed using Taq polymerase, 100 ng of the bait plasmid containing mmpL7 domain 2 (pMJ2) as the PCR template, and 18 cycles of amplification. Mutagenesis rate was approximately one mutation per kilobase. Yeast homologous recombination was used to introduce the mutant MmpL7 domain 2 PCR product into the bait vector. GFP screening was performed visually by fluorescence microscopy.
GST pulldown binding assays.
Recombinant MmpL7 domain 2–GST, MmpL7 domain 2–I611S–GST, and GST alone were expressed in E. coli strain DH5α by growing cultures to mid-logarithmic phase at 32 °C and inducing with 1 mM IPTG for 30 min. Cells were centrifuged and resuspended in 50 mM Tris, 1 mM EDTA, 150 mM NaCl, 1 mM PMSF, and 1 mM DTT, and lysed by sonication. Lysates were cleared by centrifugation at 12,000 g and incubated with Glutathione agarose beads (G 4510, Sigma, St. Louis, Missouri, United States) at 4 °C overnight. Beads were washed four times with PBS, 1 mM EDTA, and 0.5% Triton X-100, and then stored as a 50% slurry in 50 mM Tris, 1 mM EDTA, 500 mM NaCl, 20% glycerol, and 0.5% Triton X-100. PpsE-myc was expressed in M. smegmatis, and lysates were prepared by bead-beating cells into binding buffer (20 mM Tris, 1 mM EDTA, 150 mM NaCl, 1 mM PMSF, and 1 mM DTT). Lysates were incubated with 50 μl of protein-coated beads overnight at 4 °C, washed three times in binding buffer, and resuspended in SDS sample buffer. Samples were boiled to elute all proteins off the beads and resolved on 7.5% SDS-PAGE gels. PpsE-myc was detected by Western blotting using 9E10 monoclonal antibodies (kind gift of J. M. Bishop).
Biochemical analysis of PDIM and SL-1.
M. tuberculosis cultures were labeled with 14C-propionate, which is incorporated into both PDIM and SL-1. Surface-exposed lipids were extracted by resuspending the cells in hexanes and gently sonicating [9]. Cell pellets, containing the remainder lipids, were harvested by centrifugation. Lipids from both fractions were extracted by the Bligh-Dyer method [36] and analyzed by separation on 10 cm × 10 cm HPTLC plates (Alltech Associates, Deerfield, Illinois, United States) by using either hexanes/ether (9:1) to resolve PDIM or chloroform/methanol/water (60:30:6) to resolve SL-1. Lipid spots on TLC plates were quantified using a phosphorimager.
Protein preparation and analysis.
M. tuberculosis cells were grown into mid-logarithmic phase and harvested by centrifugation. Cell lysates were separated by SDS-PAGE using 12% polyacrylamide gels. Proteins were visualized by immunoblotting using monoclonal antibodies against the hemagglutinin epitope tag (HA.11, Covance, Berkeley, California, United States). Loading was normalized by total protein, and efficiency of transfer was confirmed by Ponceau S staining of the nitrocellulose membrane.
Supporting Information
Figure S1 Creation of ΔmmpL7 in M. tuberculosis by Specialized Transduction
(A) Map of the mmpL7 region in wild-type and the ΔmmpL7 mutant showing the restriction sites and probe location for Southern blot. Genomic DNA from both wild-type and ΔmmpL7 was digested with BamHI, and the blot was probed with a 1,029-bp 5′ flank to mmpL7 revealing a 2,636-bp fragment for wild-type and a 4,700-bp fragment for the mutant.
(B) Southern blot of BamHI-digested genomic DNA from indicated strains.
(C) Surface-exposed lipids (S) and lipids associated with the remaining cell pellet (P) were labeled and extracted from wild-type and ΔmmpL7 cells as described in Figure 4A and then separated by TLC to resolve PDIM.
(29 KB PDF)
Click here for additional data file.
Table S1 Strains and Plasmids Used in This Study
(71 KB DOC)
Click here for additional data file.
We thank H. Madhani, A. Sil, C. Gross, and all members of the Cox laboratory for critical reading of the manuscript. We thank all members of the Sil and Cox laboratories for helpful discussions. We thank J. M. Bishop and R. Brent for antibodies and P. Walter and R. Brent for yeast strains. MJ is supported by a Howard Hughes Medical Institute Predoctoral Fellowship. JSC gratefully acknowledges the support of the Pew Scholars Program in the Biomedical Sciences, the Sandler Family Supporting Foundation, and the W. M. Keck Foundation. This work was supported by National Institutes of Health grant AI68540.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MJ and JSC conceived and designed the experiments. MJ performed the experiments. MJ and JSC analyzed the data and wrote the paper.
Abbreviations
Masmycocerosic acid synthase
PDIMphthiocerol dimycocerosate
SL-1sulfolipid-1
TLCthin layer chromatography
TMtransmembrane
==== Refs
References
McKinney JD Jacobs WR Jr Bloom BR 1998 Persisting problems in tuberculosis Fauci A Krause R Emerging infections London Academic Press 51 146
Goren MB Brokl O Schaefer WB 1974 Lipids of putative relevance to virulence in Mycobacterium tuberculosis: Phthiocerol dimycocerosate and the attenuation indicator lipid Infect Immun 9 150 158 4271720
Chan J Fan XD Hunter SW Brennan PJ Bloom BR 1991 Lipoarabinomannan, a possible virulence factor involved in persistence of Mycobacterium tuberculosis within macrophages Infect Immun 59 1755 1761 1850379
Brennan PJ Nikaido H 1995 The envelope of mycobacteria Annu Rev Biochem 64 29 63 7574484
Ng V Zanazzi G Timpl R Talts JF Salzer JL 2000 Role of the cell wall phenolic glycolipid-1 in the peripheral nerve predilection of Mycobacterium leprae
Cell 103 511 524 11081637
Glickman MS Jacobs WR Jr 2001 Microbial pathogenesis of Mycobacterium tuberculosis: Dawn of a discipline Cell 104 477 485 11239406
Cole ST Brosch R Parkhill J Garnier T Churcher C 1998 Deciphering the biology of Mycobacterium tuberculosis from the complete genome sequence Nature 393 537 544 9634230
Cox JS Chen B McNeil M Jacobs WR Jr 1999 Complex lipid determines tissue-specific replication of Mycobacterium tuberculosis in mice Nature 402 79 83 10573420
Converse SE Mougous JD Leavell MD Leary JA Bertozzi CR 2003 MmpL8 is required for sulfolipid-1 biosynthesis and Mycobacterium tuberculosis virulence Proc Natl Acad Sci U S A 100 6121 6126 12724526
Camacho LR Ensergueix D Perez E Gicquel B Guilhot C 1999 Identification of a virulence gene cluster of Mycobacterium tuberculosis by signature-tagged transposon mutagenesis Mol Microbiol 34 257 267 10564470
Domenech P Reed MB Dowd CS Manca C Kaplan G 2004 The role of MmpL8 in sulfatide biogenesis and virulence of Mycobacterium tuberculosis
J Biol Chem 279 21257 21265 15001577
Camacho LR Constant P Raynaud C Laneelle MA Triccas JA 2001 Analysis of the phthiocerol dimycocerosate locus of Mycobacterium tuberculosis . Evidence that this lipid is involved in the cell wall permeability barrier J Biol Chem 276 19845 19854 11279114
Rao V Fujiwara N Porcelli SA Glickman MS 2005
Mycobacterium tuberculosis controls host innate immune activation through cyclopropane modification of a glycolipid effector molecule J Exp Med 201 535 543 15710652
Rousseau C Winter N Pivert E Bordat Y Neyrolles O 2004 Production of phthiocerol dimycocerosates protects Mycobacterium tuberculosis from the cidal activity of reactive nitrogen intermediates produced by macrophages and modulates the early immune response to infection Cell Microbiol 6 277 287 14764111
Reed MB Domenech P Manca C Su H Barczak AK 2004 A glycolipid of hypervirulent tuberculosis strains that inhibits the innate immune response Nature 431 84 87 15343336
Azad AK Sirakova TD Fernandes ND Kolattukudy PE 1997 Gene knockout reveals a novel gene cluster for the synthesis of a class of cell wall lipids unique to pathogenic mycobacteria J Biol Chem 272 16741 16745 9201977
Azad AK Sirakova TD Rogers LM Kolattukudy PE 1996 Targeted replacement of the mycocerosic acid synthase gene in Mycobacterium bovis BCG produces a mutant that lacks mycosides Proc Natl Acad Sci U S A 93 4787 4792 8643481
Onwueme KC Ferreras JA Buglino J Lima CD Quadri LE 2004 Mycobacterial polyketide-associated proteins are acyltransferases: Proof of principle with Mycobacterium tuberculosis PapA5 Proc Natl Acad Sci U S A 101 4608 4613 15070765
Trivedi OA Arora P Vats A Ansari MZ Tickoo R 2005 Dissecting the mechanism and assembly of a complex virulence mycobacterial lipid Mol Cell 17 631 643 15749014
Trivedi OA Arora P Sridharan V Tickoo R Mohanty D 2004 Enzymic activation and transfer of fatty acids as acyl-adenylates in mycobacteria Nature 428 441 445 15042094
Tseng TT Gratwick KS Kollman J Park D Nies DH 1999 The RND permease superfamily: An ancient, ubiquitous and diverse family that includes human disease and development proteins J Mol Microbiol Biotechnol 1 107 125 10941792
Murakami S Nakashima R Yamashita E Yamaguchi A 2002 Crystal structure of bacterial multidrug efflux transporter AcrB Nature 419 587 593 12374972
Yu EW McDermott G Zgurskaya HI Nikaido H Koshland DE Jr 2003 Structural basis of multiple drug-binding capacity of the AcrB multidrug efflux pump Science 300 976 980 12738864
Touze T Eswaran J Bokma E Koronakis E Hughes C 2004 Interactions underlying assembly of the Escherichia coli AcrAB-TolC multidrug efflux system Mol Microbiol 53 697 706 15228545
Tikhonova EB Zgurskaya HI 2004 AcrA, AcrB, and TolC of Escherichia coli form a stable intermembrane multidrug efflux complex J Biol Chem 279 32116 32124 15155734
Golemis E Serebriiskii I Finley RLJ Kolonin MG Gyuris J 1999 Interaction trap/two-hybrid system to identify interacting proteins In: Ausubel FM, Brent R, Kingston RE, Moore DD, Seidman JG, et al., editors. Current protocols in molecular biology New York John Wiley and Sons 20.1.1 20.1.40
Bennett MA Shern JF Kahn RA 2004 Reverse two-hybrid techniques in the yeast Saccharomyces cerevisiae
Methods Mol Biol 261 313 326 15064466
Seal RP Amara SG 1998 A reentrant loop domain in the glutamate carrier EAAT1 participates in substrate binding and translocation Neuron 21 1487 1498 9883740
Saier MH Jr 2003 Tracing pathways of transport protein evolution Mol Microbiol 48 1145 1156 12787345
Elkins CA Nikaido H 2002 Substrate specificity of the RND-type multidrug efflux pumps AcrB and AcrD of Escherichia coli is determined predominantly by two large periplasmic loops J Bacteriol 184 6490 6498 12426336
Tikhonova EB Wang Q Zgurskaya HI 2002 Chimeric analysis of the multicomponent multidrug efflux transporters from gram-negative bacteria J Bacteriol 184 6499 6507 12426337
Bystrykh LV Fernandez-Moreno MA Herrema JK Malpartida F Hopwood DA 1996 Production of actinorhodin-related “blue pigments” by Streptomyces coelicolor A3(2) J Bacteriol 178 2238 2244 8636024
Keenan RJ Freymann DM Stroud RM Walter P 2001 The signal recognition particle Annu Rev Biochem 70 755 775 11395422
Glickman MS Cox JS Jacobs WR Jr 2000 A novel mycolic acid cyclopropane synthetase is required for coding, persistence, and virulence of Mycobacterium tuberculosis
Mol Cell 5 717 727 10882107
Guarente L Ptashne M 1981 Fusion of Escherichia coli lacZ to the cytochrome c gene of Saccharomyces cerevisiae
Proc Natl Acad Sci U S A 78 2199 2203 6264467
Bligh EG Dyer WJ 1959 A rapid method of total lipid extraction and purification Can J Biochem Physiol 37 911 917 13671378
|
16201014
|
PMC1238737
|
CC BY
|
2021-01-05 12:11:27
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e2
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010002
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 10.1371/journal.ppat.001000305-PLPA-RA-0003R1plpa-01-01-05Research ArticleEvolutionHIV - AIDSImmunologyInfectious DiseasesVirologyVirusesPrimatesAnimalsSIVsm Quasispecies Adaptation to a New Simian Host SIV Adaptation to a New HostDemma Linda J 1¤aLogsdon John M Jr.1¤bVanderford Thomas H 1Feinberg Mark B 2¤cStaprans Silvija I 2*
1 Department of Biology, Emory University, Atlanta, Georgia, United States of America
2 Departments of Medicine and Microbiology and Immunology, and Emory Vaccine Center, Emory University School of Medicine, Atlanta, Georgia, United States of America
Koup Richard A EditorNational Institutes of Health, United States of America*To whom correspondence should be addressed. E-mail: [email protected]¤a Current address: Foodborne and Diarrheal Disease Branch, Centers for Disease Control and Prevention, Atlanta, Georgia, United States of America
¤b Current address: Department of Biological Sciences, Roy J. Carver Center for Comparative Genomics, University of Iowa, Iowa City, Iowa, United States of America
¤c Current address: Merck Vaccine Division, Merck and Company, West Point, Pennsylvania, United States of America
9 2005 30 9 2005 1 1 e310 3 2005 20 6 2005 Copyright: © 2005 Demma et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Despite the potential for infectious agents harbored by other species to become emerging human pathogens, little is known about why some agents establish successful cross-species transmission, while others do not. The simian immunodeficiency viruses (SIVs), certain variants of which gave rise to the human HIV-1 and HIV-2 epidemics, have demonstrated tremendous success in infecting new host species, both simian and human. SIVsm from sooty mangabeys appears to have infected humans on several occasions, and was readily transmitted to nonnatural Asian macaque species, providing animal models of AIDS. Here we describe the first in-depth analysis of the tremendous SIVsm quasispecies sequence variation harbored by individual sooty mangabeys, and how this diverse quasispecies adapts to two different host species—new nonnatural rhesus macaque hosts and natural sooty mangabey hosts. Viral adaptation to rhesus macaques was associated with the immediate amplification of a phylogenetically related subset of envelope (env) variants. These variants contained a shorter variable region 1 loop and lacked two specific glycosylation sites, which may be selected for during acute infection. In contrast, transfer of SIVsm to its natural host did not subject the quasispecies to any significant selective pressures or bottleneck. After 100 d postinfection, variants more closely representative of the source inoculum reemerged in the macaques. This study describes an approach for elucidating how pathogens adapt to new host species, and highlights the particular importance of SIVsm env diversity in enabling cross-species transmission. The replicative advantage of a subset of SIVsm variants in macaques may be related to features of target cells or receptors that are specific to the new host environment, and may involve CD4-independent engagement of a viral coreceptor conserved among primates.
Synopsis
Why do some infectious agents establish successful cross-species transmission while others do not? Despite the clear potential for diseases harbored by animals to become emerging human pathogens, this question remains unanswered. Certain simian immunodeficiency viruses (SIVs) responsible for the human HIV-1 and HIV-2 epidemics have succeeded in infecting new host species, including humans. This study provides clues to how an SIV adapts to a new host in an experimental cross-species transmission. Indeed, many emerging diseases are caused by highly mutation-prone RNA viruses like SIV, which exist not as a single species, but rather as a population of genetic variants within a single infection. The presence of numerous viral variants in an infected animal increases the chance that variants with the ability to enter into or multiply in a new host species are present. This study describes how an SIV population from a natural reservoir host, the sooty mangabey, adapts to a new monkey species, the rhesus macaque. A limited subset of SIV variants containing unique viral surface proteins appears well suited to multiply in the new host. This study documents how viral variation facilitates cross-species transmission, and highlights the particular importance of immunodeficiency virus envelope variants in infecting new hosts.
Citation:Demma LJ, Logsdon JM Jr, Vanderford TH, Feinberg MB, Staprans SI (2005) SIVsm quasispecies adaptation to a new simian host. PLoS Pathog 1(1): e3.
==== Body
Introduction
At least 40 primate species in Africa are infected by diverse simian immunodeficiency viruses (SIVs) assigned to six major phylogenetic lineages; however, the mosaic nature of the SIV genomes attests to the common simian-to-simian transmission of SIVs [1,2]. These African nonhuman primate reservoir hosts maintain normal CD4 T cell counts and avoid AIDS, despite lifelong SIV infection [3–6]. Our studies of naturally SIV-infected sooty mangabeys (SMs) indicate that these hosts are highly viremic, yet manifest far lower levels of aberrant immune activation and apoptosis than are seen in pathogenic SIV and HIV infections; these latter observations help to explain how SMs maintain numerically and functionally intact T lymphocyte populations [3]. Zoonotic transmission and sustained propagation of SIVcpz and SIVsm from SIV-infected chimpanzees and SMs, respectively, to humans [2,7], resulted in the human HIV-1 and HIV-2 AIDS epidemics.
SIV and HIV env sequence variation, including variation in length and glycosylation patterns, enables these viruses to utilize different coreceptors for infection, and to adapt to variation in the relative levels of the viral receptor (CD4) and coreceptors (e.g., CC-chemokine receptor 5 [CCR5]) to gain efficient entry into cells [8,9]. Env variation also enables the virus to readily escape antibody responses [10–12]. Our studies of SIVsm env diversity in naturally infected SMs demonstrate high levels of intrahost env variable region 1 and 2 (V1V2) amino acid diversity (median, 5.6%; range, 0%–38%) that are maintained by continual positive selection, presumably antibody mediated (unpublished data). Considerable V1V2 amino acid length variation and high and variable numbers of glycosylation consensus sequences are also observed. This high diversity of SIV V1V2 in the natural host environment may promote the potential for cross-species transmission by generating the env variants necessary to ensure successful infection of new hosts.
For successful cross-species transmission to occur, including the continued propagation of an infectious agent in a new host species, the agent must be able to replicate at levels in the new host that ensure its sustained passage to new individuals of that species; otherwise the newly infected host(s) will simply represent a “dead-end” infection that does not lead to secondary and sustained infections in the new species. Alternatively, the infectious agent that has been recently transmitted to a new host may require the accumulation of mutations that enable it to replicate at levels high enough to ensure continued transmission to new individuals. Thus, SIVs that are capable of quickly adapting to new hosts and replicating to high levels are most likely to successfully breach the species barrier and continue to spread in the new species. Adaptation of naturally occurring SIV quasispecies to new hosts has not been studied. In studies analyzing the adaptation of diverse HIV-1 quasispecies from identifiable human donors to newly infected “recipients,” the early expansion of viruses that are homogeneous in env sequences, macrophage-tropic, and CCR5-utilizing is described [13,14]. This sequence homogenization is not observed in gag [15], suggesting that multiple variants are transmitted, followed by selection for particular env variants during primary infection. Selection for env homogeneity has also been reported after parenteral inoculation [13], suggesting that, separate from any selective processes taking place at the mucosal barrier, there is strong selection for particular env genotypes during acute infection. Recently, a study of heterosexual HIV-1 transmission demonstrated that viruses encoding compact, glycan-restricted Envs with exposed neutralizing epitopes were significantly favored in newly infected hosts [16]. Another report confirmed these findings in transmission of subtype A but not B [17]. Whether these observations extend to other clades, cohorts, or routes of HIV infection remains to be determined [18,19].
Here we describe the adaptation of diverse SIVsm quasispecies to the new rhesus macaque (RM) host, and compare quasispecies evolution in natural SM and nonnatural RM hosts. During the first days of infection, SIVsm replicated as well in the RM host as in the original host, if not better, apparently due to the robust replicative capacity of a subset of viral variants containing a shorter V1 loop and lacking two specific glycosylation sites. This study demonstrates how viral quasispecies diversity, by providing multiple variants, some of which can replicate to high levels in new hosts, may facilitate cross-species transmission.
Results
High Diversity of the SIVsm Quasispecies Inoculum
The uncloned SIVsm inoculum consisting of plasma from a naturally infected SMs contained 4 × 106 SIV RNA copies/ml. To characterize the diversity of this source inoculum (SI), and the molecular behavior of the quasispecies upon transmission to new hosts, we analyzed a 456-nucleotide region spanning the variable V1V2 region of env and a 421-nucleotide region of the p27 capsid region of the more functionally conserved gag gene (GenBank accession numbers AY852284–AY853166). We chose to sequence only portions of the coding sequences of these two genes, as efforts to amplify full-length coding sequences resulted in poor RT-PCR amplification efficiencies that were not compatible with the reliable sampling of multiple quasispecies variants. Sequences representing actively replicating SIV were amplified directly from virion RNA in the plasma by RT-PCR. 29 V1V2 and 7 gag clone sequences analyzed using maximum parsimony, neighbor-joining (NJ), and maximum likelihood (ML) phylogenetic tree constructions demonstrated that the SI was phylogenetically distinct from commonly used laboratory SIV isolates (Figure S1).
SI V1V2 sequence length varied between 139 and 143 aa (Table 1). The range of pairwise nucleotide diversity calculated for the SI population was 0.3%–5.1% for V1V2 (mean, 2.7%; median, 2.7%) and 0.7%-4.6% for gag (mean, 2.4%; median, 2.3%). The amino acid diversity ranged from 0% to 12.8% in V1V2 (mean, 5.9%; median, 6.3%) with only six of 406 identical pairwise comparisons. The amino acid diversity in gag p27 was lower, ranging from 0% to 0.7% (mean, 0.2%; median, 0%). The minimal diversity detected in gag, which was PCR-amplified using identical conditions, suggests that the observed V1V2 diversity is not the result of PCR-introduced mutation. The average viral diversity observed in this study is similar to that reported in other studies of SIV infections of natural hosts [4,20,21]. The within-host extremes of V1V2 diversity observed in this and another study (unpublished data) is a novel observation resulting from the large number of sequences analyzed.
Table 1 V1V2 Amnio Acid Variation in RMs and SMs at Days 10 and 14 and in SI
Robust Virus Replication Demonstrates Immediate Quasispecies Adaptation to the New Nonnatural RM Host
SMs and RMs were intravenously (IV) inoculated with 1 ml of the SI described above (SIVsm). IV injection may partially recapitulate the circumstances of cross-species SIV transmission, which are thought to involve exposure to bloody flesh during hunting or butchering [22,23]. It ensures reproducible infection and enables the study of host-specific differences in response to SIV infection that lead to AIDS in RMs but not SMs [24]. At day 7 postinfection (p.i.), SIVsm replication was detected in all animals except RM2 (Figure S2). Peak viremia occurred between days 10 and 14 for all animals except RM2, whose peak likely occurred between days 14 and 28, an interval when no sampling was performed. RM1 and RM3 manifested peak viremia levels of 1.6 × 109 copies/ml of plasma and 6.1 × 108 copies/ml respectively, higher than the peak viremia for the three SMs (5.0 × 107 to 1.5 × 108 copies/ml). Viral loads declined to similar set point levels of ~1 × 106 copies/ml, except for RM2, which maintained fewer than 1,000 copies/ml after day 60. RM1 and RM3 developed AIDS 2.5 and 3.5 y p.i. and were euthanized. Divergent host responses and disease outcomes during primary SIVsm infection of SMs and RMs are described elsewhere [24].
Early Amplification of a Phylogenetically Related Subset of env Variants in RMs Contrasts with Unrestricted env Diversity in SMs
At day 14 p.i., the replicating env V1V2 sequences for all six animals were compared to each other and to the SI. Despite the more robust replication of SIVsm in the RMs, few of the SI V1V2 variants appeared among the clade containing most of the variants replicating in RMs (clade 1; Figure 1), demonstrating that a subset of genetically related env variants was amplified during acute SIVsm infection of RMs. Specifically, the proportion of variants replicating in RMs in clade 1 (97%) was significantly higher than that of either the SI variants (21%; Marascuillo Procedure, p < 0.0001) or the variants replicating in SMs (22%; p < 0.0001). In contrast, there was little selection of specific SIVsm env variants upon transfer to new naïve SMs, with no significant difference (Marascuillo Procedure, p = 99%) in the distributions of SI variants and variants replicating in SMs between clades 1 and 2 (Figure 1). Mean intrahost pairwise nucleotide diversity in the RMs at day 14 was 0.9%, significantly lower than that of the SMs (2.1%; Tukey's HSD, p < 0.05) and the SI (Newman-Keuls, p < 0.05). Mean intrahost amino acid diversity in RMs at day 14 (1.6%) was lower than SM amino acid diversity (3.4%), although not significantly. This pattern of restriction was observed as early as 10 d p.i., with 37/39 (95%) of RM variants clustering with the same six SI variants seen at day 14 (unpublished data). Thus, only a subset of env genotypes appear well suited to replicate in the new RM host environment, but this subset replicates to surprisingly high levels.
Figure 1 Phylogenetic Tree of V1V2 Variants
An NJ tree of all day 14 and SI V1V2 variants, constructed using a GTR model of evolution. Bootstrap support values greater than 50% are shown in italics at nodes and the number of multiple clones from the same animal at the ends of branches is indicated beside the symbol.
SIVsm env Variants That Are Preferentially Amplified in Newly Infected RMs Contain Shorter V1 Regions and Lack Two Glycosylation Consensus Sequences
SIV env glycosylation is important in receptor and coreceptor utilization [25], and in evading neutralizing antibodies [26,27]. Eight predicted N-linked glycosylation sites (N-gly) containing the amino acid motif NXT/S were identified along the 125 aa of V1V2 analyzed, although direct evidence of glycosylation at these sites is not explicitly demonstrated. Among the 29 SI clones analyzed, most of these positions encoded the consensus N-gly sequence (Figures 2 and S3), consistent with the observation that SIVsm V1V2 is highly glycosylated in its natural host (unpublished data).
Figure 2 Pattern and Temporal Dynamics of Protein Sequence Evolution in Envelope
V1V2 amino acid sequence for SI, SM1, and RM1 at days 10 and 578. The consensus of all sequences is indicated at the top with amino acid positions labeled above. Glycosylation consensus motifs (NXT/S) are highlighted in yellow.
Two of these predicted N-gly sites, at positions 7 and 19 in V1 (Figure 2), were immediately selected against in the newly infected RMs, before the anticipated development of an antibody response. (RMs and SMs demonstrated anti-SIV antibodies by ELISA at day 40 p.i., the first time-point assessed, with increasing titers by day 130 p.i. [unpublished data]). In the SI, 76% and 52% of V1V2 sequences exhibited the N-gly motif at positions 7 and 19. At 10 d p.i., the motif at position 7 was present in 80% of SM1 sequences, but in only 5% of RM1 sequences. At position 19, the motif was present in 70% of SM1 sequences compared to 5% of RM1 sequences (Figure 2 for SM1 and RM1 at day 10; Table 1 for frequencies in all animals). The near absence of the motif at positions 7 and 19 was observed in all RMs analyzed at days 10 and 14 (Figure S3). Furthermore, the predominant V1V2 variants in RMs at days 10 and 14 were shorter in length by two amino acids compared to the variants in SM (Table 1). Thus, a disadvantage of variants with longer V1 loops and two specific N-gly sites in V1 may explain the restricted outgrowth of specific env variants during early infection of RMs.
At days 10–100, SIVsm sequences from SMs had a greater mean number of N-gly sites per sequence than variants in RMs, but by day 578 the overall frequency of glycosylation consensus motifs increased in both species, and there was no difference between species (Figure S4; p < 0.001). This increase in mean glycosylation over time in the RMs is in part due to the reemergence of variants containing the two specific N-gly sites that were absent in the majority of early RM variants. The range of V1 region amino acid length variations also increased over time, and no species-specific differences were seen at day 100 and thereafter (unpublished data). These data demonstrate continual evolution of V1 in both SMs and RMs.
Increasing Positive Selection in SMs and RMs at Later Times Postinfection
To compare selection pressures between hosts and time points, nonsynonymous and synonymous nucleotide substitutions (dN and dS, respectively) at each codon of V1V2 were calculated for sequences at day 14 (prior to seroconversion) and day 578 (chronic infection) for each animal and the SI (Figure 3). The pattern of selection in SMs at both time points (Figure 3A and 3B) was similar to the SI (Figure 3E), suggesting few changes in selection pressure upon SIVsm transfer to naïve SMs. This is consistent with the phylogenetic analyses, which indicate that IV transfer of SIVsm does not subject the quasispecies to any significant selective pressures or bottleneck in the natural SM host, but results in considerable restriction of the SIVsm quasispecies diversity in RMs. In contrast to the SMs, the relative lack of sites under strong selection in RMs at day 14 (Figure 3C) corroborates the strong, early selection of a subset of variants from the SI. The subsequent substantial increase (Figure 3D) in the number of sites under selection and the magnitude of selection at those sites not only reflect the outgrowth of variants more similar to the SI, but also suggest the presence of immune-selective pressures in the RMs during the postacute phase of infection.
Figure 3 Evidence for Greater Positive Selection in SIV env at Later Times Postinfection
(A–E) Calculations for dN and dS were performed along the 124 amino acids of the V1V2 region using SNAP (http://hiv-web.lanl.gov/). The average dN and dS at each codon is shown for SMs at day 14 (A) and day 578 (B), as well as for RMs at day 14 (C) and day 578 (D), and for the SI (E). Yellow boxes indicate predicted N-gly sites, and asterisks indicate N-gly sites not present at early time points in RMs.
(F) Cumulative dN and dS are shown across all sites for each animal at day 14 and day 578. Raw values of cumulative dN and cumulative dS are indicated below the graph.
To quantify the magnitude of selection in the SIVsm env V1V2, cumulative dN and cumulative dS were calculated for each animal at days 14 and 578 (Figure 3F). SMs and RMs showed relative increases in cumulative dN and dS at day 578 (Wilcoxon rank sum test, p < 0.005), especially in a region of V1 (amino acid positions 22–57) described as important in antibody escape [25,28–30]. At later times, despite increases of both dN and dS in RMs, cumulative dN-dS was greater in RMs than SMs, although the difference was not statistically significant, suggesting greater positive selection pressures in the non-natural host. However, continual evolution of V1V2 occurred in both species, consistent with observations of persistent within-host positive diversifying selection in SMs (unpublished data).
Variants Related to the Original Inoculum Reemerge in RMs at Later Times Postinfection
Phylogenetic analyses of clones from day 100 p.i. showed V1V2 sequences beginning to diversify in RMs, although the variants remained clustered by host (Figure 4; see Figure S5 for parallel phylogenetic analysis of SM1). At this time, viral sequences in RM1 were more closely related to variants from the SI than to variants from 10 d p.i., suggesting a reemergence of the SI-related quasispecies during chronic infection. These results indicate that day 10 V1V2 variants are an evolutionary “dead-end,” as it is unlikely that directional evolution would result in viral quasispecies in all RMs that are highly related to the original SI quasispecies.
Figure 4 Reemergence of Donor-Related Variants in Late Infection of Nonnatural Hosts
ML tree of sequences from the SI and RM1 at days 10 and 100. Bootstrap values greater than 50% are shown at nodes. The SI variants are identified by the legend.
At all time points after infection of the three SMs, average nucleotide diversity of SIVsm V1V2 sequences remained similar to that of the original SI (~3.0%; Figure 5). In contrast, in RMs, nucleotide diversity increased after day 40 despite manifesting an initial restriction in viral diversity. Viral variants at day 578 became more animal-specific (unpublished data), as would be expected under host-specific selection pressure. The viral diversity in RMs at day 578 (averaging 4.5% ± 0.8%) was greater than both that of the SI and that of the SIVsm variants observed in SMs at late times (t-test with Bonferroni adjustment, p = 0.03). These data suggest that selection pressures change during the course of SIVsm infection of RMs; V1V2 variants that replicated to high levels in primary infection lost their replicative advantage, and previously undetected variants that were closely related to the SI became detectable. The increasing nucleotide diversity over time in RMs is consistent with the observed increase in positive diversifying selection pressures in RMs (see Figure 3).
Figure 5 Temporal Changes in Nucleotide Sequence Evolution in SMs and RMs
Mean pairwise nucleotide diversity of the V1V2 sequences for each animal at each time point, calculated using the Tamura-Nei model of nucleotide substitution in MEGA 2.1 [52]. The diversity of the SI is indicated on the y-axis. Trend lines are drawn for RMs (red) and SMs (blue).
No Early Selection for Specific gag Variants Following Intra- or Cross-Species Transmission
Despite the high levels of selection in env, no species-specific phylogenetic relationships were observed for SIVsm gag variants at day 10 p.i. (Figure 6), indicating that there was no preferential amplification of specific gag variants in association with the establishment of infection in either SMs or RMs. The average nucleotide diversity of gag variants following transmission to both species was similar to that of the SI (unpublished data). These data suggest that most gag variants were equivalent in their ability to establish successful infection of either host. At later times, some amino acid changes in gag became apparent in individual animals (Figure S6). In both RMs analyzed at days 100 and 578, there was almost complete amino acid fixation at two sites (positions 39 and 68). Only one SM manifested any evidence of amino acid fixation in gag, and this was only partial (position 126 in SM2). Fixation of amino acid changes, particularly at position 68, which occurred in two RMs, could be due to cell-mediated immune response pressures, which are thought to be stronger in RMs compared to SMs [24]. However, that these changes are due to random amino acid fixation through genetic drift cannot be ruled out, because of the large population sizes involved and the limited number of gag clones analyzed.
Figure 6 Phylogenetic Analysis of SIVsmm gag Variants in Natural and Nonnatural Hosts
An NJ unrooted phylogenetic tree of all day-10 gag variants was constructed in PAUP* [43] using the GTR model with a gamma rate distribution of shape α = 1.0. The SI variants are represented by triangles and identified by the legend. Bootstrap values greater than 50% are shown at nodes, and the number of multiple clones from the same animal at the ends of branches is indicated within the symbol.
Discussion
Identification of specific characteristics that enable pathogens to infect new species may reveal why some emerging infections become widespread while others do not. RNA virus quasispecies diversity has been posited to underlie their zoonotic success, yet no study had analyzed the behavior of diverse naturally occurring viral quasispecies upon inoculation into different host species. This study represents the first analysis, to our knowledge, of the evolution of a diverse naturally occurring SIV quasispecies, following its side-by-side inoculation into a new nonnatural host species (rhesus macaques) and the natural host species from which it was derived (sooty mangabeys). Our studies, which focused on the intensive sequencing of large numbers of viral variants in the env V1V2 region, point to the importance of diversity in this region in initiating a successful cross-species infection event. However, this does not exclude the possibility that diversity in other genome regions, including diversity in other env regions, plays an important role in cross-species transmission events.
Upon inoculation of SMs with the diverse SIVsm quasispecies, little host restriction was observed during acute infection despite continued strong positive selection pressures consistent with host-specific viral evolution and similar to our findings in a study of natural infection in SMs (unpublished data). However, a restricted, genetically related subset of SIVsm env V1V2 variants that harbored a shorter V1 loop and lacked two specific glycosylation sites was preferentially amplified in all of the RMs during acute infection. This was observed despite IV inoculation, which would have bypassed mucosal barriers, and was observed as early as 10 days p.i., likely prior to the development of immune responses. While we cannot rule out that these variants hitchhiked to a high frequency in the RMs, the observed amplification of a subset of variants appeared to represent an advantage for these envelopes that was related to specific features of target cells in the RM but not the SM. Loss of one of these glycosylation sites (position 7) has been shown to result in CD4-independent SIVs in the SIVmac239 strain [25], suggesting the possibility that viral variants that preferentially use CCR5 independently of CD4 may be selected for during acute infection of the new RM host. It remains to be determined whether loss of this same glycosylation site in SIVsm also results in CD4 independence. Because CCR5 is more highly conserved between SMs and RMs than is CD4 [31,32], it might be anticipated that CD4-independent viral variants could overcome species differences in the primary viral receptor (CD4) and have a distinct advantage in the new host environment. The possibility that efficient coreceptor utilization independent of CD4 is an important factor in establishing cross-species transmission is a topic for further study.
Recently, the selection of more compact, glycan-restricted HIV envs after heterosexual transmission was described [16,17]. If the selection of slightly shorter, glycan-restricted SIVsm envs observed in this study of cross-species transmission is a related phenomenon, then, because our IV inoculations bypassed mucosal barriers, the advantage of such variants may be a posttransmission phenomenon. One possibility is that in an antibody-naïve host environment, more compact, less glycosylated Env conformations with more accessible receptor-binding domains have a replicative advantage. However, it is noteworthy that SIVsms encoding less glycosylated V1V2 regions do not appear to have any replicative advantage in newly infected, antibody-naïve SMs. The lack of selective pressure on the SIVsm quasispecies in acutely infected SMs suggests that highly glycosylated V1V2 variants are well adapted to initiate new infections of its natural host species.
Although it might be expected that only a subset of SM-adapted SIVs could replicate well in a new species, it is intriguing that these variants replicated to levels exceeding those seen in the natural SM host. In this [24] and other studies of acute SIV infection [33,34], we have observed a relationship between the magnitude of early CD4 T cell activation and the magnitude of early virus replication. Given the higher levels of CD4 T cell activation observed in the acutely infected RMs as compared to the SMs in this study [24], it is conceivable that increased numbers of activated CD4 T cells provided additional cellular targets for infection. If this target cell-driven hypothesis of more robust SIVsm replication in RMs is correct, it raises the possibility that SIV infection-induced CD4 T cell activation in nonnatural hosts actually facilitates zoonotic transmission of these CD4 T cell-tropic lentiviruses. Additional studies are required to explore this hypothesis. It is also worth noting that activated CD4 T cells may up-regulate CCR5 (or other coreceptor) expression [35], and down-regulate CD4 expression [36]. This might provide another selective force for the observed glycan-restricted SIVsm variants that may be less CD4 dependent in newly infected RMs. Whatever the explanation for the selection of specific env V1V2 variants in RMs, selection pressures giving rise to these effects need not be strong, given the high level of diversity of the inoculum and the likely number of replication cycles involved. Nonetheless, the robust replication of these variants ensured the establishment of high viremia during infection of a new host, a characteristic that would be important for continued propagation of the virus in the new species.
At later times p.i. of RMs, SIVsm variants more closely related to the original SI quasispecies reemerged, suggesting that all variants were initially transmitted to the RMs, but that only a subset of variants replicated to high levels during the acute infection period. Variants related to the SI may have been physically sequestered in resting memory cells, as has been suggested for HIV-1 [37], or simply replicated at such low levels that they were not sampled. Studies have suggested that “archival” HIV variants are maintained in infected hosts [38]. When host selection pressures change, such as with the termination of antiretroviral therapy, these archived variants may emerge, obviating the necessity for back mutation of the most predominant viral variants at the time of change in selection pressure. This capacity to archive the variants present in a diverse, infecting swarm, referred to as the “molecular memory” of the quasispecies [39], demonstrates the significant potential of lentiviral quasispecies to respond to changing selection pressures and presents significant hurdles when considering HIV prevention or treatment measures. SI V1V2 variant emergence at later times suggests that these viruses have replicative advantages in chronically infected RMs, perhaps due to their resistance to neutralizing antibodies. Compensatory changes in other regions of the genome (e.g., in the CD4-binding region of SIV env) could also have relieved initial selection pressures against these variants.
This study demonstrates how viral quasispecies diversity enables successful cross-species transmission by providing multiple variants, some of which are able to establish high-level viremia in new hosts, which, in turn, increases the probability of successful propagation within new species. Our studies point to SIVsm env diversity in its reservoir host as a likely required, although not necessarily sufficient prerequisite for successful cross-species transmission. These observations have implications for which infectious agents may be zoonotically transmitted and efficiently propagated in a new host species. Finally, the potential roles of CD4-independent SIVs and coreceptor sequence conservation in cross-species transmission are important topics for further study.
Materials and Methods
Experimental SIV infection.
SMs and RMs were housed at the Yerkes National Primate Research Center, Atlanta, Georgia, United States, and maintained in accordance with federal guidelines [40]. Prior to the study, the absence of SIV infection was confirmed by negative SIV PCR of plasma and negative HIV-2 serology for at least 1 y. Three RMs and three SMs were experimentally infected IV with a diverse inoculum of uncloned SIVsm from a naturally infected SM (individual FQi). SMs FLn, FCo, and FGu are referred to as SM1, SM2, and SM3, respectively. RMs RHt4, RQl4, and RZw4 are referred to as RM1, RM2, and RM3, respectively. The animals were followed at multiple time points following the infection, and quantitative PCR was carried out to determine the viral dynamics of their acute SIV infection [24].
PCR.
Viral RNA was extracted from freshly thawed plasma samples from the three SMs and three RMs in this study using the Qiagen Viral RNA Kit. SIV sequences were amplified from 5 μL of template in a PCR reaction using the Qiagen One-Step RT-PCR Kit (Qiagen, Valencia, California, United States).
To amplify the env V1V2 region, a mixture of two forward primers was used, FENV1 (5′-CTTGGGAGAATACAGTCACAG-3′) corresponding to bp 6,780–6,800 of the SIVsmmH4 genome, and FENV2 (5′-CTTGGGAGAATACAGTAACAG-3′) also corresponding to bp 6,780–6,800 but containing one different base at position 6,796. The reverse env V1V2 primer was also a mixture of RENV1 (5′-TAAATCTAATAGCATCCCAATAAT-3′) and RENV2 (5′-TAAATCTAATAGCATCCCAATAGT-3′) corresponding to bp 7,221–7,244 of the SIVsmmH4 genome, and differing at bp 7,222. The primer pair amplified a 456-bp fragments spanning the V1V2 hypervariable region of env.
The gag region was amplified using shortgagF1 (5′-TTAAGTCCAAGAACATTAAATGC-3′) and shortgagR (5′-GTAGAACCTGTCTACATAGCT-3′), which correspond to bp 1,493–1,515 and 1,937–1,957 of SIVsmmH4, respectively, yielding a 421-bp product of the 5′ end of the p27 capsid protein.
Conditions for each reaction were 30 min at 50 °C and 15 min at 95 °C, followed by 40 cycles of 94 °C for 1 min, 52 °C for 30 s, and 72 °C for 1 min. A final extension time was carried out for 5 min at 72 °C. Due to extremely low viral load, RNA from RM2 could not be amplified after day 14 for either V1V2 or gag. RT-PCR sensitivity was determined to be less than 500 copies per reaction. Viral loads from each of the animals did not significantly differ at each time point (with the exception of animal RM3, in which virus was undetectable using the RT-PCR protocol after day 14). Samples were not standardized for input copy number, potentially confounding the extent of change in viral diversity that was measured over time, although this would not confound comparisons between animals at each time point since viral loads were similar.
No-template controls and negative controls from the RNA extraction were used in each set of reactions to ensure that no cross contamination occurred at either step. In addition, samples from each pair of animals, SM1/RM1, SM2/RM2, SM3/RM3 were extracted at least 3 mo apart. This ensured that contamination within species was avoided. Contamination of negative-extraction controls was detected when extracting SM2 and RM2 samples from days 70 and 100. This extraction was repeated, and virus could not be amplified from RM2 due to very low copy numbers. On one occasion, the RT-PCR reaction was contaminated with a particular molecular clone, however these sequences were easy to identify with phylogenetic analysis due to their extensive divergence from the SI. These contaminants were excluded from the analysis. RNA extracted from day 10 plasma in SM1 and RM1 was RT-PCR amplified under the same conditions as above, except that 10 μL of PCR product was removed at 25, 30, 35, and 40 cycles for cloning and sequencing to ensure that PCR bias during extended cycling was not a factor in sample diversity. Viral RNA from days 70 and 100 for RM1 and RM3 was extracted, PCR amplified, and cloned in duplicate to ensure experimental repeatability. A 1:10 dilution of SI was amplified under the same conditions, and 15 clones from this RT-PCR product were sequenced to ensure that input copy number did not bias diversity.
DNA cloning and sequencing.
PCR products from each sample were run on a 1.5% low-melt agarose gel, and the 456-bp V1V2 or 421-bp gag product was extracted and cloned into the pCR4-TOPO vector (TOPO TA Cloning Kit, Invitrogen, Carlsbad, California, United States). Between 15 and 30 V1V2 clones and 5 and 10 gag p27 clones from each time point and each animal were randomly selected and sequenced using the M13F and M13R primers with the dye terminator cycle sequencing method.
Sequence and phylogenetic analyses.
Sequences were aligned using the program CLUSTAL X [41], followed by manual adjustment using MacClade 4.0 [42]. Nonaligned regions of length variation in V1 and V2 were removed (corresponding to nucleotides 6,932–6,974), and sequences containing internal stop codons, single deletions, or double deletions were also excluded from analysis, as these are thought to be PCR artifacts [43]. Figures S1 and S2 show the resulting alignment of all sequences in V1V2 and gag, respectively.
For the SI, maximum parsimony and NJ were implemented using the PAUP 4.0b10* package for V1V2 and gag [44]. For each of the resulting trees, bootstrap support was determined with 1,000 resamplings of the sequences. The most highly supported clade in both the NJ and the parsimony trees was used as the outgroup for all subsequent phylogenetic trees (Figure S3).
For tree construction, the Modeltest program [45] was used to construct and evaluate the DNA substitution models used. Based on the Modeltest results, phylogenetic analysis on sequences obtained from successive time points during the acute infection was performed by ML using the program Treefinder [46]. The general time reversible (GTR) model, which allows for rate variation between sites [47–49], was used, and the shape parameter (α) of the gamma distribution used in this model was estimated, as were base frequencies and substitution rate parameters. Bootstrap support was determined with 1,000 resamplings of the ML tree using distance methods in PAUP4.0b10*, incorporating the estimated rate parameters.
The cumulative number of synonymous and nonsynonymous and nucleotide substitutions was estimated using Synonymous/Nonsynonymous Analysis (SNAP; http://hiv-web.lanl.gov/), which calculates rates of nucleotide substitution from a set of codon-aligned nucleotide sequences, based on the method of Nei and Gojobori [50], and incorporating a statistic developed in Ota and Nei [51]. Viral nucleotide diversity at each time point was determined by calculating the pairwise nucleic acid distances for each of the clones using the method of Tamura and Nei [52] in the program MEGA 2.1 [53]. This same method was also employed to quantify nucleotide divergence from the source, defined as the ratio of the difference in nucleotide diversity between SI and each sample of variants to the total diversity in the two groups. Amino acid diversity was calculated using the gamma distance method in the program Mega 2.1. Phylogenetic trees constructed with synonymous or nonsynonymous sites only were constructed in Mega 2.1 using distance methods incorporating the Tamura-Nei model of nucleotide substitution with gamma-distributed rates. All statistics were computed using SYSTAT 10 software (SPSS, Chicago, Illinois, United States).
Supporting Information
Figure S1 Phylogenetic Analysis of Source Inoculum V1V2 Variants with Molecular Clones
(A) NJ tree showing the most highly supported clade of SI used as the outgroup for all subsequent phylogenetic trees. (B) An unrooted ML tree of 30 SI V1V2 variants and corresponding V1V2 sequences from clones SIVmac239, SIVsmmH4, and SIVmne (obtained from the HIV sequence database [http://hiv-web.lanl.gov/content/index]) was constructed with Treefinder [46] using a GTR model and estimated gamma rate distribution, base frequencies, and substitution rates. Bootstrap values greater than 50% are shown at nodes.
(53 KB PDF)
Click here for additional data file.
Figure S2 Viral Replication Dynamics following Infection with a Diverse SIVsmm
Three SMs and three RMs were inoculated with plasma obtained from a naturally infected SM. Viral replication was monitored in SMs and RMs by quantitative RT-PCR of plasma RNA (see Materials and Methods).
(210 KB PDF)
Click here for additional data file.
Figure S3 Amino Acid Sequences of All Animals and All Time Points in the V1V2 Region of Envelope
Region corresponds to nucleotides 6,801–7,220 of SIVsmmH4. Sequences were aligned using the program CLUSTAL X [41], followed by manual adjustment using MacClade 4.0 [42]. A nonaligned region of length variation in V1 was removed, corresponding to amino acids 129–137 of SIVsmmH4 env, and is indicated by “~”. The consensus of all sequences in this study is shown above all sample sets, with codon positions labeled above. A dot indicates amino acid identity with the consensus sequence, and any amino acid changes are indicated with the appropriate symbol. The V1V2 regions are highlighted in blue on the consensus sequence, and glycosylation consensus motifs present in each sequence are highlighted in yellow.
(226 KB DOC)
Click here for additional data file.
Figure S4 Mean Number of Glycosylation Consensus Motifs in SMs and RMs for All Time Points
Frequency of glycosylation consensus motifs is lower in RMs (regression analysis, p < 0.001) and increases over time in both SMs and RMs. The number of motifs in the SI is indicated with a star on the y-axis.
(114 KB PDF)
Click here for additional data file.
Figure S5 Phylogenetic Analysis of Natural Host V1V2 Variants at Days 10–100 Shows No Specific Pattern
ML phylogenetic tree of sequences obtained from SM1 at days 10 (pink) to 100 (red) is shown, constructed with Treefinder [46] using a GTR model and estimated gamma rate distribution, base frequencies, and substitution rates. Bootstrap values greater than 50% are shown at nodes, and the number of multiple clones from the same animal at the ends of branches is indicated within the symbol. The SI variants are represented by triangles and identified by the label within.
(77 KB PDF)
Click here for additional data file.
Figure S6 Amino Acid Sequences of All Animals and All Time Points in the p27 Region of gag
Region corresponds to nucleotides 1,516–1,936 of SIVsmmH4. Sequences were aligned using the program CLUSTAL X [41], followed by manual adjustment using MacClade 4.0 [42]. The top sequence in each set corresponds to the majority consensus sequence from all sequences at all time points, with codon positions labeled above. A dot indicates amino acid identity with the consensus sequence, and any amino acid changes are indicated with the appropriate symbol.
(75 KB DOC)
Click here for additional data file.
Accession Number
The GenBank (http://www.ncbi.nlm.nih.gov/) accession number of the SIVsmmH4 genome is X14307.
We dedicate this paper to the memory of Dr. H. McClure, for his selfless devotion to advancing AIDS research in the nonhuman primate models, and for his genuine and warm collegiality. We thank Drs. F. Novembre and S. Garg for technical assistance, and Drs. E. Hunter and C. Derdeyn for valuable comments. This work was supported by National Institutes of Health grants AI4915502 and AI4476301 to MBF, RR00165 to the Yerkes National Primate Research Center, and National Institute on Allergy and Infectious Diseases Statistical Training on AIDS Grant T32-AI07442.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. LJD, MBF, and SIS conceived and designed the experiments. LJD performed the experiments. LJD and THV analyzed the data. JML contributed reagents/materials/analysis tools. LJD, THV, and SIS wrote the paper.
Abbreviations
CCR5CC-chemokine receptor 5
GTRgeneral time reversible
IVintravenous
NJneighbor-joining
MLmaximum likelihood
p.i.postinfection
RMrhesus macaque
SIsource inoculum
SIVsimian immunodeficiency virus
SMsooty mangabey
V1SIV envelope variable region 1
V2SIV envelope variable region 2
==== Refs
References
Salemi M De Oliveira T Courgnaud V Moulton V Holland B 2003 Mosaic genomes of the six major primate lentivirus lineages revealed by phylogenetic analyses J Virol 77 7202 7213 12805419
Hahn BH Shaw GM Cock KMD Sharp PM 2000 AIDS as a zoonosis: Scientific and public health implications Science 287 607 614 10649986
Silvestri G Sodora DL Koup RA Paiardini M O'Neil SP 2003 Non-pathogenic simian immunodeficiency virus infection of sooty mangabey mokeys is characterized by limited bystander immunopathology despite chronic high-level viremia Immunity 18 441 452 12648460
Rey-Cuille M-A Berthier J-L Bomsel-Demontoy M-C Chaduc Y Montagnier L 1998 Simian immunodeficiency virus replicates to high levels in sooty mangabeys without inducing disease J Virol 72 3872 3886 9557672
Goldstein S Ourmanov I Brown CR Beer BE Elkins WR 2000 Wide range of viral load in healthy African green monkeys naturally infected with simian immunodeficiency virus J Virol 74 11744 11753 11090174
Broussard SR Staprans SI White R Whitehead EM Feinberg MB 2001 Simian immunodeficiency virus replicates to high levels in naturally infected African green monkeys without inducing immunologic or neurologic disease J Virol 75 2262 2275 11160730
Korber BTM Muldoon M Theiler J Gao F Gupta R 2000 Timing the ancestor of the HIV-1 pandemic strains Science 288 1789 1796 10846155
Puffer BA Altamura LA Pierson TC Doms RW 2004 Determinants within gp120 and gp41 contribute to CD4 independence of SIV Envs Virology 327 16 25 15327894
Zhang M Gaschen B Blay W Foley B Haigwood N 2004 Tracking global patterns of N-linked glycosylation site variation in highly variable viral glycoproteins: HIV, SIV, and HCV envelopes and influenza hemagglutinin Glycobiology 14 1229 1246 15175256
Reiter JN Means RE Desrosiers RC 1998 A role for carbohydrates in immune evasion in AIDS Nat Med 4 679 684 9623976
Richman DD Wrin T Little SJ Petropoulos CJ 2003 Rapid evolution of the neutralizing antibody response to HIV type 1 infection Proc Natl Acad Sci U S A 100 4144 4149 12644702
Wei X Decker JM Wang S Hui H Kappes JC 2003 Antibody neutralization and escape by HIV-1 Nature 422 307 311 12646921
van't Wout AB, Koostra NA, Mulder-Kampinga GA, Albrecht-van Lent N, Scherpbier HJ, et al. 1994 Macrophage tropic variants initiate human immunodeficiency virus type 1 infection after sexual, parenteral, and vertical transmission J Clin Invest 94 2060 2067 7962552
Zhu T Wang N Carr A Nam DS Moor-Jankowski R 1996 Genetic characterization of human immunodeficiency virus type 1 in blood and genital secretions: Evidence for viral compartmentalization and selection during sexual transmission J Virol 70 3098 3107 8627789
Zhang LQ MacKenzie P Cleland A Holmes EC Brown AJ 1993 Selection for specific sequences in the external envelope protein of human immunodeficiency virus type 1 upon primary infection J Virol 67 3345 3356 8497055
Derdeyn CA Decker JM Bibollet-Ruche F Mokili JL Muldoon M 2004 Envelope-constrained neutralization-sensitive HIV-1 after heterosexual transmission Science 303 2019 2022 15044802
Chohan B Lang D Sagar M Korber B Lavreys L 2005 Selection for human immunodeficiency virus type 1 envelope glycosylation variants with shorter V1-V2 loop sequences occurs during transmission of certain genetic subtypes and may impact viral RNA levels J Virol 79 6528 6531 15858037
Cohen J 2004 Virology. HIV may shed some protection as it jumps to new hosts Science 303 1956 15044774
Sagar M Kirkegaard E Long EM Celum C Buchbinder S 2004 Human immunodeficiency virus type 1 (HIV-1) diversity at time of infection is not restricted to certain risk groups or specific HIV-1 subtypes J Virol 78 7279 7283 15194805
Beer BE Bailes E Goeken R Dapolito G Coulibaly C 1999 Simian immunodeficiency virus (SIV) from sun-tailed monkeys (Ceropithecus solatus): Evidence for host-dependent evolution of SIV within the C. l'hoesti superspecies J Virol 73 7734 7744 10438863
Johnson PR Fomsgaard A Allan JS Gravell M London WT 1990 Simian immunodeficiency viruses from African green monkeys display unusual genetic diversity J Virol 64 1086 1092 2304139
Peeters M Courgnaud V Abela B Auzel P Pourrut X 2002 Risk to human health from a plethora of simian immunodeficiency viruses in primate bushmeat Emerg Infect Dis 8 451 457 11996677
Gao F Bailes E Robertson DL Chen Y Rodenburg CM 1999 Origin of HIV-1 in the chimpanzee Pan troglodytes troglodytes
Nature 397 436 441 9989410
Silvestri G Fedanov A Germon S Kozyr N Kaiser W 2004 Divergent host responses during primary SIVsmm infection of natural mangabey and non-natural rhesus macaque hosts J Virol 79 4043 4054
Puffer BA Pohlmann S Edinger AL Carlin D Sanchez MD 2002 CD4 independence of simian immunodeficiency virus envs is associated with macrophage tropism, neutralization sensitivity, and attenuated pathogenicity J Virol 76 2595 2605 11861825
Rudensey LM Kimata JT Long EM Chackerian B Overbaugh J 1998 Changes in the extracellular envelope glycoprotein of variants that evolve during the course of simian immunodeficiency virus SIVMne infection affect neutralizing antibody recognition, syncytium formation, and macrophage tropism but not replication, cytopathicity, or CCR-5 coreceptor recognition J Virol 72 209 217 9420217
Burns DPW Desrosiers RC 1994 Envelope sequence variation, neutralizing antibodies, and primate lentivirus persistence Curr Top Microbiol 188 185 219
Petry H Pekrun K Hunsmann G Jurkiewicz E Luke W 2000 Naturally ocurring V1-env region variants mediate simian immunodeficiency virus SIVmac escape from high-titer neutralizing antibodies induced by a protective subunit vaccine J Virol 74 11145 11152 11070011
Jurkiewicz E Hunsmann G Schnaffner J Nisslein T Luke W 1997 Identication of the V1 region as a linear neutralizing epitope of the simian immunodeficiency virus SIVmac envelope glycoprotein J Virol 71 9475 9481 9371609
Chackerian B Rudensey LM Overbaugh J 1997 Specific N-linked and O-linked glycosylation modifications in the envelope V1 domain of simian immunodeficiency virus variants that evolve in the host alter recognition by neutralizing antibodies J Virol 71 7719 7727 9311856
Fomsgaard A Hirsch VM Johnson PR 1992 Cloning and sequences of primate CD4 molecules: Diversity of the cellular receptor for simian immunodeficiency virus/human immunodeficiency virus Eur J Immunol 22 2973 2981 1425921
Kunstman KJ Puffer B Korber BT Kuiken C Smith UR 2003 Structure and function of CC-chemokine receptor 5 homologues derived from representative primate species and subspecies of the taxonomic suborders Prosimii and Anthropoidea J Virol 77 12310 12318 14581567
Staprans SI Barry AP Silvestri G Safrit JT Kozyr N 2004 Enhanced SIV replication and accelerated progression to AIDS in macaques primed to mount a CD4 T cell response to the SIV envelope protein Proc Natl Acad Sci U S A 101 13026 13031 15326293
Garber DA Silvestri G Barry AP Fedanov A Kozyr N 2003 Blockade of T cell costimulation reveals interrelated actions of CD4+ and CD8+ T cells in control of SIV replication J Clin Invest 113 836 845
Ebert LM McColl SR 2002 Up-regulation of CCR5 and CCR6 on distinct subpopulations of antigen-activated CD4+ T lymphocytes J Immunol 168 65 72 11751947
Paillard F Sterckers G Vaquero C 1990 Transcriptional and post-transcriptional regulation of TcR, CD4 and CD8 gene expression during activation of normal human T lymphocytes EMBO J 9 1867 1872 2140772
Siliciano JD Siliciano RF 2004 A long-term latent reservoir for HIV-1: Discovery and clinical implications J Antimicrob Chemother 54 6 9 15163657
Albrecht D Zollner B Feucht HH Lorenzen T Laufs R 2002 Reappearance of HIV multidrug-resistance in plasma and circulating lymphocytes after reintroduction of antiretroviral therapy J Clin Virol 24 93 98 11744433
Ruiz-Jarabo CM Arias A Baranowski E Escarmis C Domingo E 2000 Memory in viral quasispecies J Virol 74 3543 3547 10729128
National Institutes of Health 1985 Guide for the care and use of laboratory animals, rev. ed. Department of Health and Human Services publication 85–23 Bethesda, Maryland National Institutes of Health
Thompson JD Higgins DG Gibson TJ 1994 CLUSTAL W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acid Res 22 4673 4680 7984417
Maddison WP Maddison DR 1989 Interactive analysis of phylogeny and character evolution using the computer program MacClade Folia Primatol 53 190 202 2606395
McAllister J Casino C Davidson F Power J Lawlor E 1998 Long-term evolution of the hypervariable region of hepatitis C in a common-source-infected cohort J Virol 72 4893 4905 9573256
Swofford DL 2002 PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). 4.0 ed Sunderland, Massachusetts Sinauer, Associates
Posada D Crandall KA 1998 Modeltest: Testing the model of DNA substitution Bioinformatics 14 817 818 9918953
Jobb G 2002 Treefinder [computer program] Available at: http://www.treefinder.de . Accessed 01 January 2003.
Gu X Li W-H 1996 A general additive distance with time-reversibility and rate variation among nucleotide sites Proc Natl Acad Sci U S A 93 4671 4676 8643462
Gu X Li W-H 1998 Estimation of evolutionary distances under stationary and nonstationary models of nucleotide substitution Proc Natl Acad Sci U S A 95 5899 5905 9600890
Yang Z 1994 Estimating the pattern of nucleotide substitution J Mol Evol 39 105 111 8064867
Nei M Gojobori T 1986 Simple methods for estimating the numbers of synonymous and nonsynonymous nucleotide substitutions Mol Biol Evol 3 418 426 3444411
Ota T Nei M 1994 Variance and covariances of the numbers of synonymous and nonsynonymous substitutions per site Mol Biol Evol 11 613 619 8078400
Tamura K Nei M 1993 Estimation of the number of nucleotide substitutions in the control region of mitochondrial DNA in humans and chimpanzees Mol Biol Evol 10 512 526 8336541
Kumar S Tamura K Jakobsen I Nei M 2001 MEGA2: Molecular evolutionary genetics analysis software Bioinformatics 17 1244 1245 11751241
|
16201015
|
PMC1238738
|
CC BY
|
2021-01-05 12:11:27
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e3
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010003
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 1620101610.1371/journal.ppat.001000405-PLPA-RA-0024R2plpa-01-01-06Research ArticleBiochemistryCell BiologyImmunologyInfectious DiseasesMicrobiologyVirologyVirusesIn VitroAnimalsHomo (human)Mus (mouse)Influenza Virus PB1-F2 Protein Induces Cell Death through Mitochondrial ANT3 and VDAC1 Influenza Virus PB1-F2 Induces ApoptosisZamarin Dmitriy 1García-Sastre Adolfo 1Xiao Xiaoyao 2Wang Rong 2Palese Peter 1*
1 Department of Microbiology, Mount Sinai School of Medicine, New York, New York, United States of America
2 Department of Human Genetics, Mount Sinai School of Medicine, New York, New York, United States of America
Virgin Skip EditorWashington University School of Medicine, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 1 e413 4 2005 23 6 2005 Copyright: © 2005 Zamarin et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.The influenza virus PB1-F2 is an 87-amino acid mitochondrial protein that previously has been shown to induce cell death, although the mechanism of apoptosis induction has remained unclear. In the process of characterizing its mechanism of action we found that the viral PB1-F2 protein sensitizes cells to apoptotic stimuli such as tumor necrosis factor alpha, as demonstrated by increased cleavage of caspase 3 substrates in PB1-F2-expressing cells. Moreover, treatment of purified mouse liver mitochondria with recombinant PB1-F2 protein resulted in cytochrome c release, loss of the mitochondrial membrane potential, and enhancement of tBid-induced mitochondrial permeabilization, suggesting a possible mechanism for the observed cellular sensitization to apoptosis. Using glutathione-S-transferase pulldowns with subsequent mass spectrometric analysis, we identified the mitochondrial interactors of the PB1-F2 protein and showed that the viral protein uniquely interacts with the inner mitochondrial membrane adenine nucleotide translocator 3 and the outer mitochondrial membrane voltage-dependent anion channel 1, both of which are implicated in the mitochondrial permeability transition during apoptosis. Consistent with this interaction, blockers of the permeability transition pore complex (PTPC) inhibited PB1-F2-induced mitochondrial permeabilization. Based on our findings, we propose a model whereby the proapoptotic PB1-F2 protein acts through the mitochondrial PTPC and may play a role in the down-regulation of the host immune response to infection.
Synopsis
PB1-F2 is a short polypeptide encoded by influenza viruses. While the role of this viral protein is not completely understood, it is known to localize in the mitochodria of the infected cell and to promote cell death. The authors found that PB1-F2 sensitizes cells to death through interactions with two mitochondrial proteins, ANT3 and VDAC1. These interactions promote the permeabilization of the mitochodria and facilitate the release of mitochondrial products that trigger cell death (apoptosis). PB1-F2-mediated cell death through the mitochondria is likely to contribute to the pathogenicity of the influenza virus.
Citation:Zamarin D, García-Sastre A, Xiao X, Wang R, Palese P (2005) Influenza virus PB1-F2 protein induces cell death through mitochondrial ANT3 and VDAC1. PLoS Pathog 1(1): e4.
==== Body
Introduction
Influenza virus infection results in the activation of cellular pathways aimed at inhibition of viral replication and induction of an antiviral state [1]. To overcome the antiviral signaling, influenza viruses evolved accessory proteins, such as NS1 and PB1-F2, that have been proposed to down-modulate different aspects of the host immune response [2,3]. While the NS1 protein has been shown to play a role in the inhibition of the type I interferon response, the function of the PB1-F2 protein remains elusive.
PB1-F2 is a novel 87-amino acid protein serendipitously identified in an alternate reading frame of the PB1 gene of the influenza A/PR/8/34 virus [3]. Initial studies revealed that the protein localizes to mitochondria resulting in the alteration of mitochondrial morphology, dissipation of mitochondrial membrane potential, and cell death, which was more pronounced in cells of immune origin [3]. The basic amphipathic helix in the C-terminal region of the PB1-F2 protein was subsequently determined to be responsible for its mitochondrial localization [4,5]. Synthetic peptides derived from the C-terminal domain of the protein were shown to have an ability to oligomerize and nonspecifically permeabilize lipid bilayer membranes [6,7], properties observed with some known cellular mitochondrial apoptotic mediators [8,9]. Despite these findings, however, the precise mechanism and function of PB1-F2-induced apoptosis remains unclear.
Regulation of the mitochondrial permeabilization has been implicated in the life cycle of several known human pathogens [10,11]. Indeed, stable cell lines overexpressing the antiapoptotic proteins of the Bcl-2 family are less permissive to influenza viral replication than their parental cell lines [12–14], highlighting the importance of the role of the mitochondrial apoptotic pathways in influenza virus pathogenesis.
Cellular apoptotic signaling to mitochondria proceeds through activation of the members of the proapoptotic Bcl-2 family BH3-only proteins such as Bid, which exert their effects through induction of the release of several mitochondrial apoptotic mediators, such as cytochrome c, apoptosis-inducing factor, endonuclease G, Smac/Diablo, and Omi/HtrA2 [15]. The exact mechanism leading to the mitochondrial permeabilization is still under investigation, but is known to involve cellular apoptotic mediators of the Bcl-2 family, such as Bak and Bax, and proteins constituting the permeability transition pore complex (PTPC), such as the adenine nucleotide translocator 3 (ANT3) in the inner mitochondrial membrane and the voltage-dependent anion channel 1 (VDAC1) in the outer mitochondrial membrane [16–19].
In view of the possible contribution of the PB1-F2 protein to influenza viral pathogenesis, we sought to determine the role of the protein in modulation of host immune response and to further elucidate its mechanism of action. We show that the mitochondrial permeabilization by the PB1-F2 protein renders cells sensitive to the proapoptotic effect of tumor necrosis factor alpha (TNFα) through tBid signaling. Furthermore, our results indicate that PB1-F2-induced apoptosis proceeds through a unique mechanism involving its interaction with the ANT3 and VDAC1 proteins of the PTPC at the inner and outer mitochondrial membranes, respectively. With the use of ANT3-specific inhibitor, we conclude that PB1-F2 directly permeabilizes mitochondria in an ANT3-dependent manner. The results of our studies provide a deeper insight into the function of the PB1-F2 protein and underscore the role of the PTPC in mitochondrial permeabilization and cell death in influenza virus-infected cells.
Results
PB1-F2 Protein Sensitizes Transfected Cells to Apoptotic Stimuli
We investigated whether transient expression of the influenza virus PB1-F2 protein would enhance cell death by intrinsic and extrinsic proapoptotic stimuli (Figure 1). To achieve maximal transient transfection efficiency of the PB1-F2 protein, the proapoptotic effect of the agents was initially assayed in PB1-F2-transfected 293T cells, where apoptosis was detected by cleavage of poly A ribose polymerase (PARP), a direct substrate of caspase 3 (Figure 1A). The unrelated influenza virus protein, nucleoprotein (NP), which does not induce apoptosis, was used as a control. Reduction of full-length PARP with a concomitant increase in cleaved PARP product was seen in PB1-F2-transfected cells in response to DNA damage caused by UV irradiation and cisplatin, when compared to the control (Figure 1A). Treatment of PB1-F2-transfected cells with 50 ng/ml of TNFα or with 5 ng/ml of TNF-related apoptosis-inducing ligand resulted in significant increase in PARP cleavage, as compared to the control (Figure 1A). In view of the findings that PB1-F2 sensitized cells to both extrinsic and intrinsic apoptotic stimuli, we tested whether the PB1-F2-expressing cells are also sensitized to death due to detachment from the growth matrix (anoikis). The activation of anoikis has also been shown to involve proapoptotic members of the BH3 family [20,21]. Indeed, PB1-F2 also sensitized cells to death by anoikis, as demonstrated by rapid membrane blebbing and fragmentation of trypsinized human lung epithelial A549 cells transfected with a construct expressing PB1-F2 (Figure 1B) [22]. To determine whether the PB1-F2-expressing cells were indeed undergoing apoptosis, we labeled PB1-F2-transfected A549 cells with M30 antibody to caspase-cleaved cytokeratin. Interestingly, in the absence of other apoptotic stimuli, only a small proportion of PB1-F2-expressing cells underwent apoptosis (Figure 1C), which could explain the relatively low proapoptotic effect of the PB1-F2 protein observed in the assays above. Overall, these results suggest that the PB1-F2 protein enhances the effect of cellular proapoptotic stimuli.
Figure 1 PB1-F2 Protein Sensitizes Cells to Apoptosis in Response to Cytotoxic Stimuli
(A) Expression of PB1-F2 protein enhanced cleavage of poly-A ribose polymerase (PARP) in response to cytotoxic agents. 293T cells were transfected with either PB1-F2 (F2) or influenza virus NP (NP) for 24 h and were subsequently treated for 6 h with 50 ng/ml TNFα, 5 ng/ml TNF-related apoptosis-inducing ligand, 100 μM cisplatin, or UV-irradiation at 60 J/m2 as indicated. All cells were collected 6 h later, lysed, and processed by immunoblotting for cleaved PARP.
(B) PB1-F2 protein predisposes cells to death by anoikis. A549 cells were transfected either with empty vector or with PB1-F2 and trypsinized 24 h post-transfection. Cells were incubated in PBS with 0.3% BSA for 15 min, pipetted onto microscope slides, and changes in cell morphology (membrane blebbing indicated by arrows) were visualized by phase-contrast microscopy [22].
(C) A subset of PB1-F2-expressing cells undergoes apoptosis. A549 cells were transfected with HA-tagged PB1-F2 for 24 h and stained for HA epitope and cleaved cytokeratin (M30).
Sensitization to Apoptosis by the PB1-F2 Protein Is Inhibited by Bcl-xL
Since PB1-F2 has been shown to localize to mitochondria [3,4], we investigated whether mitochondrial apoptotic mechanisms are involved in PB1-F2-induced apoptotic enhancement. Mitochondrial permeabilization by cellular apoptotic mediators is controlled by the antiapoptotic proteins such as Bcl-xL and Bcl-2. Given the dependence of PB1-F2-induced apoptosis on cellular apoptotic factors, we decided to determine whether PB1-F2-induced apoptotic sensitization can overcome the cytoprotective effect of Bcl-xL. For the purposes of our experiments, we used A549 cells to generate a cell line stably overexpressing Bcl-xL protein (A549-Bcl-xL) and a respective control cell line expressing the gene encoding neomycin resistance. The resulting cell lines both exhibited transient transfection efficiencies of around 60% as witnessed by transient expression of green fluorescent protein (GFP; unpublished data).
In a control cell line stably transfected with neomycin vector (A549-neo), expression of PB1-F2 enhanced the proapoptotic effect of TNFα, as demonstrated by immunolabeling with M30 antibody specific for cleaved cytokeratin 18 (Figure 2A). Coimmunostaining for PB1-F2 revealed that the PB1-F2-expressing cells were indeed undergoing apoptosis. However, PB1-F2-induced apoptosis in response to TNFα was inhibited in the A549 cell line stably overexpressing Bcl-xL (Figure 2B). Overall these results suggest that the mitochondrial apoptotic pathways are required for PB1-F2-mediated apoptotic enhancement.
Figure 2 PB1-F2 Protein-Mediated Enhancement of TNFα-Induced Apoptosis Is Inhibited by Bcl-xL
A549 cells containing a stably-integrated neomycin resistance gene (A549-neo) (A), or A549 cells stably overexpressing Bcl-xL (A549-Bcl-xL) (B) were transfected with either empty vector or vector encoding HA-tagged PB1-F2. At 24 h post-transfection, cells were treated with 50 ng/ml TNFα, where indicated. Then 6 h post-treatment, the cells were fixed and stained with M30 antibody to cleaved cytokeratin and with anti-HA antibody.
The Influenza Virus PB1-F2 Protein Disrupts Mitochondrial Organization and Induces the Release of Cytochrome C
To further characterize PB1-F2-mediated apoptotic enhancement, we generated a monoclonal antibody (26D3) specific for the N-terminal domain of the protein, and confirmed by microscopy that the PB1-F2 protein expressed from transfected plasmid localized to mitochondria in A549-neo and A549-Bcl-xL cells (Figure 3A). While in normal cells mitochondria are usually distributed along the microtubular networks, mitochondria of PB1-F2-transfected cells lost the normal tubuloreticular organization and displayed a punctiform appearance (Figure 3A). Interestingly, this effect was not inhibited by Bcl-xL overexpression (Figure 3A), suggesting that disruption of the mitochondrial network is not the primary mechanism for PB1-F2-induced apoptosis.
Figure 3 Influenza Virus PB1-F2 Protein Disrupts Mitochondria
(A) PB1-F2 disrupts reticulotubular mitochondrial organization. A549-neo and A549-Bcl-xL cells were transfected with HA-tagged PB1-F2 for 24 h and stained with antibody against HA tag and with human anti-mitochondrial serum as a marker for mitochondria. The cells with punctiform mitochondria are indicated by (*), while the cells with normal reticulotubular mitochondrial organization are indicated by (#).
(B) PB1-F2 protein induces release of mitochondrial cytochrome c in transfected cells. HeLa cells were transfected with HA-tagged PB1-F2 for 24 h, treated with 50 ng/ml TNFα for 8 h, and stained with anti-HA antibody (green, secondary antibody FITC), anti-cytochrome c antibody (red, secondary antibody Alexa 568), and DAPI (blue). The cells were visualized by confocal microscopy.
(C) Subcellular fractionation of 293T cells transfected with PB1-F2. 293T cells were transfected for 24 h and subsequently were either mock-treated or treated with 50 ng/ml TNFα. Cells were subsequently collected and fractionated into the cytosolic and mitochondrial fractions. The amount of cytochrome c remaining in the mitochondrial fraction was detected by Western blot. Tom20 protein was used as a loading control.
To further evaluate the effect of PB1-F2 protein on the mitochondrial apoptotic pathways, we analyzed the PB1-F2-expressing cells for release of cytochrome c from the mitochondria. HeLa cells were transfected with hemagglutinin (HA)-tagged PB1-F2, treated with 50 ng/ml TNFα for 8 h, and stained for HA tag and cytochrome c. Expression of the PB1-F2 protein resulted in the release of cytochrome c from mitochondria (Figure 3B) in a subset of cells, as witnessed by diffusion of cytochrome c in the cytoplasm. For a better quantitative measure of the amount of cytochrome c released, 293T cells were transfected with a plasmid expressing PB1-F2 or an empty vector and 24 h later were treated for 8 h with 50 ng/ml TNFα. The cells were subsequently fractionated into cytosolic and mitochondrial fractions, and the amount of cytochrome c remaining in the mitochondrial fraction was determined by Western blot (Figure 3C). Treatment of the PB1-F2 transfected cells with TNFα resulted in enhanced release of cytochrome c from the mitochondria when compared to the vector control.
PB1-F2 Directly Induces Mitochondrial Permeabilization and Sensitizes Mitochondria to the Proapoptotic Effect of tBid
To further investigate whether the effect of PB1-F2 protein on the mitochondria is direct, we incubated purified mouse liver mitochondria with recombinant PB1-F2 protein. Recombinant tBid protein was used as a control. Treatment of mitochondria with the PB1-F2 protein resulted in the release of cytochrome c, which increased at higher concentrations, suggesting a direct role for PB1-F2 in promoting mitochondrial outer membrane permeabilization (Figure 4A). This effect was not observed when a recombinant glutathione-S-transferase (GST) protein was used in this assay at similar concentrations (unpublished data). In view of the previous report that the C-terminal region of PB1-F2 protein targets it to the inner mitochondrial membrane, we proceeded to investigate whether PB1-F2-induced mitochondrial permeabilization also involves the inner mitochondrial membrane. For this purpose we used the mitochondrial potential-sensitive dye JC-1 to determine whether incubation of purified mitochondria with recombinant PB1-F2 protein would result in dissipation of the mitochondrial inner membrane potential. The uptake of the JC-1 dye by intact mitochondria results in increased red-orange fluorescence at 590 nm upon excitation at 490 nm. As can be seen from Figure 4B, incubation of the purified mitochondria for 10 min with increasing doses of PB1-F2 resulted in reduction of JC-1 fluorescence at 590 nm, suggesting that the PB1-F2 protein-induced mitochondrial permeabilization involves the inner mitochondrial membrane. Interestingly, while recombinant PB1-F2 protein was, at equimolar concentrations, less effective than tBid in releasing cytochrome c (Figure 4A), it was comparable to tBid in permeabilization of the inner mitochondrial membrane. This suggests that while loss in the mitochondrial membrane potential might not be required for tBid-induced cytochrome c release [23], it may be necessary for PB1-F2-induced mitochondrial permeabilization. These results confirmed previous findings that showed that expression of PB1-F2 protein in cells can lead to dissipation of the mitochondrial membrane potential [3]. Interestingly, as we and others showed, expression of the PB1-F2 protein itself does not cause significant levels of apoptosis [3]. This suggests that the inner membrane permeabilization by the PB1-F2 protein might not be the primary mechanism of apoptosis induction. Rather, we hypothesize that PB1-F2-induced inner membrane permeabilization acting in conjunction with another apoptotic mechanism may be responsible for PB1-F2-mediated apoptotic enhancement illustrated in Figures 1 and 2.
Figure 4 Influenza Virus PB1-F2 Protein Directly Permeabilizes Mitochondria
(A) Recombinant PB1-F2 protein caused release of cytochrome c from purified mouse mitochondria. Mouse mitochondria (50 μg total protein) were incubated for 1 h at 30 °C with recombinant protein of interest at indicated concentrations. The amount of cytochrome c released in the supernatant was assayed by immunoblotting using anti-cytochrome c antibody. Recombinant tBid protein was used as a control. The mitochondrial pellet was assayed for the Tom20 protein to ensure an equal amount of mitochondria was used.
(B) PB1-F2 and tBid cause dissipation of the mitochondrial membrane potential. Mouse mitochondria (50 μg) were incubated for 10 min at 30 °C with the protein of interest at indicated concentrations. Relative loss in membrane potential was measured by uptake of membrane potential-sensitive JC-1 dye. Data is presented as percent fluorescence relative to mock-treated mitochondria.
(C) PB1-F2 protein enhances tBid-induced cytochrome c release. Purified mouse mitochondria (50 μg) were incubated with 10 nM PB1-F2 or mock-treated for 15 min in 20 μl and were subsequently treated with recombinant tBid in the indicated concentrations for 1 h in a total volume of 30 μl. The supernatants were assayed for cytochrome c release by Western blot. The mitochondrial pellet was processed for Tom20 to ensure equal amount of mitochondria used.
In view of the fact that both intrinsic and extrinsic cellular apoptotic stimuli converge on the mitochondria through signaling of the Bcl-2 family BH3 proteins, we investigated whether the presence of PB1-F2 in suboptimal concentrations would enhance the release of cytochrome c from purified mitochondria by recombinant tBid. Incubation of the purified mitochondria with recombinant tBid results in release of cytochrome c at tBid concentrations as low as 10 nM (Figure 4C). Preincubation of the purified mitochondria with 10 nM PB1-F2 augments the tBid-induced cytochrome c release by a factor of almost ten (Figure 4C). As the PB1-F2 protein was used at a concentration at which it fails to release cytochrome c by itself, the result suggests that the effect of PB1-F2 and tBid is not simply additive. We speculate that PB1-F2-induced permeabilization of the inner mitochondrial membrane may augment cytochrome c release in response to tBid. Overall, these results suggest that sensitization of cells to apoptosis by the PB1-F2 may proceed through its potentiation of the effects of the cellular BH3 family proteins.
Previous studies revealed that the PB1-F2 protein localizes to both inner and outer mitochondrial membranes [3–5]. As the cellular apoptotic proteins were previously shown to act through the mitochondrial proteins controlling permeabilization of the inner as well as outer mitochondrial membranes, we speculate that the observed PB1-F2-induced apoptotic sensitization could proceed via components in both membranes. To investigate possible involvement of mitochondrial proteins in PB1-F2-induced apoptosis, we sought to identify potential mitochondrial interactors of the PB1-F2 protein.
The PB1-F2 Protein Interacts with ANT3 and VDAC1 Proteins of the Mitochondrial PTPC
PB1-F2 protein was expressed in 293T cells as an N-terminal GST fusion protein, and the complexes between GST-PB1-F2 and its interacting proteins were pulled down with glutathione-Sepharose beads. Proteins were subsequently eluted from the beads, separated by SDS gel electrophoresis, and visualized by silver stain. An unrelated polypeptide of similar length fused to GST (GST-Nipah-Wc, containing the last 43 amino acids of the W protein of the Nipah virus fused to GST) was used as a control. Unique protein bands at approximately 36, 55, and 80 kDa specific for the PB1-F2 fusion protein but not GST or the control protein were detected (Figure 5A, lanes 4 and 5). Mass spectrometry analysis of the proteins identified the 36-kDa band as the mitochondrial ANT3, the 55-kDa band as beta tubulin, and the 80-kDa band as cytokeratin. The latter was likely a contaminant of the protein preparation. While interaction of PB1-F2 with tubulin may be important for PB1-F2-induced mitochondrial disorganization (see Figure 3A), we chose to focus on the ANT3 protein, which is known to be involved in mitochondrial apoptosis. Coimmunoprecipitation experiments with transfected HA-tagged ANT3 confirmed the interaction between PB1-F2 and ANT3 (Figure 5B). As expected, transfected HA-tagged ANT3 localized to the mitochodria (Figure 5C).
Figure 5 PB1-F2 Protein Interacts with ANT3 and VDAC1
(A) Cellular proteins pulled down with GST-PB1-F2 were separated by 12% SDS-PAGE and silver-stained. The asterisks mark the protein bands that are unique to the GST-PB1-F2 lanes (~36 kDa, 55 kDa, and 80 kDa). Lanes 4 and 5 represent the results of two separate pulldown experiments.
(B) PB1-F2 specifically interacts with ANT3 and VDAC1 but not other outer and inner mitochondrial membrane proteins. Interaction of PB1-F2 with ANT3 and VDAC1 was confirmed in 293T cells by coimmunoprecipitation with transfected HA-tagged ANT3 and flag-tagged VDAC1 (top five images). Tom20, COXIV, and TIM 44 proteins were used as the mitochondrial coimmunoprecipitation controls.
(C) Transfected ANT3 and VDAC1 localize to mitochondria. 293T cells were transfected with Flag-tagged ANT3 and VDAC1 for 24 h and were subsequently fractionated to generate the cytosolic and mitochondrial fractions. Tom20 protein served as a control for the mitochondrial fraction. Each sample was processed in triplicate (lanes 1, 2, and 3, ANT3; lanes 4, 5, and 6, VDAC1).
(D) PB1-F2 directly interacts with ANT3 and VDAC1. 35S-labeled ANT3 and VDAC1 were expressed in vitro using a rabbit reticulocyte lysate system and subjected to pulldown with either 5 μg GST (left lane) or GST-PB1-F2 (right lane). Proteins were separated by 12% SDS-PAGE and visualized by autoradiography.
(E) PB1-F2 expressed during viral infection interacts with ANT3 and VDAC1. 293T cells were transfected with Flag-tagged ANT3 and VDAC1 proteins and 24 h later were infected with wild-type PR8 virus (PR8) or the virus knocked out for PB1-F2 expression (del) at an MOI of 2. 15 h after infection, the cells were collected and the immunoprecipitation with anti-PB1-F2 polyclonal serum was performed. Influenza virus NP was used as a control to show equal levels of infection between the samples. WB, Western blot.
To verify that the interaction between PB1-F2 and ANT3 was direct, we expressed 35S-labeled ANT3 in vitro and performed GST pulldowns on the labeled protein with either GST or GST-PB1-F2 proteins. Only GST-PB1-F2 was able to coprecipitate the in vitro-translated ANT3 protein, suggesting that direct interaction between PB1-F2 and ANT3 is likely (Figure 5D).
Previous studies revealed that the predicted basic amphipathic helix formed by amino acids 69–82 is sufficient for targeting PB1-F2 to the inner mitochondrial membrane [4]. However, immunoelectron microscopy studies had also previously shown that the full-length PB1-F2 protein localized to both inner and outer mitochondrial membranes [3], suggesting that the N terminus of the protein may play a role in targeting the protein to the outer mitochondrial membrane. Thus, while ANT3 was the only component of the PTPC identified by mass spectrometry to interact with PB1-F2, we could not exclude the possibility that PB1-F2 exerts its effect in conjunction with other proteins of the pore complex. Since ANT3 is exclusively an inner membrane protein, we speculated that the outer membrane PB1-F2 may be acting through an additional mechanism. Indeed, immunoprecipitation experiments revealed that PB1-F2 protein also interacted with VDAC1 (Figure 5B), and GST pulldown experiments with in vitro-translated VDAC1 confirmed that the interaction is likely direct (Figure 5D). These findings were surprising, since VDAC1 protein was not identified in the original experiment by mass spectrometry. While it is likely that the VDAC1 protein was missed in our analysis, it is also possible that the N-terminal GST tag may have prevented the interaction of the PB1-F2 protein with VDAC1 within the cell.
ANT3 and VDAC1 appear to be critical components of the pore complex [19]. ANT3 is an inner mitochondrial membrane protein that functions as an antiporter catalyzing the exchange of ATP for ADP [19,24]. In the presence of apoptotic stimuli, ANT3 is believed to undergo major conformational changes resulting in the formation of nonspecific pores in the inner mitochondrial membrane. Apoptotic stimuli also trigger conformational changes in the outer membrane protein VDAC1, which forms pores in the outer mitochondrial membrane. ANT3 and VDAC1 are believed to interact, forming the PTPC, which leads to dissipation of the inner mitochondrial membrane potential and the release of apoptotic mediators from the mitochondrial intermembrane space [24–26].
To eliminate the possibility of nonspecific interaction of PB1-F2 with mitochondrial proteins, we performed PB1-F2 coimmunoprecipitation experiments with the outer mitochondrial membrane transport protein Tom20, the inner mitochondrial membrane protein COXIV, and the protein Tim44, which localizes to the matrix side of the inner mitochondrial membrane (Figure 5B). PB1-F2 protein failed to coimmunoprecipitate with any of these proteins, further confirming the specificity of its interaction with ANT3 and VDAC1 (Figure 5B).
To confirm that the transfected tagged ANT3 and VDAC1 proteins were properly targeted to the mitochondria, we performed subcellular fractionation and determined that the majority of the expressed ANT3 and VDAC1 are indeed present within the mitochondrial fraction (Figure 5C). Furthermore, to determine whether PB1-F2 interacts with ANT3 and VDAC1 within the context of viral infection, we infected Flag-ANT3- and Flag-VDAC1-transfected 293T cells with either a wild-type PR8 virus or with its correspondent virus knocked out for PB1-F2 protein expression. Immunoprecipitation experiments with anti-PB1-F2 polyclonal serum revealed that the PB1-F2 protein interacts with ANT3 and VDAC1 during the viral infection (Figure 5E).
Due to lack of availability of good antibodies specific for ANT3 and VDAC1, we were unable to demonstrate the interaction of PB1-F2 with the endogenous proteins, apart from the interaction shown by mass spectrometry. Thus, to further confirm the specificity of the interaction, we proceeded to identify the interaction domains within the PB1-F2 protein.
The C-Terminal Domain of PB1-F2 Protein Is Responsible for the Interaction with ANT3, While Both C- and N-Terminal Domains Interact with VDAC1
Expression of HA-tagged N- or C-terminal domains of PB1-F2 was unsuccessful and resulted in fragments which appeared to be unstable (unpublished data). To stabilize each part of the protein, we generated HA-tagged GFP-fusion protein constructs of each domain (HA-nF2-GFP and HA-cF2-GFP; Figure 6A). The N-terminal region (HA-nF2-GFP) included amino acids 1–38, while the C-terminal fusion protein (HA-cF2-GFP) included amino acids 39–87 (Figure 6A). In transfected HeLa cells, full-length HA-PB1-F2-GFP fusion protein and HA-cF2-GFP protein localized to the mitochondria, while the HA-nF2-GFP fusion protein was diffusely distributed in the cytoplasm and the nucleus (Figure 6B). Since possible interaction of the N terminus of the protein with its cellular target could be influenced by its subcellular localization, we generated an additional construct fusing the N terminus to the mitochondrial targeting signal of the cytochrome oxidase, which targets the protein to the inner mitochondrial membrane (HA-MTS-nF2-GFP) [27]. The resultant fusion protein localized to mitochondria (Figure 6B). We next determined whether the fusion proteins interacted with Flag-tagged ANT3 and VDAC1 in transfected 293T cells. Both HA-PB1-F2-GFP and HA-cF2-GFP fusion proteins proved to interact with both VDAC1 and ANT3, while the HA-nF2-GFP interacted only with VDAC1 (Figure 6C and 6D). HA-MTS-nF2-GFP protein possessing the inner mitochondrial membrane targeting sequence still failed to interact with ANT3, confirming that the N terminus of the protein is not responsible for the interaction (Figure 6C). Interestingly, the interaction of HA-MTS-nF2-GFP protein with VDAC1 was also greatly reduced when compared to the HA-nF2-GFP protein lacking the MTS (Figure 6D). It is possible that forced localization of the nF2 to the inner mitochondrial membrane may have prevented its interaction with VDAC1 in the outer membrane.
Figure 6 Identification of PB1-F2 Protein Domains Responsible for Interaction with ANT3 and VDAC1
(A) PB1-F2 N- and C-terminal domains were cloned separately as N-terminal HA- and C-terminal GFP fusion proteins as follows. (1) GFP control; (2) full-length fusion protein (HA-PB1-F2-GFP); (3) C-terminal 38–87-amino acid domain fusion protein (HA-cF2-GFP); (4) N-terminal 1–37 amino acid domain fusion protein (HA-nF2-GFP); (5) N-terminal domain fusion protein with cytochrome oxidase mitochondrial targeting sequence (HA-MTS-nF2-GFP); and (6) control MTS-GFP fusion protein (HA-MTS-GFP).
(B) Localization of fusion proteins was determined in transfected HeLa cells (green, GFP-fusion protein; red, Mitotracker dye staining for mitochondria; blue, DAPI nuclear stain).
(C and D) Interaction of fusion proteins with ANT3 and VDAC1 was determined by cotransfecting 293T cells with each fusion construct and a vector encoding Flag-tagged ANT3 or VDAC1, respectively. Immunoprecipitation was performed with an anti-HA antibody, with subsequent immunoblotting for Flag-tagged ANT3 or VDAC1.
The PB1-F2 Protein-Mediated Mitochondrial Permeabilization Is Attenuated by ANT3 Blockers
We further investigated the role of the PTPC in PB1-F2-induced mitochondrial permeabilization. We turned to the known pore complex inhibitors: bongkrekic acid (BA), which was shown to inhibit ANT3 [28], and cyclosporine A (CsA), which binds to the mitochondrial matrix cyclophilin D and also inhibits ANT3. Purified mouse liver mitochondria were treated for 30 min with recombinant PB1-F2 protein in the presence or absence of BA (Figure 7A). Incubation of mitochondria with increasing doses of the PB1-F2 protein resulted in dissipation of the mitochondrial membrane potential as measured by JC-1 fluorescence. This effect was attenuated, although not completely inhibited, when the mitochondria were preincubated with 50 μM BA (Figure 7A).
Figure 7 PB1-F2 Protein Induces Mitochondrial Permeabilization in ANT3-Dependent Manner
(A) Recombinant PB1-F2 protein was incubated with 50 μg of purified mouse liver mitochondria in the presence or absence of 50 μM BA for 30 min. The mitochondria were further processed for assessment of membrane potential by JC-1 fluorescence at 590 nm.
(B) PB1-F2 induces loss of mitochondrial membrane potential in transfected cells. HeLa cells were transfected with GFP fusion constructs of PB1-F2, and 12 h later were treated with 50 ng/ml TNFα for 8 h, where indicated. Cells were subsequently stained with Mitotracker CMXRos Red dye. Cells with dissipated membrane potential are indicated by (*).
(C) The mitochondrial permeability transition inhibitor CsA inhibits PB1-F2-induced loss of mitochondrial membrane potential. HeLa cells in presence of CsA were transfected and treated as in (B) and stained with Mitotracker dye.
To confirm the involvement of ANT3 in PB1-F2-induced dissipation of the mitochondrial membrane potential in cells, and to identify the domain of the protein responsible for this permeabilization, we transfected HeLa cells with the PB1-F2-GFP fusion constructs described in Figure 6 and stained the cells with the mitochondrial membrane potential-sensitive Mitotracker CMXRos Red dye. GFP protein was used as a control. Consistent with previous reports [4,5], we found that in the absence of other stimuli, the C-terminal domain of the protein was more effective in dissipating the inner membrane potential than the full-length protein (Figure 7B, upper photomicrographs), as revealed by decreased Mitotracker Red staining in these cells. This is further consistent with our findings that the C-terminal domain of the PB1-F2 protein is responsible for its interaction with ANT3 (Figure 6). Treatment of the HeLa cells expressing the PB1-F2-GFP fusion construct with TNFα resulted in dissipation of the membrane potential, further confirming that the full-length protein requires additional apoptotic stimuli for its effect (Figure 7B). To confirm that the PB1-F2-induced permeability transition indeed proceeds in ANT3-dependent manner, we performed the same experiment in the presence of CsA, a known ANT3 inhibitor (Figure 7C). Treatment of HeLa cells with CsA resulted in preservation of the mitochondrial membrane potential in the cells expressing the cF2-GFP and in the TNFα-treated cells expressing the PB1-F2-GFP protein.
PB1-F2-Induced Apoptosis Proceeds in ANT3-Dependent Manner
To confirm that the PB1-F2-mediated sensitization to apoptosis proceeds in ANT3-dependent manner, we investigated whether BA, an ANT3 blocker, could inhibit this effect. A549 cells were transfected with PB1-F2 for 24 h and were subsequently treated with TNFα either in the presence or absence of 50 μM BA. As can be seen from Figure 8A, BA inhibited TNFα-induced apoptosis in these cells, as compared to the untreated control. Overall, these results, along with the interaction studies, suggest that the PTPC plays a role in PB1-F2-induced mitochondrial permeabilization.
Figure 8 PB1-F2 Induces Apoptosis Acting through Components of the PTPC
(A) ANT3 blocker BA inhibits PB1-F2-mediated sensitization of cells to apoptosis. A549 cells were transfected with PB1-F2 for 20 h and then treated with TNFα as indicated, either in the presence or absence of 50 μM BA. The cells were collected 8 h later and stained with M30 antibody to cleaved cytokeratin.
(B) Proposed models of PB1-F2 action during infection. During early stages of the infection, PB1-F2 localizes to mitochondria, where it interacts with ANT3 and VDAC and predisposes the mitochondria to permeability transition. Later in the infection, when more PB1-F2 is synthesized, and upon induction of antiviral apoptotic signaling pathways, the mitochondria undergo the permeability transition, which results in the induction of apoptosis. PB1-F2-induced mitochondrial permeabilization can proceed through three possible mechanisms, as indicated on the right graphic: (1) enhancement of the pore complex formation through direct interaction with ANT3 and VDAC1; (2) independent permeabilization of the inner and outer mitochondrial membranes with the help of ANT3 and VDAC1, respectively; and (3) direct permeabilization of the mitochondrial membranes.
Discussion
Mitochondrial control of apoptosis is a critical gateway for many cellular apoptotic pathways, whereby mitochondrial permeabilization and release of mediator proteins represent the point of no return in the execution of apoptotic cascades [29]. Permeabilization of mitochondria is thought to occur through two proposed mechanisms, proceeding in either permeability transition-dependent or -independent manner [16,17]. The mechanisms are not mutually exclusive, and it is generally believed that both play a role in the mitochondrial induction of apoptosis, whereby the contribution of each mechanism is dependent upon the apoptotic stimulus and/or perhaps differential tissue-specific regulation [17].
Several viral and bacterial proteins have been shown to induce or inhibit apoptosis through a direct effect on different mitochondrial components [10,11]. Viruses of the herpesvirus family evolved to regulate apoptosis by different mechanisms, with some members encoding Bcl-2 family homologs [30]. Porin proteins of Neisseria and the hepatitis B virus X protein interact with VDAC [31–33]. Cytomegalovirus vMIA protein and HIV Vpr proteins affect the permeability transition by targeting ANT3 [34,35], and the M11 protein of myxoma virus targets the peripheral benzodiazepine receptor, another component of the pore complex [36]. Still other viral proteins exert their effect on the mitochondria through yet unidentified mechanisms [10,11].
Knowing the role of influenza virus PB1-F2 protein in apoptosis, we decided to elucidate its mechanism of action and the role that the protein may play in viral infection. The sensitivity to apoptosis of PB1-F2-expressing cells was greatly enhanced in response to different cellular apoptotic stimuli, such as DNA damage, anoikis, and signaling through death receptors (see Figure 1). Similar findings were previously reported for the hepatitis B virus X protein [37]. Since proteins of the BH3 family such as Bid have been implicated in both intrinsic and extrinsic apoptotic signaling [20,22,38], we hypothesized that PB1-F2 may sensitize mitochondria to proapoptotic effects of Bid. Indeed, treatment of purified mitochondria with recombinant tBid in the presence of PB1-F2 resulted in enhanced cytochrome c release (see Figure 4). Furthermore, the proapoptotic effect of PB1-F2 on cells in response to TNFα was blocked by Bcl-xL (see Figure 2), which could occur through inhibition by Bcl-xL of either outer or inner mitochondrial membrane permeabilization. Interestingly, Bcl-xL overexpression did not inhibit PB1-F2-induced disorganization of mitochondria (see Figure 3A). This suggests that alteration of mitochondrial organization by PB1-F2 may proceed through a different mechanism, possibly through its interaction with tubulin (see Figure 5A).
Due to the unstable nature of the PB1-F2 protein and its low expression levels within the cells, the studies described above were performed in cell lines with good transfection efficiencies, such as A549 and 293T cells. The major drawback of this strategy is the transformed nature of these cells, which may alter their apoptotic responses. In particular, as seen from Figures 1 and 2, these cells are normally resistant to proapoptotic effects of TNFα, suggesting that the survival/proinflammatory pathways activated by TNFα exert a dominant effect. Nevertheless, expression of the PB1-F2 protein shifted this balance toward the proapoptotic pathways in both cell lines.
We further evaluated the mechanism of PB1-F2-induced mitochondrial permeabilization and determined that the PB1-F2 protein directly induced cytochrome c release and loss of the mitochondrial inner membrane potential in purified mouse liver mitochondria (see Figure 4A and 4B), suggesting that PB1-F2 may be acting in permeability transition-dependent manner. This speculation was further supported when proteins of the PTPC ANT3 and VDAC1 were identified in our PB1-F2 interaction screen.
To confirm that PB1-F2-induced mitochondrial permeabilization proceeds through the pore complex in an ANT3-dependent manner, we used two known ANT3 blockers, BA and CsA. These inhibitors attenuated the PB1-F2-induced loss of the mitochondrial membrane potential both in purified mitochondria (Figure 7A), and in live cells (Figure 7B). Furthermore, BA inhibited PB1-F2-mediated sensitization of cells to apoptosis by TNFα (Figure 8A). This suggests that, while similarly to the proteins of the Bcl-2 family, the PB1-F2 protein may permeabilize membranes nonspecifically [6,8,9], it induces a permeability transition in the inner mitochondrial membrane in an ANT3-dependent manner.
ANT3 is localized to the inner mitochondrial membrane and contributes to apoptosis induction by forming nonspecific pores that result in dissipation of the inner mitochondrial membrane potential. Direct interaction of ANT3 with several cellular and viral proteins (including the HIV Vpr protein) has been shown to result in mitochondrial permeabilization, leading to apoptosis [10,11]. The idea of the importance of ANT3 involvement in apoptosis has recently been challenged by the finding that the knockout of the mouse homologs of ANT3 did not alter the induction of apoptosis [39]. The results of that work nevertheless suggested that ANT3 does have an essential role in regulating permeability transition by modulating the sensitivity of the pore complex to Ca2+ activation and ANT ligands [39]. Similarly to the HIV Vpr, the influenza virus PB1-F2 protein may act as a direct ligand to ANT3 [40].
Interaction of the PB1-F2 protein with both ANT3 and VDAC1 was a surprising finding, since to our knowledge this is the first viral protein shown to interact with components of the pore complex located in both inner and outer mitochondrial membranes. We further determined that, while interaction of PB1-F2 protein with ANT3 occurs through its C-terminal domain (see Figure 6), interaction of PB1-F2 protein with VDAC1 is mediated through both N- and C-terminal regions. Overall, these data suggest a possible complex formation between VDAC1, ANT3, and PB1-F2, whereby PB1-F2 may bridge VDAC1 and ANT3, thus promoting formation of the PTPC (see Figure 8B). Further experiments will be needed to determine whether such a complex is indeed formed within the cell. In this study, we did not further identify the PB1-F2 protein residues responsible for the interaction with ANT3 and VDAC1. In a previous study, the inner mitochondrial membrane targeting signal was mapped to the amphipathic helix within the C-terminal region of the protein, consistent with our finding that the PB1-F2 C-terminal region interacts with the inner membrane ANT3 [4]. The amphipathic helix by itself was sufficient to dissipate the mitochondrial membrane potential, and mutations within the amphipathic helix abolished the mitochondrial localization of the protein [4]. At this point, however, we are unable to separate the mitochondrial-targeting and the ANT3-binding domains of the PB1-F2 protein, since lack of mitochondrial targeting also abolished PB1-F2-ANT3 interaction. In support of our findings, the N-terminal region of the PB1-F2 protein targeted to mitochondria with a heterologous inner mitochondrial membrane targeting sequence failed to interact with ANT3, confirming the specificity of the ANT3 interaction with the PB1-F2 C-terminus.
The induction of permeability transition by PB1-F2 could proceed through three possible mechanisms (Figure 8B). Through its interaction with both ANT3 and VDAC1, PB1-F2 could potentially act as a bridge between the two proteins, enhancing the formation of the pore complex (mechanism 1). Alternatively, the protein could act separately at the inner and outer mitochondrial membranes, in conjunction with ANT3 and VDAC1, respectively (mechanism 2). The latter model is supported by the evidence that PB1-F2 has been shown to localize to both inner and outer mitochondrial membranes and to directly induce membrane permeabilization [3,6]. Recent studies also indicate that, similar to the members of the Bcl-2 family, the PB1-F2 protein is capable of forming multimeric complexes, which are probably responsible for its membrane-permeabilizing activity [7]. Thus, direct permeabilization of the mitochondrial membrane by PB1-F2 protein complexes could account for the third possible mechanism.
We chose to focus our studies on the VDAC1 and ANT3 proteins primarily because these proteins were identified in our interaction screens, while other proteins localized to the outer and inner mitochondrial membranes and the mitochondrial matrix (Tom20, COXIV, and Tim44) were excluded. In addition, these protein isoforms are specifically known to be involved in the mitochondrial permeability transition. We cannot, however, exclude the possibility that the PB1-F2 protein may also interact with other isoforms of ANT and VDAC. In particular, a recent study showed that the VDAC2 isoform is implicated in suppression of mitochondrial apoptosis [41]. The fact that the ANT3-specific inhibitor BA did not completely inhibit PB1-F2-induced inner membrane permeabilization suggests that additional players may be involved (Figure 7A). In addition, while the results of our work propose the role of ANT3 in PB1-F2-induced permeabilization of the inner mitochondrial membrane, further experiments are needed to determine the role of the VDAC1 protein and the function of the PB1-F2 protein in the outer membrane. Our studies are currently limited by the lack of availability of specific inhibitors against the VDAC1 protein. Experiments are currently in progress to individually characterize the role of VDAC1 in PB1-F2-induced mitochondrial permeabilization.
Since the PB1-F2 protein is relatively short-lived and is expressed mainly during the later stages of the infection cycle [3], its proapoptotic effect most likely is not inhibitory to viral replication. Based on our findings, we propose the following model for the mode of PB1-F2 protein action within the cell (Figure 8). Early in the infection cycle, PB1-F2 localizes to the mitochondria, but does not induce a permeability transition due to its low levels present within the cell. This allows for maintenance of cell viability, which supports normal viral replication. During later stages of infection, when enough PB1-F2 protein is synthesized or when cellular mitochondrial apoptotic signaling pathways are activated, the mitochondria undergo a permeability transition.
While induction of apoptosis by influenza virus may at first seem counterintuitive to efficient viral production, it has been shown to be important in influenza viral replication. First of all, influenza virus production is inhibited in stable cell lines overexpressing Bcl-2, while activation of caspase 3 has been shown to be important in replication and propagation of influenza viruses [12–14,42]. In addition, activation of the apoptotic cascade has been suggested to play a role in processing of influenza viral proteins and maturation of viral particles [43]. Moreover, we speculate that sensitization of cells to death by TNFα rather than direct induction of apoptosis by PB1-F2 could have several additional advantageous effects for the virus. TNFα has been shown to exert a strong antiviral effect against influenza virus [44], which is likely to proceed through nuclear factor kappa-B- and c-Jun N-terminal kinase-dependent activation of antiviral gene expression and secretion of antiviral cytokines, such as type I interferon [45]. Sensitization of cells to the proapoptotic effects of TNFα would minimize the antiviral signaling to other cells. In support of this theory, induction of apoptosis by influenza virus has been shown to limit the release of proinflammatory cytokines [46]. Studies are currently underway to analyze a possible differential effect on cytokine induction by the wild-type and PB1-F2 mutant viruses. Finally, as suggested by previous work, immune cells seem to be more sensitive to induction of apoptosis by PB1-F2, as indicated by the finding that PB1-F2-knockout influenza virus induced less cell death than the wild-type virus in infected human monocytes. This observation suggests that the protein may play a role in down-regulation of the host immune response to the infection [3]. Interestingly, the ANT3 and VDAC1 proteins are expressed at different levels in a variety of tissues and cell types, including lymphocytes [47–50]. It is thus possible that these proteins may play a role in the regulation of differential apoptotic responses in different cell types following infection with influenza virus. Further animal studies will be needed to evaluate the function of the PB1-F2 protein in modulation of the immune response and its overall role in the pathogenesis of influenza virus infection.
Materials and Methods
Cell lines, antibodies, and reagents.
293T, A549, and HeLa cells were obtained from ATCC and were maintained in DMEM culture medium (Gibco, San Diego, California, United States) supplemented with 10% fetal calf serum (Hyclone, South Logan, Utah, United States), 100 U/ml of penicillin G sodium and 100 μg/ml of streptomycin sulfate (Gibco). A549 cells stably overexpressing Bcl-xL were generated by retroviral integration using the pLNCX2 vector system with neomycin selection marker from Clontech (Palo Alto, California, United States). Goat polyclonal antibodies to ANT3 and VDAC1 were obtained from Santa Cruz Biotechnologies (Santa Cruz, California, United States); monoclonal antibodies to Bcl-xL, cytochrome c, Tom20, Tim44 and PARP were obtained from Pharmingen (San Diego, California, United States); rabbit polyclonal antibody to GFP was obtained from Clontech; M30 antibody against cleaved cytokeratin was from Roche (Basel, Switzerland); and antibodies to Flag and HA epitopes were obtained from Sigma (St. Louis, Missouri, United States). Human anti-mitochondrial serum was obtained from Immunovision (Springdale, Arizona, United States). Monoclonal anti-PB1-F2 antibody (clone 26D3) was generated in mice immunized with full-length recombinant PB1-F2 protein produced in E. coli. Polyclonal anti-PB1-F2 serum was generated in rabbits immunized with full-length recombinant PB1-F2 protein. BA, cisplatin, and recombinant tBid were obtained from Sigma; G418 and CsA were from Calbiochem (San Diego, California, United States); Mitotracker Red and JC-1 dyes, DAPI, monoclonal anti-COXIV antibody, and secondary fluorochrome-conjugated antibodies were obtained from Molecular Probes (Eugene, Oregon, United States).
Constructs and cloning.
The pCAGGS vector for the expression of proteins under control of chicken β-actin promoter has been described previously [51]. cDNAs for the full-length ANT3, VDAC1, Bcl-xL, Bak, and Bax were obtained by reverse transcription with oligo-dT primer from RNA obtained from A549 cells. PCR for each gene was performed with gene-specific primers. An N-terminal HA or Flag tag was introduced into each construct by PCR with 5′ gene-specific primers possessing the tag sequences. Each tagged gene was introduced into the pCAGGS vector for mammalian expression and into the pcDNA 3.1+ vector (Invitrogen, Carlsbad, California, United States) for in vitro transcription. The Bcl-xL gene was in addition cloned into the pLNCX2 retroviral vector (Clontech) for stable integration into A549 cells. The PB1-F2 gene was reverse-transcribed and amplified with gene-specific primers by RT-PCR from viral RNA of the influenza virus A/PuertoRico/8/34 strain. N-terminal HA and Flag tags were introduced by PCR as outlined above. Tagged and untagged constructs were cloned into the pCAGGS vector for mammalian expression and the pGEX6P-1 vector (Amersham, Little Chalfont, United Kingdom) for bacterial expression of the GST fusion protein. The pCAGGS-GFP vector was generated by subcloning the GFP gene from the pEGFP vector (Clontech). The HA-tagged PB1-F2-GFP fusion construct was generated by insertion of the full-length HA-tagged PB1-F2 gene into the pCAGGS-GFP vector. HA-tagged constructs expressing either C- or N-terminally truncated PB1-F2-GFP fusion proteins (HA-nF2-GFP and HA-cF2-GFP, respectively) were generated by insertion of PCR-amplified C- or N-terminal domains into the pCAGGS-GFP vector (see Figure 4). The HA-MTS-nF2-GFP construct was generated by trimolecular ligation of the HA-tagged cytochrome oxidase MTS sequence cloned from A549 cells (sequence available from Clontech) and the N-terminal domain of PB1-F2 using the pCAGGS-GFP vector. Sequences of each generated construct were confirmed by automated sequencing performed at the Mount Sinai sequencing core facility. All primer sequences are available upon request.
Recombinant protein purification from E. coli.
Competent BL-21 cells (Stratagene, La Jolla, California, United States) transformed with pGEX6p-1 vector were cultured to an OD600 of 0.6 in 2XYT medium. The cells were induced for 2 h at 37 °C with 1 mM IPTG, collected in PBS, and frozen. Purification of the GST fusion proteins was performed using the GSH Sepharose resin (Amersham) according to the manufacturer's protocol. Purified protein was either eluted with glutathione buffer (Amersham) as a GST fusion protein, or cleaved from GST on the column with Prescission Protease (Pharmacia, Uppsala, Sweden) and eluted with PBS.
Transfections for localization and apoptosis assays.
A549 and HeLa cells grown on coverslips were transfected in 24-well dishes with 1 μg of vector of interest using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After 24 h, the cells were fixed with 5% formaldehyde in PBS and permeabilized with 1% Triton X-100. Proteins of interest were visualized by indirect immunofluorescence. Cells were probed with specific primary antibody for 2 h at room temperature, washed, and labeled with secondary antibody conjugated to a specific fluorophore. Labeled cells were visualized by laser scanning confocal microscopy (TCS-SP; Leica, Wetzlar, Germany) with TCS-SP software for image capture. A Zeiss (Oberkochen, Germany) Axiovert 200 microscope with Zeiss Axiovision software was used for fluorescence analysis of live cells. Protocols for assessment of anoikis have been described elsewhere [22]. Briefly, cells were transfected for 24 h, trypsinized, and resuspended in serum-free medium. Cellular morphology was analyzed for blebbing and fragmentation under a phase-contrast microscope over the next hour.
Mitochondrial purification and cytochrome c release assay.
Freshly isolated Balb/c mouse livers were homogenized and fractionated according to the protocols described previously [52]. Purified mitochondria were resuspended in MRM-S buffer (250 mM sucrose, 10 mM Hepes, 1 mM ATP, 5 mM succinate, 0.08 mM ADP, and 2 mM K2HPO4 [pH 7.4]) to a final concentration of 10 mg/ml protein. For cytochrome c release assays, 50 μg of mitochondria (by total protein) were incubated with recombinant PB1-F2 or tBid protein for 1 h at 30 °C in a total volume of 25 μl. After incubation, mitochondria were pelleted and the supernatant was collected and analyzed for cytochrome c release by immunoblotting for cytochrome c, while the mitochondrial pellet was analyzed by Western blot for Tom20.
Determination of the mitochondrial membrane potential with JC-1 fluorescence.
Purified mitochondria (50 μg) were incubated with proteins of interest for indicated times as described above. After incubation, 1 ml of 200 nM JC-1 dye in MRM-S buffer was added to the mitochondria and incubated at RT for 10 min. JC-1 fluorescence was measured in a Bio-Rad (Hercules, California, United States) fluorometer with the excitation filter of 490 nm and emission filter of 590 nm.
In vivo GST-fusion protein expression and GST pulldowns.
For identification of cellular interactors, 293T cells in 15-cm dishes were transfected with 15 μg of mammalian expression vector encoding GST-fusion proteins of interest. Cells were collected 30 h post-transfection and lysed in coimmunoprecipitation buffer (see below). Lysates were incubated with glutathione beads for 4 h, and beads were spun down and washed ten times in lysis buffer. Proteins were eluted off the beads with glutathione elution buffer, as recommended by the manufacturer's instructions (Amersham).
Mass spectrometry and protein identification.
The fraction of the eluted proteins was initially separated by SDS-PAGE (10%) and analyzed by silver stain. For mass spectrometry analysis, the eluted proteins were separated on the SDS-PAGE (10%) and visualized by Coomassie blue staining. Bands of interest were cut out from the gel, destained, reduced, alkylated, and digested with trypsin. Micro-HPLC analysis of the tryptic peptides was conducted by using an LCQ electrospray ionization ion trap mass spectrometer (ThermoFinnigan, Waltham, Massachusetts, United States) coupled to an online MicroPro-HPLC system (Eldex Laboratories, Napa, California, United States). Proteins were identified by using the peptide molecular masses and their MS/MS fragment data to search the National Center for Biotechnology Information nonredundant DNA and protein sequence databases with the program KNEXUS (Genomic Solutions, Ann Arbor, Michigan, United States).
Immunoprecipitations.
For all coimmunoprecipitation experiments, 293T cells in 35-mm dishes were transfected with 1 μg of each appropriate expression vector using Lipofectamine 2000 transfection reagent (Invitrogen) according to the manufacturer's protocol. Cells were collected 30 h post-transfection and lysed in the coimmunoprecipitation buffer: 0.5% NP-40, 150 mM NaCl, 20 mM Hepes (pH 7.4), 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 10% glycerol, and Complete Protease Inhibitor Cocktail and PMSF (Roche). Proteins were incubated at 4 °C overnight with 1 μg of the appropriate antibody, and protein complexes were precipitated with protein G agarose beads (Roche) for 2 h. Beads were washed five times in lysis buffer and resuspended in protein sample buffer. Proteins were subsequently separated by 12% SDS-PAGE and detected by Western blotting.
In vitro transcription/translation and GST pulldowns.
Proteins were in vitro transcribed and translated by use of a T7-coupled rabbit reticulocyte system (Promega, Madison, Wisconsin, United States) and 35S protein labeling mix (Perkin Elmer, Wellesley, California, United States) according to the manufacturers' protocols. For binding experiments, 35S-labeled proteins were incubated with 5 μg of either GST-PB1-F2 or GST with glutathione-Sepharose beads for 2 h at 4 °C in binding buffer (PBS with 0.25% NP-40 and 0.1% BSA) and washed with binding buffer three times. The beads were then resuspended in protein sample buffer, separated by 12% SDS-PAGE, and analyzed by fluorography for bound proteins. To visualize 35S-labeled proteins by fluorography, the gels were fixed, incubated in Amplify (Amersham), and dried before exposure to film.
We thank Dr. Domenico Tortorella for his help with generation of the stable cell lines and Heike Dorninger for her help with the apoptosis assays. Microscopy and mass spectrometry were performed at the Mount Sinai School of Medicine Core Facilities (Microscopy Shared Resource Facility and Mass Spectrometry Proteomics Laboratory, respectively), and supported in part, with funding from National Institutes of Health (NIH)-National Cancer Institute shared resources grants (R24 CA095823 and CA88325). This work was partially supported by NIH grants (PP and AGS) and the NIH training grant AI007647 (DZ). PP is an Ellison Medical Foundation Scholar in Global Infectious Diseases.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DZ, AGS, and PP conceived and designed the experiments. DZ, XX, and PP performed the experiments. DZ, AGS, XX, RW, and PP analyzed the data. DZ, CC, RW, and PP contributed reagents/materials/analysis tools. DZ and PP wrote the paper.
Abbreviations
ANT3adenine nucleotide translocator 3
BAbongkrekic acid
CsAcyclosporine A
GFPgreen fluorescent protein
GSTglutathione-S-transferase
HAhemagglutinin
NPnucleoprotein
PARPpoly A ribose polymerase
PTPCpermeability transition pore complex
TNFαtumor necrosis factor alpha
VDAC1voltage-dependent anion channel 1
==== Refs
References
Garcia-Sastre A 2001 Inhibition of interferon-mediated antiviral responses by influenza A viruses and other negative-strand RNA viruses Virology 279 375 384 11162793
Garcia-Sastre A Egorov A Matassov D Brandt S Levy DE 1998 Influenza A virus lacking the NS1 gene replicates in interferon-deficient systems Virology 252 324 330 9878611
Chen W Calvo PA Malide D Gibbs J Schubert U 2001 A novel influenza A virus mitochondrial protein that induces cell death Nat Med 7 1306 1312 11726970
Gibbs JS Malide D Hornung F Bennink JR Yewdell JW 2003 The influenza A virus PB1-F2 protein targets the inner mitochondrial membrane via a predicted basic amphipathic helix that disrupts mitochondrial function J Virol 77 7214 7224 12805420
Yamada H Chounan R Higashi Y Kurihara N Kido H 2004 Mitochondrial targeting sequence of the influenza A virus PB1-F2 protein and its function in mitochondria FEBS Lett 578 331 336 15589841
Chanturiya AN Basanez G Schubert U Henklein P Yewdell JW 2004 PB1-F2, an influenza A virus-encoded proapoptotic mitochondrial protein, creates variably sized pores in planar lipid membranes J Virol 78 6304 6312 15163724
Henklein P Bruns K Nimtz M Wray V Tessmer U 2005 Influenza A virus protein PB1-F2: Synthesis and characterization of the biologically active full length protein and related peptides J Pept Sci 11 481 490 15641124
Vieira HL Boya P Cohen I El Hamel C Haouzi D 2002 Cell permeable BH3-peptides overcome the cytoprotective effect of Bcl-2 and Bcl-X(L) Oncogene 21 1963 1977 11960369
Basanez G Zhang J Chau BN Maksaev GI Frolov VA 2001 Pro-apoptotic cleavage products of Bcl-xL form cytochrome c-conducting pores in pure lipid membranes J Biol Chem 276 31083 31091 11399768
Boya P Roumier T Andreau K Gonzalez-Polo RA Zamzami N 2003 Mitochondrion-targeted apoptosis regulators of viral origin Biochem Biophys Res Commun 304 575 581 12729592
Boya P Roques B Kroemer G 2001 New EMBO members' review: Viral and bacterial proteins regulating apoptosis at the mitochondrial level EMBO J 20 4325 4331 11500358
Nencioni L Iuvara A Aquilano K Ciriolo MR Cozzolino F 2003 Influenza A virus replication is dependent on an antioxidant pathway that involves GSH and Bcl-2 FASEB J 17 758 760 12594179
Olsen CW Kehren JC Dybdahl-Sissoko NR Hinshaw VS 1996 Bcl-2 alters influenza virus yield, spread, and hemagglutinin glycosylation J Virol 70 663 666 8523590
Hinshaw VS Olsen CW Dybdahl-Sissoko N Evans D 1994 Apoptosis: A mechanism of cell killing by influenza A and B viruses J Virol 68 3667 3673 8189504
van Gurp M Festjens N van Loo G Saelens X Vandenabeele P 2003 Mitochondrial intermembrane proteins in cell death Biochem Biophys Res Commun 304 487 497 12729583
Zamzami N Kroemer G 2001 The mitochondrion in apoptosis: How Pandora's box opens Nat Rev Mol Cell Biol 2 67 71 11413468
Martinou JC Green DR 2001 Breaking the mitochondrial barrier Nat Rev Mol Cell Biol 2 63 67 11413467
Beutner G Ruck A Riede B Brdiczka D 1998 Complexes between porin, hexokinase, mitochondrial creatine kinase and adenylate translocator display properties of the permeability transition pore. Implication for regulation of permeability transition by the kinases Biochim Biophys Acta 1368 7 18 9459579
Vyssokikh MY Brdiczka D 2003 The function of complexes between the outer mitochondrial membrane pore (VDAC) and the adenine nucleotide translocase in regulation of energy metabolism and apoptosis Acta Biochim Pol 50 389 404 12833165
Valentijn AJ Gilmore AP 2004 Translocation of full-length Bid to mitochondria during anoikis J Biol Chem 279 32848 32857 15148322
Puthalakath H Villunger A O'Reilly LA Beaumont JG Coultas L 2001 Bmf: A proapoptotic BH3-only protein regulated by interaction with the myosin V actin motor complex, activated by anoikis Science 293 1829 1832 11546872
Frisch SM 1999 Evidence for a function of death-receptor-related, death-domain-containing proteins in anoikis Curr Biol 9 1047 1049 10508612
Shimizu S Tsujimoto Y 2000 Proapoptotic BH3-only Bcl-2 family members induce cytochrome c release, but not mitochondrial membrane potential loss, and do not directly modulate voltage-dependent anion channel activity Proc Natl Acad Sci U S A 97 577 582 10639121
Halestrap AP Brennerb C 2003 The adenine nucleotide translocase: A central component of the mitochondrial permeability transition pore and key player in cell death Curr Med Chem 10 1507 1525 12871123
Vyssokikh M Brdiczka D 2004 VDAC and peripheral channelling complexes in health and disease Mol Cell Biochem 256–257(1–2): 117–126.
Shimizu S Narita M Tsujimoto Y 1999 Bcl-2 family proteins regulate the release of apoptogenic cytochrome c by the mitochondrial channel VDAC Nature 399 483 487 10365962
Rizzuto R Brini M Pizzo P Murgia M Pozzan T 1995 Chimeric green fluorescent protein as a tool for visualizing subcellular organelles in living cells Curr Biol 5 635 642 7552174
Henderson PJ Lardy HA 1970 Bongkrekic acid. An inhibitor of the adenine nucleotide translocase of mitochondria J Biol Chem 245 1319 1326 4245638
Danial NN Korsmeyer SJ 2004 Cell death: Critical control points Cell 116 205 219 14744432
Kroemer G 1997 The proto-oncogene Bcl-2 and its role in regulating apoptosis Nat Med 3 614 620 9176486
Massari P Ho Y Wetzler LM 2000
Neisseria meningitidis porin PorB interacts with mitochondria and protects cells from apoptosis Proc Natl Acad Sci U S A 97 9070 9075 10922061
Muller A Gunther D Brinkmann V Hurwitz R Meyer TF 2000 Targeting of the pro-apoptotic VDAC-like porin (PorB) of Neisseria gonorrhoeae to mitochondria of infected cells EMBO J 19 5332 5343 11032801
Rahmani Z Huh KW Lasher R Siddiqui A 2000 Hepatitis B virus X protein colocalizes to mitochondria with a human voltage-dependent anion channel, HVDAC3, and alters its transmembrane potential J Virol 74 2840 2846 10684300
Goldmacher VS Bartle LM Skaletskaya A Dionne CA Kedersha NL 1999 A cytomegalovirus-encoded mitochondria-localized inhibitor of apoptosis structurally unrelated to Bcl-2 Proc Natl Acad Sci U S A 96 12536 12541 10535957
Jacotot E Ferri KF El Hamel C Brenner C Druillennec S 2001 Control of mitochondrial membrane permeabilization by adenine nucleotide translocator interacting with HIV-1 viral protein rR and Bcl-2 J Exp Med 193 509 519 11181702
Everett H Barry M Sun X Lee SF Frantz C 2002 The myxoma poxvirus protein, M11L, prevents apoptosis by direct interaction with the mitochondrial permeability transition pore J Exp Med 196 1127 1139 12417624
Su F Schneider RJ 1997 Hepatitis B virus HBx protein sensitizes cells to apoptotic killing by tumor necrosis factor alpha Proc Natl Acad Sci U S A 94 8744 8749 9238048
Rytomaa M Martins LM Downward J 1999 Involvement of FADD and caspase-8 signalling in detachment-induced apoptosis Curr Biol 9 1043 1046 10508619
Kokoszka JE Waymire KG Levy SE Sligh JE Cai J 2004 The ADP/ATP translocator is not essential for the mitochondrial permeability transition pore Nature 427 461 465 14749836
Jacotot E Ravagnan L Loeffler M Ferri KF Vieira HL 2000 The HIV-1 viral protein R induces apoptosis via a direct effect on the mitochondrial permeability transition pore J Exp Med 191 33 46 10620603
Cheng EH Sheiko TV Fisher JK Craigen WJ Korsmeyer SJ 2003 VDAC2 inhibits BAK activation and mitochondrial apoptosis Science 301 513 517 12881569
Wurzer WJ Planz O Ehrhardt C Giner M Silberzahn T 2003 Caspase 3 activation is essential for efficient influenza virus propagation EMBO J 22 2717 2728 12773387
Lowy RJ 2003 Influenza virus induction of apoptosis by intrinsic and extrinsic mechanisms Int Rev Immunol 22 425 449 12959753
Seo SH Webster RG 2002 Tumor necrosis factor alpha exerts powerful anti-influenza virus effects in lung epithelial cells J Virol 76 1071 1076 11773383
Levy DE Garcia-Sastre A 2001 The virus battles: IFN induction of the antiviral state and mechanisms of viral evasion Cytokine Growth Factor Rev 12 143 156 11325598
Brydon EW Smith H Sweet C 2003 Influenza A virus-induced apoptosis in bronchiolar epithelial (NCI-H292) cells limits pro-inflammatory cytokine release J Gen Virol 84 2389 2400 12917460
Doerner A Pauschinger M Badorff A Noutsias M Giessen S 1997 Tissue-specific transcription pattern of the adenine nucleotide translocase isoforms in humans FEBS Lett 414 258 262 9315697
Stepien G Torroni A Chung AB Hodge JA Wallace DC 1992 Differential expression of adenine nucleotide translocator isoforms in mammalian tissues and during muscle cell differentiation J Biol Chem 267 14592 14597 1378836
Kayser H Kratzin HD Thinnes FP Gotz H Schmidt WE 1989 [Identification of human porins. II. Characterization and primary structure of a 31-lDa porin from human B lymphocytes (Porin 31HL).] Biol Chem Hoppe Seyler 370 1265 1278 2559745
Blachly-Dyson E Zambronicz EB Yu WH Adams V McCabe ER 1993 Cloning and functional expression in yeast of two human isoforms of the outer mitochondrial membrane channel, the voltage-dependent anion channel J Biol Chem 268 1835 1841 8420959
Niwa H Yamamura K Miyazaki J 1991 Efficient selection for high-expression transfectants with a novel eukaryotic vector Gene 108 193 199 1660837
Wei MC Lindsten T Mootha VK Weiler S Gross A 2000 tBID, a membrane-targeted death ligand, oligomerizes BAK to release cytochrome c Genes Dev 14 2060 2071 10950869
|
16201016
|
PMC1238739
|
CC BY
|
2021-01-05 12:11:28
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e4
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010004
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 10.1371/journal.ppat.001000505-PLPA-RA-0011R2plpa-01-01-03Research ArticleEvolutionInfectious DiseasesMicrobiologyMolecular Biology - Structural BiologyEubacteriaAncient Origin and Gene Mosaicism of the Progenitor of Mycobacterium
tuberculosis
Origin and Evolution of Tubercle BacilliGutierrez M. Cristina 1*Brisse Sylvain 2Brosch Roland 3Fabre Michel 4Omaïs Bahia 1Marmiesse Magali 3Supply Philip 5Vincent Veronique 1
1 Laboratoire de Référence des Mycobactéries, Institut Pasteur, Paris, France
2 Unité de Biodiversité des Bactéries Pathogènes Emergentes, Institut Pasteur, Paris, France
3 Unité de Génétique Moléculaire Bactérienne, Institut Pasteur, Paris, France
4 Laboratoire de Biologie Clinique, HIA Percy, Clamart, France
5 INSERM U629, Institut Pasteur de Lille, Lille, France
Ramakrishnan Lalita EditorUniversity of Washington, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 1 1 e528 3 2005 6 7 2005 Copyright: © 2005 Gutierrez et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.The highly successful human pathogen Mycobacterium tuberculosis has an extremely low level of genetic variation, which suggests that the entire population resulted from clonal expansion following an evolutionary bottleneck around 35,000 y ago. Here, we show that this population constitutes just the visible tip of a much broader progenitor species, whose extant representatives are human isolates of tubercle bacilli from East Africa. In these isolates, we detected incongruence among gene phylogenies as well as mosaic gene sequences, whose individual elements are retrieved in classical M. tuberculosis. Therefore, despite its apparent homogeneity, the M. tuberculosis genome appears to be a composite assembly resulting from horizontal gene transfer events predating clonal expansion. The amount of synonymous nucleotide variation in housekeeping genes suggests that tubercle bacilli were contemporaneous with early hominids in East Africa, and have thus been coevolving with their human host much longer than previously thought. These results open novel perspectives for unraveling the molecular bases of M. tuberculosis evolutionary success.
Synopsis
Mycobacterium tuberculosis, the agent of tuberculosis, is a highly successful human pathogen and kills nearly 3 million persons each year. This pathogen and its close relatives sum up in a single and compact clonal group dating back only a few tens of thousands of years. Using genetic data, the researchers have discovered that human tubercle bacilli from East Africa represent extant bacteria of a much broader progenitor species from which the M. tuberculosis clonal group evolved. They estimate that this progenitor species is as old as 3 million years. This suggests that our remote hominid ancestors may well have already suffered from tuberculosis. In addition, the researchers show that tubercle bacilli are able to exchange parts of their genome with other strains, a process that is known to play a crucial role in adaptation of pathogens to their hosts. Thus, the M. tuberculosis genome appears to be a composite assembly, resulting from ancient horizontal DNA exchanges before its clonal expansion. These findings open novel perspectives for unraveling the origin and the molecular bases of M. tuberculosis evolutionary success, and lead to reconsideration of the impact of tuberculosis on human natural selection.
Citation:Gutierrez MC, Brisse S, Brosch R, Fabre M, Omaïs B, et al. (2005) Ancient origin and gene mosaicism of the progenitor of Mycobacterium tuberculosis. PLoS Pathog 1(1): e5.
==== Body
Introduction
Most bacterial species consist of a wide spectrum of distinct clones or clonal complexes [1–3] that differ from one another by 1% or more at synonymous nucleotide sites [4,5]. Intraspecies genetic diversity is usually generated both by mutations and by horizontal genetic exchanges. However, some important human pathogens such as Salmonella enterica serotype Typhi [6] and Yersinia pestis [1] essentially consist of a single specialized clone that recently evolved from a well-known more diversified progenitor species. Members of the Mycobacterium tuberculosis complex (MTBC), the agents responsible for tuberculosis, are among the most successful human pathogens. The MTBC as defined here comprises the so-called M. tuberculosis, M. bovis, M. microti, M. africanum, M. pinnipedii, and M. caprae species. Although the members of the MTBC display different phenotypic characteristics and mammalian host ranges, they represent one of the most extreme examples of genetic homogeneity, with about 0.01%–0.03% synonymous nucleotide variation [7–12] and no significant trace of genetic exchange among them [8,13–15]. Therefore, it is believed that the members of the MTBC are the clonal progeny of a single successful ancestor, resulting from a recent evolutionary bottleneck that occurred 20,000 to 35,000 y ago [7,8,11,16].
However, the nature and the boundaries of the bacterial pool that existed prior to the putative bottleneck, as well as the time of the transition to pathogenicity for mammalian hosts, have not yet been identified. A preliminary report suggested that M. canettii, a rare tubercle bacillus with an unusual smooth colony phenotype [17], could represent the most ancestral lineage of the MTBC [18]. However, this speculation relied only on the identification of one to four nucleotide polymorphisms in a single gene. Here, based on an extensive genetic analysis including seven genes, we found that M. canettii and other smooth tubercle bacilli actually correspond to pre-bottleneck lineages, belonging to a much broader progenitor species from which the MTBC emerged.
Results/Discussion
Identification of Clonal Groups of Smooth Tubercle Bacilli
We extensively characterized 37 pulmonary and extra-pulmonary isolates of smooth tubercle bacilli (see Material and Methods; Table S1) from European and African patients, mostly immunocompetent subjects who live or have lived in Djibouti, East Africa. Genotyping with a broad set of repetitive DNA and long sequence polymorphism markers led to recognition of eight clonal groups, designated A to I, within which the markers were virtually identical (Figure S1; Table S2). According to these markers, only groups A and C/D corresponded to M. canettii isolates, as defined by van Soolingen et al. [17] and Brosch et al. [16]. Group B was closely related to M. canettii but differed by the presence of RD12can, characteristically deleted in M. canettii, and by the absence of IS1081 insertion sequence. The five other groups of smooth tubercle bacilli were remarkably distinctive from M. canettii and the thousands of MTBC strains globally investigated up to now, notably by lacking IS1081 and/or the direct repeat (DR) locus [19].
Smooth Tubercle Bacilli and MTBC Form a Single Mycobacterial Species
To determine the positions of the smooth tubercle bacilli within the Mycobacterium genus, we classically sequenced portions of six housekeeping genes (katG, gyrB, gyrA, rpoB,
hsp65, and sodA) and the complete 16S rRNA gene of all isolates of groups A, B, E, F, G, H, and I, of representative isolates of group C/D, and of representative strains of the MTBC members (Table S3). Consistent with the analysis involving repetitive DNA and long sequence polymorphism markers, all gene fragments were identical for smooth strains belonging to the same group, but differed between the groups. The comparison of the sequences of 16S rRNA (Figure 1) and these housekeeping genes (data not shown) with those of other mycobacterial species demonstrated that the eight groups of the smooth strains and MTBC members form a single species, defined by a compact phylogenetic clade remote from the other species of the Mycobacterium genus. The 1,537-bp 16S rRNA sequences of smooth groups E to I were identical to their MTBC counterparts, whereas the sequences of groups A to D differed only by a single nucleotide from the MTBC.
Figure 1 Phylogenetic Position of the Tubercle Bacilli within the Genus Mycobacterium
The blue triangle corresponds to tubercle bacilli sequences that are identical or differing by a single nucleotide. The sequences of the genus Mycobacterium that matched most closely to those of M. tuberculosis were retrieved from the BIBI database (http://pbil.univ-lyon.fr/bibi/) and aligned with those obtained for 17 smooth and MTBC strains. The unrooted neighbor-joining tree is based on 1,325 aligned nucleotide positions of the 16S rRNA gene. The scale gives the pairwise distances after Jukes-Cantor correction. Bootstrap support values higher than 90% are indicated at the nodes.
Population Structure of the Tubercle Bacilli Species
The DNA sequences of multiple housekeeping genes can be used to infer the population structure and the phylogenetic history of bacterial species [1–4]. To investigate the population structure of the tubercle bacilli species, we aligned the 3,387 nucleotides sequenced in the six housekeeping genes of the representative smooth and MTBC isolates. The alignment revealed no insertions or deletions. We identified 52 polymorphic nucleotide sites (1.54%), of which 46 were synonymous substitutions. Two of the six nonsynonymous sites were located in the katG and gyrA genes. These two mutations, together with the presence of the TbD1 and RD9 genomic regions in all the smooth isolates, classify the smooth strains among the most ancient phylogenetic lineages of tubercle bacilli [7,16].
Each unique gene sequence was assigned a different allele number, resulting in two to 11 alleles per gene. The distances between the various alleles were calculated using the mean percent divergence at synonymous (Ks) and nonsynonymous sites (Ka). The distances between the alleles of the MTBC strains were always much smaller than those between the alleles of the smooth strains (Table 1). Furthermore, the distances between the MTBC alleles and the smooth tubercle bacilli alleles were within the range observed in the smooth strains alone, with the minor exception of hsp65. These results show that the whole MTBC is only a subset of the larger tubercle bacillus species defined by the smooth groups. Consistently, phylogenetic analysis using a split decomposition graph showed that the MTBC forms a single compact bifurcating branch, rooted within the much larger array constituted by the smooth groups (Figure 2).
Table 1 Mean Percent Pairwise Differences at Synonymous (Ks) and Nonsynonymous (Ka) Sites
DOI: 10.1371/journal.ppat.0010005.t001
Figure 2 Splits Graph of the 17 Concatenated Sequences of the Six Housekeeping Genes
The nodes represent strains and are depicted as small red (smooth tubercle bacilli) or blue (MTBC members) squares. The scale bar represents Hamming distance. Numbers at the edges represent the percent bootstrap support of the splits obtained after 1,000 replicates. The fit was 61.7%. Note that the branching order of MTBC strains is weakly supported, and it should therefore not be seen as contradicting previous evolutionary hypotheses based on deletion patterns [16].
The mean synonymous distance among distinct alleles in the tubercle bacilli (0.0083–0.039) was similar to that observed in many bacterial species known to be diverse, such as Staphylococcus aureus (0.023–0.037) [4,5,20]. Most of the synonymous nucleotide substitutions were found only in the smooth tubercle bacilli (41/46). Our fluctuation tests [21] showed that the frequency of spontaneous drug resistance mutations in the smooth and the MTBC bacilli was similar (data not shown), arguing against the possibility that the observed nucleotide diversity of the smooth bacilli is caused by hypermutation. Likewise, the ratio of synonymous to nonsynonymous substitutions of the smooth tubercle bacilli (Ks/Ka = 33.3) is close to values observed in other bacteria (ranging from 7.2 to 39.6) [22,23], but much higher than the value of 1.6 found when comparing the whole genomes of M. tuberculosis CDC1551 and H37Rv strains [10]. This high Ks/Ka value is consistent with purifying selection acting against amino acid changes over long time periods, leading to relative accumulation of synonymous versus nonsynonymous mutations. In contrast, the low Ks/Ka value observed within the MTBC is consistent with recent expansion [4,10].
These results demonstrate that, similar to Y. pestis or S. enterica serotype Typhi [1,6], the MTBC consists of a successful clonal population that recently emerged from a much more ancient and large bacterial species, engulfing M. canettii and the other smooth groups. This supports the bottleneck hypothesis [7,16]. We propose to name this species M. prototuberculosis, to reflect its status as the M. tuberculosis progenitor (Figure 2).
Gene Mosaicism of Tubercle Bacilli
To investigate the contribution of horizontal DNA exchanges to the genetic diversity of M. prototuberculosis, we investigated split decomposition of concatenated sequences [24] and the congruence of individual gene phylogenies [25]. The network structure linking the smooth strains in the splits graph (Figure 2) revealed incongruence between their gene sequences. We also found strong inconsistencies among phylogenies of individual gene sequences (Figure S2). Furthermore, the detection of several sequence mosaics in the gyrB and gyrA gene sequences provided direct evidence of intragenic recombination among the smooth strains (see boxes in Figure 3). These two genes form a single operon. As an example of mosaics, the gyrB and gyrA sequences of smooth groups C/D and E are composed of two large blocks separated by gyrA position 461. One of these blocks is almost identical to the sequence of M. tuberculosis and the other is identical to the sequence in groups H and I. The significance of sequence mosaicism was supported by maximum chi-square (p < 0.005) and Sawyer's (p < 0.05) statistical tests. In contrast, the rare minor allele differences among the smooth strains, such as those between gyrB alleles 9 and 10, are probably due to point mutations rather than recombination. Altogether, these observations provide evidence that both mutations and DNA recombination have occurred during the evolution of smooth tubercle bacilli.
Figure 3 Nucleotide Polymorphism Detected in the Six Housekeeping Genes for the 17 Sequenced Strains
(A) Location of the genes on the genome of M. tuberculosis H37Rv. Note that gyrB and gyrA are adjacent.
(B) Pattern of polymorphic sites revealing mosaicism of sequences. Colored blocks correspond to sequence stretches in the smooth strains that are similar or identical to the sequences in the MTBC. Boxes correspond to blocks of consecutive nucleotides in smooth strains that differ by at least three nucleotides from M. tuberculosis H37Rv. The last column indicates the allele number for each gene. Letters N and s indicate nonsynonymous and synonymous substitutions, respectively.
In contrast, using the same analysis, no evidence of recombination was detected among the MTBC strains, consistent with their previously reported clonal population structure [13–15]. Remarkably, however, when compared to M. prototuberculosis, the concatenated sequences of the six housekeeping genes of the MTBC strains appear to be constituted of a mosaic of patches identical or nearly identical to sequence patches from different smooth groups (see colored blocks in Figure 3). This sequence patchwork suggests that the chromosomal framework of the MTBC, despite its present clonal and highly conserved structure, is actually a composite assembly of genetic sequences resulting from multiple remote horizontal gene transfer events. These DNA transfer events likely took place in the pool of the progenitor tubercle bacilli before the expansion of the MTBC clone. Therefore, the apparent absence of recombination among the MTBC strains after the bottleneck could have several potential explanations: the MTBC strains could have lost the capacity of horizontal gene transfer, horizontal gene transfer events are too rare among tubercle bacilli to have occurred since the MTBC bottleneck, or the MTBC ecological niche differs from that of M. prototuberculosis and offers no opportunity for recombination events.
Ancient Origin of the Tubercle Bacilli Species
Synonymous nucleotide diversity can be used to estimate the minimal age of the last common ancestor of a species [22,23]. The average pairwise difference at synonymous sites (Ks) across the six housekeeping genes for the 17 sequenced strains was 0.0148 (Protocol S1). Given previous studies that estimated the age of M. tuberculosis to be approximately 35,000 y based on bacterial synonymous substitution rates of 0.0044–0.0047 per site per million years [11,26,27], we estimated that the minimal time needed to accumulate the observed amount of synonymous divergence in the tubercle bacilli species was between 2.6 and 2.8 million y. As both smooth bacilli and M. tuberculosis are isolated from human tuberculosis cases, the most parsimonious hypothesis is that the last common ancestor of the tubercle bacilli species could already have caused human tuberculosis. Therefore, our results change the current paradigm of the recent origin of tuberculosis [7] by suggesting that its causative agent is as old as 3 million years. Tuberculosis could thus be much older than the plague [1], typhoid fever [6], or malaria [28], and might have already affected early hominids. Consistent with this speculative scenario, nearly all smooth tubercle bacilli isolated so far come from East Africa, a region where early hominids were present 3 million years ago [29]. The distribution of diversity between the variable smooth tubercle bacilli from Djibouti and the uniform worldwide MTBC is remarkably reminiscent of the distribution of human genetic diversity among world populations, with larger genetic distances observed within Africa [30]. Our findings thus suggest that, similarly to humans [31], tubercle bacilli emerged in Africa and then underwent early diversification followed by much more recent expansion of a successful clone to the rest of the world, possibly coinciding with the waves of human migration out of Africa. However, we cannot exclude the possibility that the geographical confinement of the smooth bacilli to Africa reflects failure to recognize smooth isolates found elsewhere as being genuine tubercle bacilli.
Implications for Research
A longer interaction of tubercle bacilli with humans and the occurrence of recombination among tubercle bacilli have profound implications for debated questions such as the natural selection effect of tuberculosis on human populations, and the way tubercle bacilli have evolved their exceptional ability to persist for decades in host tissues [32–34]. These issues should be re-examined in the light of this new evolutionary perspective. Future studies will show whether the extensive sequence polymorphism observed in housekeeping genes goes hand in hand with nonsynonymous mutations in antigen-encoding genes or in genes encoding potential drug or diagnostic targets. Our findings may also have important consequences for strategies of research for immunoprotective and therapeutic targets, which until now have been based on the assumption of the intrinsically confined genetic variation of the pathogen restraining the possibilities of emergence of potential escape variants [7,35]. Comparative and functional genomic analyses of smooth tubercle bacilli, apparently confined to East Africa, and classical tubercle bacilli, found worldwide, will shed light on the selective advantages that led the latter to such a successful clonal expansion.
Materials and Methods
Mycobacterial isolates.
The tubercle bacilli isolates used in this study are listed in Table S1 (smooth isolates) and Table S3 (MTBC isolates). Most of the smooth tubercle isolates were recovered from African or European patients attending two French Military Medical Centres (Bouffard and Paul Faure) in Djibouti, East Africa. Three smooth isolates originally obtained by Georges Canetti and one smooth isolate obtained from Switzerland were included as references [36]. We also included type strains of each member of the MTBC as references.
Distribution of repetitive DNA sequences and long sequence polymorphism markers.
Southern blots of genomic PvuII-digested tubercle bacilli DNA were sequentially probed with probes specific for IS6110, IS1081, DR region [17], region of difference RD12can [16], and M. canettii ISMyca1 transposase. The probe specific for this transposase is a 650-bp DNA fragment obtained by using 5′-CAAGGTCAAGACGCGTACC-3′ and 5′-TGAGCTTGTCGATTTGAGCTT-3′ primers. PCR amplification of the fragments of ISMyca1 flanking the transposase was perfomed using 5′-CTCGAACAGGTTCTGCTCATC-3′ and 5′-CGAAGTTCCCCCTTGTAGG-3′ primers. RD12can flanking regions were also amplified as previously described [16] and sequenced. To detect regions of difference RD9 and TbD1, two PCR assays were done for each strain as previously described [16]. MIRU-VNTR analysis was performed via an automated technique using the target loci previously reported [37–40].
DNA sequencing.
The whole 16S rRNA gene was amplified by using 5′-GCCGTTTGTTTTGTCAGGAT-3′ and 5′-GCTCGCAACCACTATCCAGT-3′ primers. The resulting product was sequenced using the following primers: 5′-GCCGTTTGTTTTGTCAGGAT-3′, 5′-CTGAGATACGGCCCAGACTC-3′, 5′-GCGCAGATATCAGGAGGAAC-3′, 5′-TCATGTTGCCAGCACGTAAT-3′, 5′-CCTACCGTCAATCCGAGAGA-3′, 5′-TGCATGTCAAACCCAGGTAA-3′, and 5′-TTCGGGTGTTACCGACTTTC-3′. To analyze polymorphisms in housekeeping genes, fragments of katG, gyrA, gyrB, hsp65, rpoB, and sodA genes were amplified and sequenced using previously published primers [7,41]. Each experiment was performed three times using different PCR products.
Phylogenetic analyses.
Neighbor-joining trees were constructed using PAUP* version 4.0b10 with Jukes-Cantor distance correction (http://paup.csit.fsu.edu/). Trees were drawn using TreeView version 1.5 (http://taxonomy.zoology.gla.ac.uk/rod/treeview.html). Bootstrap analysis was performed with 1,000 replicates. Numbers of synonymous substitutions per synonymous site (Ks) and nonsynonymous substitutions per nonsynonymous site (Ka) were estimated using DNASP version 4.00, using the Nei and Gojobori method after Jukes-Cantor correction for multiple substitutions [42]. The program RDP version 2 [43] was used to detect mosaic sequences using the Sawyer's and chi-square methods. The RDP GENECONV algorithm (which looks for regions within a sequence alignment in which sequence pairs are sufficiently similar to suspect recombination) was used for Sawyer's test, with a g-scale parameter of one and using both sequence triplets or sequence pairs scanning methods. p-Values were obtained with the KA method. The chi-square method was implemented using the MaxChi algorithm of RDP. Given an alignment, MaxChi examines sequence pairs and seeks recombination breakpoints by comparing the number of variable and nonvariable sites on both sides of the breakpoint. Split decomposition analysis was performed using SplitsTree version 4b06 [24].
Supporting Information
Figure S1 Genotypic Patterns of 37 Smooth Tubercle Bacilli
Lanes 1 to 37 correspond to strains 1 to 37, respectively; line 38 corresponds to the reference strain M. tuberculosis Mt14323. Strains 1 and 6 are the reference strains M. canettii 140010059 and NZM 217/94, respectively; strains 8 and 17 are previously reported M. canettii strains (see Table S1). Lane groups A to I indicate the groups with identical genotypic patterns.
(A) DR region analysis by spoligotyping.
(B–E) Southern blot analysis with DNA probes against (B) the DR region, (C) IS1081, (D) IS6110, and (E) ISMyca1, a 1.8-kb insertion sequence related to the IS4 family (see Protocol S2).
(F) Southern blot analysis with a DNA probe directed against region RD12can. PCR using primers targeting the regions flanking RD12can and further sequencing of these amplification products demonstrated an identical deletion in groups A, C/D, E, and H, whereas deletion in group F overlapped RD12can.
(373 KB DOC)
Click here for additional data file.
Figure S2 Gene Phylogenies of gyrA, gyrB, hsp65, katG, and rpoB Sequences from the Eight Smooth Tubercle Bacilli Groups and the MTBC Members
The unrooted trees were obtained using Megalign version 5.53 (DNASTAR, Madison, Wisconsin, United States).
(343 KB DOC)
Click here for additional data file.
Protocol S1 Estimation of Ks Value
(25 KB DOC)
Click here for additional data file.
Protocol S2 ISMyca1, a New Insertion Sequence
(27 KB DOC)
Click here for additional data file.
Table S1 Strains of Smooth Tubercle Bacilli
(57 KB DOC)
Click here for additional data file.
Table S2 MIRU-VNTR Patterns of Smooth Tubercle Bacilli
(361 KB DOC)
Click here for additional data file.
Table S3 MTBC Strains Used in This Study
(26 KB DOC)
Click here for additional data file.
Accession Numbers
The EMBL (http://www.ebi.ac.uk/embl/) accession numbers for the sequenced portions of katG, gyrB, gyrA, rpoB, and hsp65 genes of the smooth tubercle bacilli are AJ749904–AJ749948. The M. canettii ISMyca1 sequence has been deposited in the EMBL database under accession number AJ619854.
We thank Mark Achtman, Stewart T. Cole, and Genevieve Milon for critical reading of the manuscript, and Marie Gonçalvez, Eve Willery, and Sarah Lesjean-Pottier for excellent technical assistance. This study was supported in part by the Projet Transversal de Recherche Programme from the Institut Pasteur (PTR35). PS is a researcher of the Centre National de la Recherche Scientifique.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MCG and VV conceived and designed the experiments. MCG, BO, MM, and PS performed the experiments. MCG and SB analyzed the data. MCG, SB, RB, MF, and VV contributed reagents/materials/analysis tools. MCG, SB, RB, PS, and VV wrote the paper.
Abbreviations
DRdirect repeat
MTBC
Mycobacterium tuberculosis complex
==== Refs
References
Achtman M Zurth K Morelli G Torrea G Guiyoule A 1999
Yersinia pestis, the cause of plague, is a recently emerged clone of Yersinia pseudotuberculosis
Proc Natl Acad Sci U S A 96 14043 14048 10570195
Spratt BG 2004 Exploring the concept of clonality in bacteria Methods Mol Biol 266 323 352 15148426
Maiden MC 2000 High-throughput sequencing in the population analysis of bacterial pathogens of humans Int J Med Microbiol 290 183 190 11045923
Feil EJ Spratt BG 2001 Recombination and the population structures of bacterial pathogens Annu Rev Microbiol 55 561 590 11544367
Palys T Nakamura LK Cohan FM 1997 Discovery and classification of ecological diversity in the bacterial world: The role of DNA sequence data Int J Syst Bacteriol 47 1145 1156 9336922
Kidgell C Reichard U Wain J Linz B Torpdahl M 2002
Salmonella typhi, the causative agent of typhoid fever, is approximately 50,000 years old Infect Genet Evol 2 39 45 12797999
Sreevatsan S Pan X Stockbauer KE Connell ND Kreiswirth BN 1997 Restricted structural gene polymorphism in the Mycobacterium tuberculosis complex indicates evolutionarily recent global dissemination Proc Natl Acad Sci U S A 94 9869 9874 9275218
Gutacker MM Smoot JC Migliaccio CA Ricklefs SM Hua S 2002 Genome-wide analysis of synonymous single nucleotide polymorphisms in Mycobacterium tuberculosis complex organisms: Resolution of genetic relationships among closely related microbial strains Genetics 162 1533 1543 12524330
Cole ST Brosch R Parkhill J Garnier T Churcher C 1998 Deciphering the biology of Mycobacterium tuberculosis from the complete genome sequence Nature 393 537 544 9634230
Fleischmann RD Alland D Eisen JA Carpenter L White O 2002 Whole-genome comparison of Mycobacterium tuberculosis clinical and laboratory strains J Bacteriol 184 5479 5490 12218036
Hughes AL Friedman R Murray M 2002 Genomewide pattern of synonymous nucleotide substitution in two complete genomes of Mycobacterium tuberculosis
Emerg Infect Dis 8 1342 1346 12453367
Garnier T Eiglmeier K Camus JC Medina N Mansoor H 2003 The complete genome sequence of Mycobacterium bovis
Proc Natl Acad Sci U S A 100 7877 7882 12788972
Smith NH Dale J Inwald J Palmer S Gordon SV 2003 The population structure of Mycobacterium bovis in Great Britain: Clonal expansion Proc Natl Acad Sci U S A 100 15271 15275 14657373
Supply P Warren RM Banuls AL Lesjean S Van Der Spuy GD 2003 Linkage disequilibrium between minisatellite loci supports clonal evolution of Mycobacterium tuberculosis in a high tuberculosis incidence area Mol Microbiol 47 529 538 12519202
Hirsh AE Tsolaki AG DeRiemer K Feldman MW Small PM 2004 Stable association between strains of Mycobacterium tuberculosis and their human host populations Proc Natl Acad Sci U S A 101 4871 4876 15041743
Brosch R Gordon SV Marmiesse M Brodin P Buchrieser C 2002 A new evolutionary scenario for the Mycobacterium tuberculosis complex Proc Natl Acad Sci U S A 99 3684 3689 11891304
van Soolingen D Hoogenboezem T de Haas PE Hermans PW Koedam MA 1997 A novel pathogenic taxon of the Mycobacterium tuberculosis complex, Canetti: Characterization of an exceptional isolate from Africa Int J Syst Bacteriol 47 1236 1245 9336935
Fabre M Koeck JL Le Fleche P Simon F Herve V 2004 High genetic diversity revealed by variable-number tandem repeat genotyping and analysis of hsp65 gene polymorphism in a large collection of “Mycobacterium canettii” strains indicates that the M. tuberculosis complex is a recently emerged clone of “M. canettii”
J Clin Microbiol 42 3248 3255 15243089
Kamerbeek J Schouls L Kolk A van Agterveld M van Soolingen D 1997 Simultaneous detection and strain differentiation of Mycobacterium tuberculosis for diagnosis and epidemiology J Clin Microbiol 35 907 914 9157152
Enright MC Robinson DA Randle G Feil EJ Grundmann H 2002 The evolutionary history of methicillin-resistant Staphylococcus aureus (MRSA) Proc Natl Acad Sci U S A 99 7687 7692 12032344
David HL 1970 Probability distribution of drug-resistant mutants in unselected populations of Mycobacterium tuberculosis
Appl Microbiol 20 810 814 4991927
Falush D Kraft C Taylor NS Correa P Fox JG 2001 Recombination and mutation during long-term gastric colonization by Helicobacter pylori: Estimates of clock rates, recombination size, and minimal age Proc Natl Acad Sci U S A 98 15056 15061 11742075
Ochman H Elwyn S Moran NA 1999 Calibrating bacterial evolution Proc Natl Acad Sci U S A 96 12638 12643 10535975
Huson DH 1998 SplitsTree: Analyzing and visualizing evolutionary data Bioinformatics 14 68 73 9520503
Dykhuizen DE Green L 1991 Recombination in Escherichia coli and the definition of biological species J Bacteriol 173 7257 7268 1938920
Sharp PM 1991 Determinants of DNA sequence divergence between Escherichia coli and Salmonella typhimurium: Codon usage, map position, and concerted evolution J Mol Evol 33 23 33 1909371
Smith NG Eyre-Walker A 2001 Nucleotide substitution rate estimation in enterobacteria: Approximate and maximum-likelihood methods lead to similar conclusions Mol Biol Evol 18 2124 2126 11606709
Joy DA Feng X Mu J Furuya T Chotivanich K 2003 Early origin and recent expansion of Plasmodium falciparum
Science 300 318 321 12690197
Semaw S Simpson SW Quade J Renne PR Butler RF 2005 Early Pliocene hominids from Gona, Ethiopia Nature 433 301 305 15662421
Yu N Chen FC Ota S Jorde LB Pamilo P 2002 Larger genetic differences within Africans than between Africans and Eurasians Genetics 161 269 274 12019240
Templeton A 2002 Out of Africa again and again Nature 416 45 51 11882887
Sousa AO Salem JI Lee FK Vercosa MC Cruaud P 1997 An epidemic of tuberculosis with a high rate of tuberculin anergy among a population previously unexposed to tuberculosis, the Yanomami Indians of the Brazilian Amazon Proc Natl Acad Sci U S A 94 13227 13232 9371828
Lipsitch M Sousa AO 2002 Historical intensity of natural selection for resistance to tuberculosis Genetics 161 1599 1607 12196403
Lillebaek T Dirksen A Baess I Strunge B Thomsen VO 2002 Molecular evidence of endogenous reactivation of Mycobacterium tuberculosis after 33 years of latent infection J Infect Dis 185 401 404 11807725
Musser JM Amin A Ramaswamy S 2000 Negligible genetic diversity of Mycobacterium tuberculosis host immune system protein targets: Evidence of limited selective pressure Genetics 155 7 16 10790380
Pfyffer GE Auckenthaler R van Embden JD van Soolingen D 1998
Mycobacterium canettii, the smooth variant of M. tuberculosis, isolated from a Swiss patient exposed in Africa Emerg Infect Dis 4 631 634 9866740
Supply P Lesjean S Savine E Kremer K van Soolingen D 2001 Automated high-throughput genotyping for study of global epidemiology of Mycobacterium tuberculosis based on mycobacterial interspersed repetitive units J Clin Microbiol 39 3563 3571 11574573
Le Fleche P Fabre M Denoeud F Koeck JL Vergnaud G 2002 High resolution, on-line identification of strains from the Mycobacterium tuberculosis complex based on tandem repeat typing BMC Microbiol 2 37 12456266
Frothingham R Meeker-O'Connell WA 1998 Genetic diversity in the Mycobacterium tuberculosis complex based on variable numbers of tandem DNA repeats Microbiology 144 1189 1196 9611793
Roring S Scott A Brittain D Walker I Hewinson G 2002 Development of variable-number tandem repeat typing of Mycobacterium bovis: Comparison of results with those obtained by using existing exact tandem repeats and spoligotyping J Clin Microbiol 40 2126 2133 12037076
Vincent V Brown-Elliot B Jost K Wallace R 2003
Mycobacterium: Phenotypic and genotypic identification Murray P Manual of clinical microbiology Washington (DC) ASM Press 560 584
Rozas J Sanchez-DelBarrio JC Messeguer X Rozas R 2003 DnaSP, DNA polymorphism analyses by the coalescent and other methods Bioinformatics 19 2496 2497 14668244
Martin DP Williamson C Posada D 2005 RDP2: Recombination detection and analysis from sequence alignments Bioinformatics 21 260 262 15377507
|
16201017
|
PMC1238740
|
CC BY
|
2021-01-05 12:11:28
|
no
|
PLoS Pathog. 2005 Sep 19; 1(1):e5
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010005
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 1620101810.1371/journal.ppat.001000605-PLPA-RA-0028R1plpa-01-01-02Research ArticleHIV - AIDSVirologyVirusesNatural Variation in Vif: Differential Impact on APOBEC3G/3F and a Potential Role in HIV-1 Diversification Vif, APOBEC3G/3F Neutralization, HIV-1 EvolutionSimon Viviana *Zennou Veronique ¤Murray Deya Huang Yaoxing Ho David D Bieniasz Paul D *Aaron Diamond AIDS Research Center, The Rockefeller University, New York, New York, United States of AmericaMalim Michael H EditorKing's College London, United Kingdom*To whom correspondence should be addressed. E-mail: [email protected] (VS); [email protected] (PDB)¤ Current address: Regeneron Pharmaceuticals, Tarrytown, New York, United States of America
9 2005 22 7 2005 1 1 e622 4 2005 14 6 2005 Copyright: © 2005 Simon et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.The HIV-1 Vif protein counteracts the antiviral activity exhibited by the host cytidine deaminases APOBEC3G and APOBEC3F. Here, we show that defective vif alleles can readily be found in HIV-1 isolates and infected patients. Single residue changes in the Vif protein sequence are sufficient to cause the loss of Vif-induced APOBEC3 neutralization. Interestingly, not all the detected defects lead to a complete inactivation of Vif function since some mutants retained selective neutralizing activity against APOBEC3F but not APOBEC3G or vice versa. Concordantly, independently hypermutated proviruses with distinguishable patterns of G-to-A substitution attributable to cytidine deamination induced by APOBEC3G, APOBEC3F, or both enzymes were present in individuals carrying proviruses with completely or partly defective Vif variants. Natural variation in Vif function may result in selective and partial neutralization of cytidine deaminases and thereby promote viral sequence diversification within HIV-1 infected individuals.
Synopsis
Host cells express DNA-editing enzymes, termed APOBEC3G and APOBEC3F, that are capable of profoundly attenuating the replication of retroviruses. However, human immunodeficiency viruses use an efficient approach to “neutralize” these cellular enzymes: specifically, the viral gene vif counters the antiretroviral activity of APOBEC3G and APOBEC3F by inducing their degradation.
This study systematically analyzes natural variation in the ability of Vif to neutralize APOBEC-mediated HIV-1 editing and demonstrates that some Vif mutants selectively fail to efficiently “silence” APOBEC3 enzymes. These findings extend the conventional view that errors of reverse transcription and recombination are largely responsible for the rapid evolution of HIV-1 in vivo. Indeed, variation in Vif function may profoundly impact the extent and direction of viral sequence evolution within HIV-1-infected individuals. Thus, innate cellular defenses that likely evolved to restrict retroviral replication might have been usurped by HIV to accelerate viral sequence diversification and escape from immune control and inhibition by antiretroviral drugs.
Citation:Simon V, Zennou V, Murray D, Huang Y, Ho DD, et al. (2005) Natural variation in Vif: Differential impact on APOBEC3G/3F and a potential role in HIV-1 diversification. PLoS Pathog 1(1): e6.
==== Body
Introduction
A variety of intrinsic mechanisms that protect organisms from retroviral infection and corresponding viral escape strategies have evolved as a consequence of the coexistence of retroviruses and vertebrates [1–4]. Lentiviruses, e.g., express Vif proteins that counter the antiretroviral activities of DNA-editing enzymes APOBEC3G [5–8] and APOBEC3F [9–12] by inducing their degradation by proteasomes [7,8,13–15].
While both APOBEC3G and APOBEC3F (and, to a lesser extent, APOBEC3B) exhibit anti-HIV-1 activity [9–12,16], APOBEC3F appears less potent than APOBEC3G and is partially resistant to HIV-1 Vif [11]. Importantly, the expression of APOBEC3G and APOBEC3F seems to be coordinated, probably because they have arisen by gene duplication/ recombination and possess highly homologous promoters. Both are expressed in cell populations susceptible to HIV-1 infection [9,11,17].
APOBEC3G- or APOBEC3F-catalyzed deamination is not highly sequence-specific, and many cytidines on the minus strand of nascent retroviral genomes or reporter genes can be deaminated [12,14,18–20]. However, dinucleotide context strongly influences the efficiency with which cytidine deamination occurs, such that 5′-dCdC and 5′-dTdC are the favored dinucleotides targeted by APOBEC3G and APOBEC3F, respectively [9,11,20,21]. Cytidine deamination events can be manifested as G-to-A changes on the plus strand of the viral genome [11,12,14,19–21], and incomplete suppression of cytidine deaminases by HIV-1 Vif might explain the adenosine biased nucleotide composition of the HIV-1 genome. Similarly, the occasional occurrence of hypermutated genomes containing numerous G-to-A changes [22–26] could reflect occasional inactivating mutations in HIV-1 Vif. Overall, it seems likely that Vif-mediated protection of HIV-1 genomes from host-mediated viral cDNA deamination in vivo is not absolute.
Thus far, analyses of natural variation in HIV-1 Vif function have been restricted to prediction of defects resulting from gross mutations (e.g., missing translation initiation or premature stop codons). Indeed, some reports suggest that long-term non-progressors (LTNPs) harbor grossly mutated Vif more frequently than do those with progressive HIV disease [27]. However, this is not a consistent finding [28,29]. Because natural variation in the function of apparently intact HIV-1 vif alleles remains largely unexplored, we examined the extent of variation in APOBEC3G and APOBEC3F neutralizing activity among naturally occurring HIV-1 Vif variants. Surprisingly, we found that apparently intact but inactive Vif variants are frequently detected in viral sequences from very different sources. Importantly, we also found Vif variants that selectively fail to neutralize APOBEC3G and/or APOBEC3F activity, as well as proviruses in LTNPs that appeared to be independently hypermutated as a result of APOBEC3G or APOBEC3F action. These studies indicate that sporadic inactivation of Vif likely occurs rather frequently in vivo, and that natural variation in Vif function is likely to profoundly impact the extent and direction of viral sequence evolution within HIV-1-infected individuals.
Results
Naturally Occurring Vif Diversity
Two distinct sources of Vif alleles were selected in order to either maximize or minimize the occurrence of inactive variants. The first group was derived from uncultured DNA isolated from peripheral blood mononuclear cells (PBMCs) of three extensively studied individuals (P1, P2, and P3) with non-progressive HIV-1 infection for more than 20 y [28, 30–35]. We reasoned that the very low (undetectable) level of HIV-1 replication that occurs in these individuals would maximize the chances of detecting naturally occurring defective vif alleles, because replication-defective HIV-1 variants should not be obscured by superimposed replicating virus. Moreover, some viral DNA sequences in these individuals contained numerous G-to-A changes in the 5′LTR (P2) [33] and in gag (P3) [31]. Amplification and sequence analysis of APOBEC3G mRNA from all three LTNPs excluded the possibility that hypermutation of viral sequences was due to the existence of a naturally occurring Vif-resistant APOBEC3G mutant (data not shown).
A second set of Vif variants was derived from four short-term HIV-1 isolates (V1, V2, V3, and V4) [36]. Vif variants were cloned after a limited period of co-culture of patient cells with “non-permissive” PBMCs, so as to provide a selective pressure for active vif alleles while minimizing the introduction of new mutations. V1, V3, and V4 were obtained from recently infected individuals, and V2 was obtained from a chronically infected patient. Table S1 provides a summary of characteristics of the patients from which viral isolates and DNA were obtained.
The phylogenetic relationships among a total of 79 independent vif sequences obtained from these sources is depicted in Figure 1. For each of the LTNPs, a set of Vif sequences collected in 1993–1994 [28] as well as from one or two additional time-points were included (see Material and Methods). Because the neighbor-joining analysis revealed no discernable temporal influence on phylogeny in P1, P2, and P3, the observed intra-individual sequence variation likely represents a spectrum of archived proviruses rather than ongoing sequence evolution. Consensus vif sequences from each individual or isolate diverged by 3–9% from a prototype vif allele (from HIV-1 NL4–3), a value typical of North American subtype-B vif sequences (Figure 1).
Figure 1 Naturally Occurring vif Sequence Variation
The phylogenetic relationships among 79 independent vif sequences derived from patients (P1, P2, P3) and viral isolates (V1, V2, V3, V4) were analyzed using the Neighbor-joining method. Seven subtype B reference sequences and a consensus subtype B sequence were also included. A cluster of hypermutated vif sequences found in P3 is indicated. The 40 distinct protein variants selected for functional testing are identified by •. For each patient and isolate, an individual vif consensus sequence was generated and the percent divergence of each vif consensus sequence from the NL4–3 vif referenced is shown.
While the majority of vif sequences derived from LTNPs and viral isolates encoded a full-length protein (192 residues), six sequences (from P2, P3, and V4) contained one or more premature stop codons. Interestingly, each of these resulted from a G-to-A mutation in the context of a Trp codon (TGG to TAG). In addition, a subpopulation of obviously defective Vif sequences from P3 (Figure 1) contained numerous G-to-A substitutions. Thus, expression of intact Vif proteins can be abolished by relatively rare G-to-A substitutions at tryptophan sites, or by overt hypermutation.
Frequent Occurrence of Non-Functional Vif
Simple inspection of sequences for obviously inactivating mutations likely underestimates the frequency of inactive vif alleles, and would not identify weakly active Vif proteins. Therefore, we employed a functional test to measure Vif activity. We generated VSV-G pseudotyped HIV-1 vector particles in the presence of APOBEC3G or APOBEC3F and patient/isolate-derived Vif proteins and measured their infectivity using a reporter cell-line (Figure 2). Under these conditions, HIV-1 infectivity was reduced more than 100-fold by APOBEC3G and restored by HIV-1 (NL4–3) Vif to 35–45% of the level observed in the absence of APOBEC3G (Figure 2A). APOBEC3F was a less potent inhibitor of HIV-1 and reduced infectivity by 15- to 20-fold, as has been recently reported [11]. Additionally, APOBEC3F was more resistant to neutralization by NL4–3 Vif, which restored infectivity to about 15% to 25% of its uninhibited level, even at the maximum level of Vif expression tested. The linear range of the assay with respect to the amount of transfected Vif expression plasmid was determined (Figure 2A), and levels of Vif expression within this range were used in subsequent experiments.
Figure 2 Activity of Vif Variants from Patients and Viral Isolates
(A) Infectivity of HIV-1 vector particles generated by transient transfection of 293T cells in the presence or absence of fixed amounts of APOBEC3G or APOBEC3F and the indicated amounts of HIV-1 NL4–3 Vif expression plasmids was determined, as described in the Material and Methods. Representative results from one out of three independent experiments are depicted. Infectivity measurements were performed in duplicate assays, and the error bars represent the standard deviation of the RLU values. RLU, relative light units.
(B) APOBEC3G neutralization by Vif proteins from LTNPs (P1, P2, P3) and viral isolates (V1, V2, V3, V4). The infectivity of particles generated in the presence of APOBEC3G and each Vif protein is expressed relative to the infectivity of particles generated in the absence of APOBEC3G and Vif. Each Vif protein is identified by its source (e.g., P1, V1) and a variant number (e.g., P1–2). The data represents the average infectivity values of at least three independent experiments, with the error bars showing the standard deviations.
(C) Summary of the properties of Vif variants. Independent sequences were defined as alleles that were derived from different PCR reactions or had different nucleotide sequences. Because some changes are synonymous, not all independent sequences encode variant proteins. Vif variants with gross defects (e.g., premature stop codons) as well as those that were found to be inactive in the functional assay are designated “defective Vif.” The overall frequency of inactive Vif proteins is expressed as a percentage relative to the number of independent sequences.
We next selected a panel of 40 representative and apparently intact Vif proteins from the 79 independent clones and measured their ability to neutralize APOBEC3G antiviral activity. The majority of patient- and isolate-derived Vif proteins exhibited similar activity to that of NL4–3 Vif (Figure 2B). However, each set of patient-derived and isolate-derived Vif proteins contained at least one variant that was either inactive or only weakly active (Figure 2B). In fact, nine of the 40 apparently intact Vif proteins were non-functional despite the fact that they were well expressed. One additional amino-terminally truncated Vif variant (V4–8) that contained a stop at codon 11 but was expressed via an alternative translation initiation site at codon 16 was found to be inactive. Five additional vif variants (P2, P3) that contained premature stop codons were assumed to be inactive. Therefore, of the 79 sampled vif alleles, comprising 61 distinct proteins, at least 15 were unable or poorly able to counteract APOBEC3G antiviral activity (Figure 2C). Additionally, even among functional Vif variants, there was significant variation in the extent to which they restored HIV-1 infectivity that was not explained by variation in expression level (Figure 2B, and data not shown).
Mutations Conferring Loss of APOBEC3G- and/or APOBEC3F-Neutralizing Activity
To determine the amino acid substitutions responsible for functional inactivation of apparently intact Vif proteins, we identified the closest functional relative of each non-functional variant. In most cases, we could identify functional and non-functional Vif proteins that differed by only one or two amino acids (Figure S1). Of note, two mutations (G138R and L150P) that were found only in non-functional variants arose independently in unrelated Vif proteins (from P1, P2, and/or V1, Figure 3A). The only non-functional Vif variant for which no very closely related functional counterpart was identified was from patient DNA (P2–7) which differed at four positions from its closest functional relative (W11R, K63E, G75R, G185R) and was not analyzed further.
Figure 3 Closely Related Vif Proteins Display Distinct Functional Properties
(A) Summary of the residues implicated as causing Vif defects by comparison of functional and non-functional Vif variants. Amino acid substitutions that occurred exclusively in non-functional Vif proteins are depicted relative to the NL4–3 Vif sequence. Changes identified by * are caused by G-to-A mutations.
(B, C) Function of closely related vif alleles assessed by quantitation of the infectivity of particles produced in the presence of APOBEC3G (B) or APOBEC3F (C). Amino acid substitutions in the non-functional partner of the “matched” functional Vif variant are given in parentheses in (B). The dotted line in (C) indicates the level of infectivity observed for the vector generated in the presence of APOBEC3F and in the absence of Vif. The data represent the average infectivity values of at least three independent experiments, with the error bars showing the standard deviations.
We tested the panel of “matched” functional and non-functional Vif variants for their relative ability to neutralize APOBEC3G and APOBEC3F. Like NL4–3 Vif, those that neutralized APOBEC3G also neutralized APOBEC3F, albeit less efficiently (Figure 3B and 3C). Generally, the matched variants that were inactive or weakly active against APOBEC3G also failed to neutralize APOBEC3F. However, two Vif variants (V2–5, P3–5) that contained mutations K22E and Y40H, respectively, were unusual in that they did not neutralize APOBEC3G but retained partial activity against APOBEC3F (Figure 3B and 3C).
To confirm these findings and to unambiguously identify the mutations responsible for loss of function, we constructed a panel of NL4–3 Vif mutants containing naturally occurring mutations that were associated with loss of APOBEC3G neutralization. The W11R mutant was added to this panel, as this was the only substitution in the non-functional P2–7 variant (which contained three additional substitutions, see above) that occurred at a position that is invariant in the HIV-1 sequence database.
In the presence of APOBEC3G, NL4–3 Vif harboring the mutations K22E, S32P, Y40H, E45G, F115S, G138R, or L150P failed to restore HIV-1 infectivity (Figure 4A). Conversely, NL4–3 Vif containing the mutations W11R or G143R displayed activity close to that of unmutated NL4–3 Vif. Most of the mutants exhibited similar loss or retention of activity against APOBEC3F as against APOBEC3G (compare Figures 4A and 4B). Western blotting of transfect cell lysates revealed that expression levels of the Vif mutants that fail to rescue particle infectivity were comparable or superior to the expression level observed for NL4–3 Vif (Figure 4C). Interestingly, mutants W11R, K22E, Y40H, and E45G displayed a selective propensity to neutralize one of the two APOBEC3 proteins. These mutants were analyzed in more detail by measuring viral infectivity in the presence of APOBEC3G or APOBEC3F and varying levels of Vif (Figure 4D). This analysis revealed that the K22E mutant retained partial activity against APOBEC3F but was inactive against APOBEC3G. The Y40H and E45G Vif mutants neutralized APOBEC3F activity almost as efficiently as wild-type NL4–3 Vif, but displayed only weak activity against APOBEC3G. Conversely, the W11R mutant exhibited the precisely opposite phenotype and neutralized APOBEc3G but not APOBEC3F. Thus, naturally occurring mutations can result in Vif variants that neutralize one but not the other cytidine deaminase.
Figure 4 Effect of Naturally Occurring Single Amino Acid Substitutions on NL4–3 Vif APOBEC3G and APOBEC3F Neutralization Activity
The HIV-1 vector infectivity generated in the presence of APOBEC3G (A) and APOBEC3F (B) and a fixed dose of NL4–3 Vif mutant was measured. The dotted line in (B) indicates the level of infectivity observed for the vector generated in the presence of APOBEC3F and in the absence of Vif. The data represent the average infectivity as determined by at least three independent experiments, with the error bars showing the standard deviations. (C) Protein expression levels of the NL4–3 Vif mutants were determined by Western blotting of transfected 293T cell lysates. (D) Infectivity of HIV-1 vector generated in the presence of APOBEC3G (filled symbols) and APOBEC3F (open symbols) and varying levels of selected Vif mutants.
In Vivo Sequence Diversity Consistent with Selective Exposure to APOBEC3G or APOBEC3F
In principle, mutations that selectively affect the ability of Vif to neutralize APOBEC3G and/or APOBEC3F could profoundly affect HIV-1 sequence evolution. Because we found defective Vif variants in patients who also contained G-to-A hypermutated viral DNA (P2 and P3) [31,33], we next obtained a more extensive sample of Gag-Pol sequences from patients and viral isolates harboring defective Vif alleles (a total of 25 independent Gag-Pol sequences of which 13 representative sequences are depicted in Figure 5).
Figure 5 Analysis of Gag-pol Sequences Derived from LTNP and Viral Isolates
(A) Graphic representation of the G-to-A changes (compared to HIV-1 NL4–3) present in Gag-pol. Analysis was performed using the HYPERMUT program [48].
(B) Quantitative summary of the observed changes. “# diff” is the number of positions at which the patient sequence differs from NL4–3. “# G-A (%)” represents the absolute number and percentage of all substitutions that are G-to-A changes. The dinucleotide context (GG, GA, GC, GT) reflects the two contiguous bases, with G-to-A mutations occurring in the first position. The numbers of differences (“# diff”) for the APOBEC3G- and APOBEC3F-induced changes in the in vitro assay are given as an average (± standard deviation) of those occurring in 10 to 15 clones of hrGFP (450 nucleotides).
In general, the Gag-Pol sequences from the LTNPs contained about 3-fold fewer changes relative to an NL4–3 HIV-1 reference sequence than did the viral isolate sequences. This is perhaps due to the fact that the three LTNPs became infected more than 20 y ago, quite early in the North-American HIV-1 epidemic and around the time the viral isolate NY5, which comprises the 5′ portion of the reference sequence NL4–3 [37], was obtained. Conversely, the viral isolates were derived from patients infected in the late 1990s. Nonetheless, the fraction of changes relative to the reference sequence that were G-to-A mutations was quite similar for patient and viral isolate sequences (about 25% of all changes) and higher than would be expected if nucleotide substitutions were random (Figure 5B). This phenotype was exaggerated in several P2- and P3-derived sequences that were clearly hypermutated. In some cases, examples of which are shown in Figure 5A and 5B, about 65–70% of all changes relative to the reference sequence were G-to-A changes. In P1, no obviously hypermutated viral sequences were detected, while in P2 and P3, viral DNA was often hypermutated and defective. Indeed, hypermutated Vif sequences were previously detected in P3 (see Figure 1).
In principle, G-to-A mutations can occur in one of four different dinucleotide contexts. We sequenced hrGFP-containing viral DNA generated following infection with HIV-1 vectors assembled in the presence of APOBEC3G or APOBEC3F. Consistent with recent reports [9,11], APOBEC3F primarily induced GA-to-AA changes and very few GG-to-AG changes in hrGFP. Conversely, APOBEC3G caused twice as many GG-to-AG as GA-to-AA changes (Figure 5B). In the sequences from all the isolate-derived DNA as well as in some patient-derived Gag-Pol clones (e.g., P1–29, P3–11; Figure 5B), GA-to-AA changes were more prevalent than GG-to-AG changes, relative to the reference sequence, suggesting that HIV-1 is more often mutated by APOBEC3F than by APOBEC3G. In contrast, in the extensively hypermutated Gag-Pol clones (e.g., P2–21, P2–31, P3–2, Figure 5A and 5B) there was an approximately 2:1 ratio of GG-to-AG versus AG-to-AA substitutions, which closely recapitulated the result of APOBEC3G- (and not APOBEC3F-) induced hrGFP mutagenesis.
Because the p17MA region of Gag provided a reasonable representation of mutations present in Gag-Pol, and because p17MA sequences present in P3 had previously been extensively sampled [31], we restricted a more detailed phylogenetic analysis to this region. Thus, the larger dataset comprised a total of 70 p17MA sequences, and the phylogenetic analysis is depicted in Figure 6A. Notably, the p17MA sequences obtained from a single individual harboring numerous hypermutated proviruses (P3) were exceptionally heterogeneous and exhibited greater diversity than that observed among a selection of subtype B reference sequences. While very diverse hypermutated sequences co-existed in a patient and sometimes within a single sample (Figure 6B), sampling bias due to the very low number of proviruses present in these LTNPs makes it difficult to derive conclusions about temporal variation in the proportion of hypermutated to “normal” sequences. Nonetheless, it was clear that in some individuals at some time-points, hypermutated sequences predominated. Notably, individual p17MA clones contained distinct patterns of G-to-A substitutions that reflect the generation of multiple, independently hypermutated proviruses within a single individual. Moreover, some clones almost exclusively harbored GG-to-AG changes relative to the reference sequence (e.g., P2–3, P2–5, P3–8, P3–9, Figure 6B), while in others GA-to-AA changes predominated (e.g., P3–13, P3–14, Figure 6B). In other clones, GG-to-AG and GA-to-AA substitutions occurred in variable proportions. This bias recapitulated the footprints expected to result from mutation induced either by APOBEC3G, APOBEC3F, or a mixture of both and is likely, therefore, the result of variation in the ability of Vif, within an individual, to neutralize these cytidine deaminases.
Figure 6 Phylogenetic Relationships and Hypermutation in p17MA Sequences
(A) Neighbor joining tree representing the phylogenetic relationships among 70 independent p17MA sequences derived from patients (P1, P2, P3) and viral isolates (V1, V2, V3, V4). Additional subtype B reference sequences were also included.
(B) Graphic representation of the G-to-A changes (compared to HIV-1 NL4–3) present in p17MA sequences of P2 and P3. Analysis was performed using the HYPERMUT program [48]. All possible G-to-A substitutions in the context of the reference sequence (NL4–3) are shown with the dinucleotide context color coded (bottom four sequences).
Discussion
Host cells use DNA-editing enzymes such as APOBEC3G and APOBEC3F to inhibit the replication of exogenous and endogenous retroviruses [38,39]. In turn, lentivirus Vif proteins counter the antiretroviral activity of these cytidine deaminases [5–12] by inducing their degradation.
While defective vif alleles should render the majority of HIV-1 progeny virions noninfectious, establishment of proviruses that have APOBEC3G- and/or APOBEC3F-induced mutations should endow the proviral population with additional diversity beyond that resulting from error-prone reverse transcription [1,3]. Here, we show that defective vif alleles are easily found, even in a relatively small sample of viral isolates and uncultured viral DNA from infected individuals. Because active and defective Vif variants were detected in similar proportions (about 19%) in LTNP DNA and viral isolates, it seems probable that Vif is sporadically inactivated at some frequency by reverse transcriptase errors or cytidine deamination in all HIV-1-infected individuals. We also found multiple, independent hypermutated proviruses in two of the three LTNPs studied. It is likely that the relative ease of detection of extensively hypermutated proviruses is an unusual property of LTNPs [26] and hypermutated proviruses are only rarely detected in other situations because they are obscured by superimposed replication competent virus. Put another way, the ease of detection of hypermutated HIV-1 is likely a consequence, rather than a cause, of LTNP status.
A recent report provides independent evidence for the occurrence of sporadic Vif inactivation in vivo: indeed, more than 9% of proviral sequences derived from resting CD4+ cells of HAART-treated patients with plasma viremia below the level of detection were found to be hypermutated [40].
Site-directed mutagenesis studies revealed a number of residues and domains located throughout Vif that are essential for infectivity and viral replication in non-permissive cells [41–44]. We identified here several examples of naturally occurring mutations in Vif that induce selective defects in APOBEC3G- or APOBEC3F-neutralizing activity. Interestingly, these occurred toward the amino-terminus of Vif, while those causing general defects occurred either near the amino-terminus or close to the functionally important SLQ motif [41,42,45]. These findings suggest the possibility that the Vif amino-terminal domain contains determinants that confer specific binding to APOBEC3G or APOBEC3F. Such mutations should cause nascent viral DNA to be exposed either to APOBEC3G or APOBEC3F or both, with the induction of very different patterns of G-to-A changes in the viral quasi-species. Due to the nature of sampling proviral DNA in a patient, it is practically impossible to link a particular Vif point mutant to a subsequently generated hypermutated provirus. Nonetheless, even within a tiny fraction of the total burden of, e.g., P3′s HIV-1 sequences, Vif variants that selectively failed to neutralize APOBEC3G as well as p17MA sequences bearing the footprint of selective mutation by either cytidine deaminase were rather easily detected (Figure 6). Selective and sporadic loss of cytidine deaminase neutralization can result in massive sequence variation. Indeed, the extent of p17MA sequence diversity within one of the individuals studied herein (P3) surpassed the variation observed among a selection of subtype B reference sequences and contemporary isolates. While the bulk of hypermutated proviruses are replication defective, they could, in principle, provide a genetic “resource,” and selected fragments could easily be recombined into the circulating viral population as cells harboring them become super-infected. Moreover, while intermittent inactivation of Vif has the measurable consequence of deposition of hypermutated proviruses, this likely represents an extreme manifestation of the loss-of-function phenotype. It is quite likely that more subtle (and less easily measurable) variation in Vif's anti-APOBEC3G and anti-APOBEC3F activities occurs at greater frequency than the roughly 20% inactivation documented herein. It is also interesting that some of the mutations that affected Vif function were themselves the result of G-to-A changes. Thus, “feedback” loops of Vif mutation resulting in more or less mutation of Vif and other viral genes could result from variation in Vif function.
The high adenosine content of lentiviral genomes [22] and the sporadic detection of hypermutated genomes in vivo [23–26] suggest that cytidine deamination provides an important source of viral diversification. Thus, an activity that likely evolved as a host defense against viruses might have been usurped and may facilitate HIV-1 escape from immunological and pharmacological inhibition. In addition, an appreciation of the respective contributions of viral and host-mediated mutagenesis to viral diversity may be important for determining the potential efficacy and the potential risks associated with the use of antiretroviral strategies that target vif:APOBEC3G/3F interactions.
Materials and Methods
Patient-derived and NL4–3 mutant vif alleles.
DNA was extracted from either patient's PBMCs (P1, P2, and P3), or from PBMCs used for viral propagation (V1, V2, V3, V4) using DNeasy DNA isolation kit (Qiagen, Valencia, California, United States). Samples were obtained in 1995 and 2000 for P1, in 2000 and 2001 for P2, and 2003 for P3. Because clinical materials from the mid-1990s were not available for P3, we reconstructed the two P3′s vif alleles that were previously described and contained amino acid substitutions relative to a P3 sequence from 2003 (Y40H and F115S [28]). Full-length vif genes were amplified by nested PCR using high-fidelity polymerase and cloned into the expression vector pCRV1, as previously described [46]. In order to minimize sampling bias due to the very low proviral load in some LTNP samples (e.g., less than five copies/106 PBMCs [30]), multiple parallel PCR reactions were done. Cloned Vif variants were sequenced in both directions and aligned using the DNAStar (Madison, Wisconsin, United States) Version 4.03 software package. Phylogenetic relationships were assessed by the Neighbor Joining method (PAUP software). Site-directed mutations were introduced into NL4–3 Vif using mutagenic oligonucleotides and recombinant PCR resulting in NL4–3 Vif mutants encoding W11R, K22E, S32P, Y40H, E45G, F115S, G138R, G143R, and L150P as well as the previously described mutants C114S and C133S [47]. Wild-type and mutant NL4–3 Vif proteins as well as the patient- and primary-isolate-derived Vif variants were expressed using pCRV1.
Gag-Pol and p17MA sequence analysis.
A region spanning the Gag, protease (PR), and half of the reverse transcriptase gene (RT; total 3,094 nucleotides) was amplified from all three LTNPs and the four viral isolates using nested PCR and the same DNA samples used as a source of vif variants. PCR fragments were cloned into TOPO XL vector (Invitrogen, Carlsbad, California, United States) and sequenced bidirectionally. To ensure appropriate sampling in the setting of low proviral load, we performed multiple PCR reactions in parallel and cloned the fragments from five to seven PCR reactions separately. Phylogenetic relationships were assessed by the Neighbor Joining method (PAUP software). Analysis of the G-to-A substitutions was performed using the HYPERMUT program [48].
Assay for Vif function.
HIV-1 vector particles were generated by transfecting 293T cells with plasmids expressing HIV-1 gag-pol (pCRV1/Gag-pol)[49], a packagable HIV-1 RNA genome that encodes only Tat, Rev, Vpu, and GFP (pV1/hrGFP) and the G protein from vesicular stomatitis virus (pHCMV VSV-G) in a 5:5:1 ratio. To measure Vif function, cells were co-transfected with this plasmid mixture and additional plasmids expressing amino-terminally HA-tagged APOBEC3G or APOBEC3F and a pCRV1/Vif variant. Cells were transfected in 24 well plates using LipoFectamine Plus reagent (Invitrogen) and the supernatant harvested 48 h later and filtered. Infectivity was measured in duplicate assays using TZM-bl cells, which carry an HIV-1 Tat responsive β-galactosidase indicator gene under the transcriptional control of the HIV-1 promoter. Infection was done in the presence of 8μg/ml Polybrene, and β-galactosidase activity was quantified 48 h later using chemiluminescent substrate as previously described [50].
Measurement of APOBEC3G- and APOBEC3F-driven hrGFP mutagenesis.
HIV-1 vector particles were generated by transfection of 293T cells, as for the analysis of Vif function with two differences: First, APOBEC3G or APOBEC3F but no Vif expression plasmid was included in the transfection mixture. Second, to minimize the possibility that transfected DNA rather than de novo synthesized viral DNA would be amplified and sequenced, the pV1/hrGFP plasmid vector was omitted from the transfection mixture and a 293T-derived cell line carrying an integrated V1/hrGFP vector genome was used to generate vector particles. Vector particles generated by this method were used to infect MT4 cells and DNA was extracted 8 to 10 h later using a DNeasy DNA isolation kit. hrGFP sequences were amplified by PCR, cloned into TOPO vector, and sequenced bidirectionally. Sequences (450-nucleotide fragment) from 10 to 15 clones were aligned using the DNAStar software package.
Supporting Information
Figure S1 The Sequences of Functional and Non-Functional Vif Alleles (Underlined) Were Compared in Order to Identify Positions Relevant for Inactivation of Vif
(24 KB PDF)
Click here for additional data file.
Table S1 Summary of Clinical and Virological Characteristics of the Four Patients from Whom Virus Was Cultivated As Well As from the Three LTNPs Studied
(16 KB PDF)
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession numbers for the vif and gag-pol sequences are DQ097739–DQ097768.
We thank C. Cheng-Mayer and L. C. F. Mulder for helpful discussions. TZM-bl cells were obtained from J. C. Kappes, X. Wu, and Transzyme Inc. through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, National Institutes of Health. This work was supported by NIH grants R21AI54185 (to VS) and R01AI50111 (to PDB). VS is a recipient of the Mark S. Bertuch AIDS Research Fund Award. PDB is an Elizabeth Glaser Scientist of the Elizabeth Glaser Pediatric AIDS Foundation.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. VS, DDH, and PDB conceived and designed the experiments. VS, VZ, DM, and YH performed the experiments. VS and PDB wrote the paper.
Abbreviations
LTNPlong-term non-progressor
PBMCperipheral blood mononuclear cell
==== Refs
References
Harris RS Liddament MT 2004 Retroviral restriction by APOBEC proteins Nat Rev Immunol 4 868 877 15516966
KewalRamani VN Coffin JM 2003 Virology. Weapons of mutational destruction Science 301 923 925 12920286
Bieniasz PD 2004 Intrinsic immunity: A front-line defense against viral attack Nat Immunol 5 1109 1115 15496950
Cullen BR 2003 HIV-1 Vif: Counteracting innate antiretroviral defenses Mol Ther 8 525 527 14565218
Mariani R Chen D Schrofelbauer B Navarro F Konig R 2003 Species-specific exclusion of APOBEC3G from HIV-1 virions by Vif Cell 114 21 31 12859895
Stopak K de Noronha C Yonemoto W Greene WC 2003 HIV-1 Vif blocks the antiviral activity of APOBEC3G by impairing both its translation and intracellular stability Mol Cell 12 591 601 14527406
Mehle A Strack B Ancuta P Zhang C McPike M 2004 Vif overcomes the innate antiviral activity of APOBEC3G by promoting its degradation in the ubiquitin-proteasome pathway J Biol Chem 279 7792 7798 14672928
Sheehy AM Gaddis NC Malim MH 2003 The antiretroviral enzyme APOBEC3G is degraded by the proteasome in response to HIV-1 Vif Nat Med 9 1404 1407 14528300
Wiegand HL Doehle BP Bogerd HP Cullen BR 2004 A second human antiretroviral factor, APOBEC3F, is suppressed by the HIV-1 and HIV-2 Vif proteins Embo J 23 2451 2458 15152192
Zheng YH Irwin D Kurosu T Tokunaga K Sata T 2004 Human APOBEC3F is another host factor that blocks human immunodeficiency virus type 1 replication J Virol 78 6073 6076 15141007
Liddament MT Brown WL Schumacher AJ Harris RS 2004 APOBEC3F properties and hypermutation preferences indicate activity against HIV-1 in vivo Curr Biol 14 1385 1391 15296757
Bishop KN Holmes RK Sheehy AM Davidson NO Cho SJ 2004 Cytidine deamination of retroviral DNA by diverse APOBEC proteins Curr Biol 14 1392 1396 15296758
Conticello SG Harris RS Neuberger MS 2003 The Vif protein of HIV triggers degradation of the human antiretroviral DNA deaminase APOBEC3G Curr Biol 13 2009 2013 14614829
Harris RS Bishop KN Sheehy AM Craig HM Petersen-Mahrt SK 2003 DNA deamination mediates innate immunity to retroviral infection Cell 113 803 809 12809610
Marin M Rose KM Kozak SL Kabat D 2003 HIV-1 Vif protein binds the editing enzyme APOBEC3G and induces its degradation Nat Med 9 1398 1403 14528301
Yu Q Chen D Konig R Mariani R Unutmaz D 2004 APOBEC3B and APOBEC3C are potent inhibitors of simian immunodeficiency virus replication J Biol Chem 8 8
Jarmuz A Chester A Bayliss J Gisbourne J Dunham I 2002 An anthropoid-specific locus of orphan C to U RNA-editing enzymes on chromosome 22 Genomics 79 285 296 11863358
Lecossier D Bouchonnet F Clavel F Hance AJ 2003 Hypermutation of HIV-1 DNA in the absence of the Vif protein Science 300 1112 12750511
Bishop KN Holmes RK Sheehy AM Malim MH 2004 APOBEC-mediated editing of viral RNA Science 305 645 15286366
Yu Q Konig R Pillai S Chiles K Kearney M 2004 Single-strand specificity of APOBEC3G accounts for minus-strand deamination of the HIV genome Nat Struct Mol Biol 18 18
Beale RC Petersen-Mahrt SK Watt IN Harris RS Rada C 2004 Comparison of the differential context-dependence of DNA deamination by APOBEC enzymes:Correlation with mutation spectra in vivo J Mol Biol 337 585 596 15019779
Berkhout B Grigoriev A Bakker M Lukashov VV 2002 Codon and amino acid usage in retroviral genomes is consistent with virus-specific nucleotide pressure AIDS Res Hum Retroviruses 18 133 141 11839146
Koulinska IN Chaplin B Mwakagile D Essex M Renjifo B 2003 Hypermutation of HIV type 1 genomes isolated from infants soon after vertical infection AIDS Res Hum Retroviruses 19 1115 1123 14709248
Janini M Rogers M Birx DR McCutchan FE 2001 Human immunodeficiency virus type 1 DNA sequences genetically damaged by hypermutation are often abundant in patient peripheral blood mononuclear cells and may be generated during near-simultaneous infection and activation of CD4(+) T cells J Virol 75 7973 7986 11483742
Vartanian JP Henry M Wain-Hobson S 2002 Sustained G→A hypermutation during reverse transcription of an entire human immunodeficiency virus type 1 strain Vau group O genome J Gen Virol 83 801 805 11907329
Wei M Xing H Hong K Huang H Tang H 2004 Biased G-to-A hypermutation in HIV-1 proviral DNA from a long-term non-progressor AIDS 18 1863 1865 15316355
Yamada T Iwamoto A 2000 Comparison of proviral accessory genes between long-term nonprogressors and progressors of human immunodeficiency virus type 1 infection Arch Virol 145 1021 1027 10881687
Zhang L Huang Y Yuan H Tuttleton S Ho DD 1997 Genetic characterization of vif, vpr, and vpu sequences from long-term survivors of human immunodeficiency virus type 1 infection Virology 228 340 349 9123841
Hassaine G Agostini I Candotti D Bessou G Caballero M 2000 Characterization of human immunodeficiency virus type 1 vif gene in long-term asymptomatic individuals Virology 276 169 180 11022005
Cao Y Qin L Zhang L Safrit J Ho DD 1995 Virologic and immunologic characterization of long-term survivors of human immunodeficiency virus type 1 infection N Engl J Med 332 201 208 7808485
Huang Y Zhang L Ho DD 1998 Characterization of gag and pol sequences from long-term survivors of human immunodeficiency virus type 1 infection Virology 240 36 49 9448687
Connor RI Sheridan KE Lai C Zhang L Ho DD 1996 Characterization of the functional properties of env genes from long-term survivors of human immunodeficiency virus type 1 infection J Virol 70 5306 5311 8764041
Zhang L Huang Y Yuan H Chen BK Ip J 1997 Genotypic and phenotypic characterization of long terminal repeat sequences from long-term survivors of human immunodeficiency virus type 1 infection J Virol 71 5608 5613 9188635
Huang Y Zhang L Ho DD 1995 Biological characterization of nef in long-term survivors of human immunodeficiency virus type 1 infection J Virol 69 8142 8146 7494338
Huang Y Zhang L Ho DD 1995 Characterization of nef sequences in long-term survivors of human immunodeficiency virus type 1 infection J Virol 69 93 100 7983771
Simon V Padte N Murray D Vanderhoeven J Wrin T 2003 Infectivity and replication capacity of drug-resistant human immunodeficiency virus type 1 variants isolated during primary infection J Virol 77 7736 7745 12829813
Adachi A Gendelman HE Koenig S Folks T Willey R 1986 Production of acquired immunodeficiency syndrome-associated retrovirus in human and nonhuman cells transfected with an infectious molecular clone J Virol 59 284 291 3016298
Esnault C Heidmann O Delebecque F Dewannieux M Ribet D 2005 APOBEC3G cytidine deaminase inhibits retrotransposition of endogenous retroviruses Nature 433 430 433 15674295
Sawyer SL Emerman M Malik HS 2004 Ancient adaptive evolution of the primate antiviral DNA-editing enzyme APOBEC3G PLoS Biol 2 e275 15269786
Kieffer TL Kwon P Nettles RE Han Y Ray SC 2005 G→A hypermutation in protease and reverse transcriptase regions of human immunodeficiency virus type 1 residing in resting CD4+ T cells in vivo J Virol 79 1975 1980 15650227
Yu Y Xiao Z Ehrlich ES Yu X Yu XF 2004 >Selective assembly of HIV-1 Vif-Cul5-ElonginB-ElonginC E3 ubiquitin ligase complex through a novel SOCS box and upstream cysteines Genes Dev 18 2867 2872 15574593
Mehle A Goncalves J Santa-Marta M McPike M Gabuzda D 2004 Phosphorylation of a novel SOCS-box regulates assembly of the HIV-1 Vif-Cul5 complex that promotes APOBEC3G degradation Genes Dev 18 2861 2866 15574592
Fujita M Sakurai A Yoshida A Miyaura M Koyama AH 2003 Amino acid residues 88 and 89 in the central hydrophilic region of human immunodeficiency virus type 1 Vif are critical for viral infectivity by enhancing the steady-state expression of Vif J Virol 77 1626 1632 12502880
Simon JH Sheehy AM Carpenter EA Fouchier RA Malim MH 1999 Mutational analysis of the human immunodeficiency virus type 1 Vif protein J Virol 73 2675 2681 10074113
Yu X Yu Y Liu B Luo K Kong W 2003 Induction of APOBEC3G ubiquitination and degradation by an HIV-1 Vif-Cul5-SCF complex Science 302 1056 1060 14564014
Zennou V Perez-Caballero D Gottlinger H Bieniasz PD 2004 APOBEC3G incorporation into human immunodeficiency virus type 1 particles J Virol 78 12058 12061 15479846
Ma XY Sova P Chao W Volsky DJ 1994 Cysteine residues in the Vif protein of human immunodeficiency virus type 1 are essential for viral infectivity J Virol 68 1714 1720 8107232
Rose PP Korber BT 2000 Detecting hypermutations in viral sequences with an emphasis on G→A hypermutation Bioinformatics 16 400 401 10869039
Hatziioannou T Cowan S Bieniasz PD 2004 Capsid-dependent and -independent postentry restriction of primate lentivirus tropism in rodent cells J Virol 78 1006 1011 14694132
Martin-Serrano J Zang T Bieniasz PD 2001 HIV-1 and Ebola virus encode small peptide motifs that recruit Tsg101 to sites of particle assembly to facilitate egress Nat Med 7 1313 1319 11726971
|
16201018
|
PMC1238741
|
CC BY
|
2021-01-05 12:11:27
|
no
|
PLoS Pathog. 2005 Sep 22; 1(1):e6
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010006
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 10.1371/journal.ppat.001000705-PLPA-RA-0029R2plpa-01-01-07Research ArticleImmunologyVirologyEukaryotesAnimalsMus (Mouse)Inhibition of MHC Class I Is a Virulence Factor in Herpes Simplex Virus Infection of Mice MHC Class I Inhibition by HSV-1Orr Mark T 1Edelmann Kurt H 1Vieira Jeffrey 23Corey Lawrence 234Raulet David H 5Wilson Christopher B 16*
1 Department of Immunology, University of Washington, Seattle, Washington, United States of America
2 Department of Laboratory Medicine, University of Washington, Seattle, Washington, United States of America
3 Program in Infectious Diseases, Fred Hutchinson Cancer Research Center, Seattle, Washington, United States of America
4 Department of Medicine, University of Washington, Seattle, Washington, United States of America
5 Department of Molecular and Cell Biology and Cancer Research Laboratory, University of California, Berkeley, California, United States of America
6 Department of Pediatrics, University of Washington, Seattle, Washington, United States of America
Ganem Donald EditorUniversity of California at San Francisco, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 1 e722 4 2005 20 7 2005 Copyright: © 2005 Orr et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Herpes simplex virus (HSV) has a number of genes devoted to immune evasion. One such gene, ICP47, binds to the transporter associated with antigen presentation (TAP) 1/2 thereby preventing transport of viral peptides into the endoplasmic reticulum, loading of peptides onto nascent major histocompatibility complex (MHC) class I molecules, and presentation of peptides to CD8 T cells. However, ICP47 binds poorly to murine TAP1/2 and so inhibits antigen presentation by MHC class I in mice much less efficiently than in humans, limiting the utility of murine models to address the importance of MHC class I inhibition in HSV immunopathogenesis. To address this limitation, we generated recombinant HSVs that efficiently inhibit antigen presentation by murine MHC class I. These recombinant viruses prevented cytotoxic T lymphocyte killing of infected cells in vitro, replicated to higher titers in the central nervous system, and induced paralysis more frequently than control HSV. This increase in virulence was due to inhibition of antigen presentation to CD8 T cells, since these differences were not evident in MHC class I-deficient mice or in mice in which CD8 T cells were depleted. Inhibition of MHC class I by the recombinant viruses did not impair the induction of the HSV-specific CD8 T-cell response, indicating that cross-presentation is the principal mechanism by which HSV-specific CD8 T cells are induced. This inhibition in turn facilitates greater viral entry, replication, and/or survival in the central nervous system, leading to an increased incidence of paralysis.
Synopsis
While animal models are often instructive in understanding human diseases, many factors that influence disease differ between mouse and man. Although mice can be experimentally infected with HSV-1, this virus has evolved as a human pathogen. One facet of this evolution is HSV's mechanisms to evade the immune response, allowing the virus to persist for the lifetime of the human host. This evasion includes preventing CD8 T cells from recognizing and killing infected cells by inhibiting the expression of the molecule that presents viral peptides to CD8 T cells: major histocompatibility complex (MHC) class I. HSV is unable to inhibit mouse MHC class I, thus rendering this immune-evasion strategy inoperative in the mouse. To better understand the biology of HSV infection and the immune response to this virus in humans, the authors corrected this deficiency by inserting a gene which inhibits murine MHC class I. This recombinant virus demonstrates that MHC class I inhibition is an important determinant of disease progression. The authors found that the recombinant HSV still effectively elicits a CD8 T-cell response, but this response is ineffective in controlling the infection. This finding reveals the important distinction between the size of the immune response and the effectiveness of the response, which may be important to HSV vaccine studies.
Citation:Orr MT, Edelmann KH, Vieira J, Corey L, Raulet DH, et al. (2005) Inhibition of MHC class I is a virulence factor in herpes simplex virus infection of mice. PLoS Pathog 1(1): e7.
==== Body
Introduction
Herpesviruses are distinguished by their ability to establish lifelong infection cycling between lytic and latent phases. One challenge to this lifestyle is that the immune system of the vertebrate hosts has the opportunity to be repeatedly primed, thereby increasing the potential for the host to eradicate the pathogen. To cope with this challenge, herpesviruses have evolved multiple mechanisms to evade immune detection or clearance. These mechanisms target all aspects of the immune response, including antibodies, chemokines, cytokines, natural killer (NK) cells, and CD4 and CD8 T cells [1,2].
Major histocompatibility complex (MHC) class I molecules are a particularly attractive target for immune evasion by viruses, because decreasing expression and/or antigen presentation by MHC class I can attenuate CD8 T-cell-mediated recognition of infected cells [3–5]. Inhibition of MHC class I antigen presentation is a hallmark of the herpesvirus family with all family members having at least one mechanism to achieve this. For example, the murine cytomegalovirus (MCMV) m152 gene product gp40 binds to the MHC class I/peptide complex in the ER/cis-Golgi compartment preventing export to the cell surface [6,7]. The human cytomegalovirus (HCMV) US11 gene product binds nascent MHC class I heavy chain in the endoplasmic reticulum and targets it to the cytosol for proteasomal degradation [8,9]. However, any strategy that lowers surface expression of MHC class I carries with it the undesirable (from the perspective of the virus) inverse effect of reducing the inhibitory signal that MHC class I exerts on NK cell activation. Some herpesviruses compensate for this decrease in MHC class I by expressing proteins that inhibit cell-surface expression of ligands that activate NK cells [10]. For example, MCMV m152 inhibits expression of ligands for the activating NK cell receptor NKG2D, while MCMV m157 binds to members of the Ly49 family of NK cell receptors, which include both inhibitory and activating receptors [11–13].
Herpes simplex virus (HSV) is an α-herpesvirus that establishes lifelong infection in neuronal cells from which it periodically reactivates [14]. Like the β-herpesviruses MCMV and HCMV, HSV inhibits antigen presentation on MHC class I to CD8 T cells, having evolved two distinct mechanisms by which to do so: the viral host shutoff protein (vhs) and the immediate early US12 gene product ICP47. Vhs targets host mRNA for destruction, thus nonspecifically shutting down antigen presentation at several steps [15,16]. ICP47 directly targets MHC class I antigen presentation by binding to the transporter associated with antigen presentation (TAP) 1/2 complex, preventing transport of peptides from the cytosol to the endoplasmic reticulum where peptides are loaded into the nascent MHC class I heavy-chain β2 microglobulin (β2m) complex [17–19]. However, genes that would compensate for the agonistic effect of reduced MHC class I expression on NK cell activation have not been identified in HSV.
Although HSV has no known murine homolog, HSV can infect mice in experimental models. In mice, as in humans, HSV spreads from peripheral tissues to the dorsal root ganglia in which it can establish latency or from which it can spread to the central nervous system (CNS) producing paralysis and death [20]. Consequently, murine models have been used extensively to study the pathogenesis and immunological control of HSV infection. One limitation of current murine models of HSV infection is that ICP47 poorly inhibits TAP in mouse cells [21]. This is due to ~100-fold decreased binding of ICP47 to murine TAP1/2 as compared to human TAP1/2 [18]. Consistent with this, ICP47 protects HSV-infected human fibroblasts from destruction by cytotoxic T lymphocytes (CTLs), while murine fibroblasts are not protected [22]. Despite the limited ability of ICP47 to inhibit murine TAP, a role for ICP47 in evasion of CD8 T-cell-mediated immunity in mice was suggested by studies in which CD8 T cells were able to protect mice from a mutant HSV lacking ICP47 but not from wild-type virus [23]. Given the difference in the capacity of ICP47 to inhibit peptide loading by murine versus human TAP, the contribution of CD8 CTL to HSV immunity in mice may overestimate their role in control of HSV in humans.
To develop an experimental system in which inhibition of MHC class I by HSV in mice would more closely parallel the situation in humans, we generated recombinant herpes simplex viruses (rHSVs) expressing MCMV m152 or HCMV US11. Both of these proteins inhibit antigen presentation by murine MHC class I [24,25]. We report here that these rHSVs prevented CTL recognition of infected cells in vitro, resulted in increased viral burden in the CNS, and increased the frequency of paralysis-induction compared to mice infected with control HSV. By contrast, these differences were not observed in MHC class I-deficient mice or in mice in which CD8 T cells were depleted. The generation of HSV-specific CD8 T cells was not affected, suggesting that the greater pathogenicity of these viruses resulted from evasion of CD8 T-cell recognition in the CNS, not impaired priming of the adaptive immune response.
Results
rHSVs Are Generated
We generated rHSVs expressing HCMV US11 (27US11), MCMV m152 (27m152), or a mock recombinant expressing only the gfp/gpt selection cassette (27gfp) as described in Materials and Methods and shown schematically in Figure 1A. Proper homologous recombination was confirmed by Southern blots for gpt, US11, m152, or the UL26–UL27 junction region. Insertion of the gfp selection cassette resulted in a 2.4-kB band shift compared to the parental KOS strain, while selection cassettes containing US11 or m152 gave band shifts of 3.8 kB or 4.2 kB, respectively (Figure 1B). These corresponded with the predicted band sizes indicating correct targeting (Figure 1B). The revertant virus generated from 27gfp (27gfpR) appeared identical to KOS. All viral genomic DNA hybridized with the HSV 26–27 probe as shown. Conversely, only 27gfp, 27US11, and 27m152 hybridized with a gpt probe, and only 27m152 and 27US11 hybridized with an m152 and US11 probe, respectively (data not shown). Thus all recombinant viruses contain the appropriate genes inserted into the UL26–UL27 junction region.
Figure 1 Generation of rHSVs
(A) rHSVs expressing MCMV m152 (27m152), HCMV US11 (27US11), or only the selection cassette (27gfp) were generated via homologous recombination with KOS strain HSV-1. Contents and location of insertions are indicated. Arrows indicate direction of transcription. The probe used to isolate correctly recombined viruses is indicated.
(B) Genomic DNA from KOS, 27US11, 27m152, 27gfp, and 27gfpR was digested with the indicated restriction enzymes and probed for the junction of HSV UL26–UL27. The predicted band sizes are indicated in the tabular inset.
27US11 and 27m152 Specifically Inhibit Murine MHC Class I
To determine whether 27m152 and 27US11 inhibit surface expression of murine MHC class I more efficiently than the control 27gfp, we analyzed expression on infected murine fibroblasts by flow cytometry. The control 27gfp demonstrated a modest reduction in surface expression of MHC class I 18 h after infection (Figure 2A and 2B), which was similar to the parental KOS strain (data not shown) and consistent with nonspecific vhs-mediated inhibition. Both 27US11 and 27m152 inhibited MHC class I surface expression to a greater extent than the 27gfp control. While 27m152 inhibited all tested murine MHC class I alleles, 27US11 inhibited Dd, Kb, and Db but not Kd, which is consistent with results reported by others [24].
Figure 2 27US11 and 27m152 Inhibit Murine MHC Class I Preventing Lysis by HSV-Specific CTL
(A) K-BALB (H-2d) and (B) MC57G (H-2b) fibroblast cell lines were uninfected (solid line) or infected (dashed line) at an MOI of 5:1 with 27gfp, 27US11, or 27m152 for 18 h and analyzed for surface MHC class I expression. Filled histograms are isotype controls. The percentage reduction in mean fluorescent intensity from uninfected to infected cells is indicated.
(C) and (D) Cells were infected as for (A) and (B) and co-incubated with CTL isolated from HSV-infected (C) BALB/c or (D) BALB.B mice at the indicated effector-to-target ratio.
Owing to the slow turnover of MHC class I on the cell surface, total MHC class I expression will underestimate the impact of US11 and m152 on presentation of viral peptides on newly synthesized MHC class I to CD8 T cells. To test directly the effects of rHSVs on recognition of infected targets, we examined the lysis of infected fibroblasts by CTL in vitro. HSV-specific CTL, which efficiently lysed 27gfp-infected fibroblasts, lysed fibroblasts infected with 27US11 or 27m152 less effectively (Figure 2C and 2D). Although, as determined by flow cytometry, overall inhibition of cell-surface MHC class I was greater in the H-2b (Figure 2B) than in the H-2d cells (Figure 2A), the decrease in lysis efficiency of cells infected with 27US11 and 27m152 compared to 27gfp (and KOS, data not shown) was observed both with infected H-2d (Figure 2C) and H-2b (Figure 2D) targets. These data demonstrate that 27US11 and 27m152 have a gain-of-function resulting in increased inhibition of MHC class I antigen presentation and inhibition of CTL-mediated lysis of infected murine cells.
27m152, but Not 27US11, Inhibits NKG2D Ligands, Preventing NK-Cell-Mediated Lysis
Because MHC class I molecules are ligands for inhibitory receptors on NK cells, inhibition of MHC class I surface expression would be predicted to render cells infected with 27US11 and 27m152 more vulnerable to NK-cell-mediated clearance [26]. However, MCMV m152 also inhibits expression of the Rae-1 family of ligands for the NKG2D-activating receptor on NK cells; thus 27m152 should also antagonize NK cell recognition [12,13]. By contrast, HCMV US11 is not known to inhibit the expression of NKG2D ligands, and therefore cells infected with 27US11 should be more vulnerable to lysis by NK cells. Consistent with these predictions, 27m152, but not 27US11, inhibited the expression of NKG2D ligands on infected fibroblasts (Figure 3A), and 27US11-infected fibroblasts were more readily lysed by NK cells than 27gfp-infected fibroblasts. Conversely, lysis of 27m152-infected cells was similar to lysis of cells infected with the control 27gfp virus (Figure 3B). Since the only reported function of HSV ICP47 is to block MHC class I antigen presentation, 27US11 appears to more closely parallel in mice the immune-evasion profile of HSV in humans.
Figure 3 27m152, but Not 27US11, Evades NK Recognition by Inhibiting NKG2D Ligands
(A) BALB/c fibroblasts were uninfected (solid line) or infected (dashed line) for 18 h at an MOI of 5:1 with 27gfp, 27US11, or 27m152 (dashed line) and stained with NKG2D tetramer. Solid histogram is uninfected cells stained with an irrelevant tetramer. The percentage reduction in mean fluorescent intensity from uninfected to infected cells is indicated.
(B) Cells were prepared as in (A) and co-incubated with splenocytes from naïve RAG1−/− BALB/c mice treated with polyI:C 24 h earlier at the indicated effector-to-target ratio.
Recombinant Viruses Grow as Well as Wild-Type in Vitro
All three gene products flanking the insertion area (UL26, UL26.5, and UL27) are required for in vitro growth [27]. To confirm that recombination did not alter neighboring gene products, we analyzed the single-step growth kinetics of each virus. Growth curves over a 24-h period revealed that each recombinant virus grew at the same rate as the parent virus (Figure 4A and 4B). Thus, growth of recombinant viruses is not impaired, demonstrating that no genes essential to in vitro growth, including UL26, UL26.5, and UL27, were altered. This finding was confirmed by quantitative RT-PCR for UL26, UL26.5, and UL27 mRNA. Expression for each gene was similar for all three recombinant viruses and for KOS virus (data not shown); expression of GFP and GPT was also similar for all three recombinant viruses.
Figure 4 Single-Step Growth Kinetics of rHSVs Are Similar to KOS
Vero cells were infected at an MOI of 5:1 with KOS plus (A) 27m152 or 27US11, or (B) 27gfp. Cells and supernatants were harvested at indicated times and viral titers were determined on vero cells.
The gfp-Containing Selection Cassette Attenuates Recombinant Viruses in Vivo
Recently, Halford et al. reported a KOS-strain rHSV expressing gfp driven by the HCMV immediate early promoter inserted between UL26 and UL27 [28]. This virus demonstrated a 50% increase in time-to-death after ocular infection of scid mice of C57Bl/6 or BALB/c background. To determine whether the gfp/gpt selection cassette used to generate 27gfp, 27US11, and 27m152 resulted in a similar attenuation in vivo, we compared the neuroinvasiveness and neurovirulence of 27gfp to the parental KOS strain and the revertant 27gfpR. Six days post-infection, viral titers from the footpad, dorsal root ganglia, and spinal cord were similar between BALB/c mice infected with KOS or 27gfpR. Although the viral burden in the footpads of 27gfp-infected mice was similar to KOS- or 27gfpR-infected mice, there was a ~10-fold reduction of virus in the dorsal root ganglia and ~100-fold reduction of virus in the spinal cord (Figure 5A). Thus, insertion of the selection cassette reduced neuroinvasiveness, while removal of the cassette restored it. This decrease in neuroinvasiveness correlated with decreased neurovirulence, as 80% of mice infected with KOS or 27gfpR succumbed to paralysis by day 10, while all mice infected with 27gfp retained full mobility (Figure 5B). As the revertant is identical to the parental virus in both neuroinvasiveness and neurovirulence, and 27US11 and 27m152 contain this selection cassette, we used 27gfp as a control to determine the effects of MHC class I inhibition by 27US11 or 27m152 in vivo.
Figure 5 The Selection Cassette Attenuates Neuroinvasiveness and Neurovirulence
BALB/c mice were infected in the hind footpads with 2.5 × 105 pfu of KOS, 27gfp, or 27gfpR.
(A) The indicated tissues were isolated on day 6, homogenized, and viral titers were determined on vero cells.
(B) Ten (27gfp) or 20 (KOS and 27gfpR) mice per virus were monitored for paralysis induction for 14 d. Mice displaying ataxia or paralysis were euthanized.
Inhibition of MHC Class I Is a Virulence Factor
There was no difference in the viral burden in the dorsal root ganglia or hind footpad of BALB/c mice inoculated with 27gfp, 27US11, or 27m152. However ~100-fold more virus was recovered from the spinal cord of mice infected with 27US11 or 27m152 compared to 27gfp (Figure 6A). Consistent with this increase in neuroinvasiveness, 27US11 and 27m152 induced paralysis in 70% of mice, while 27gfp did not induce paralysis with this inoculum (Figure 6B). Similar results were obtained in BALB.B mice (data not shown).
Figure 6 Inhibition of MHC Class I Increases Neuroinvasiveness and Neurovirulence
BALB/c were infected in the hind footpads with 2.5 × 105 pfu of 27gfp, 27US11, or 27m152.
(A) The indicated tissues were isolated on day 6, homogenized, and viral titers were determined on vero cells.
(B) Ten mice per virus were monitored for paralysis induction for 14 d. Mice displaying ataxia or paralysis were euthanized.
Increased Neurovirulence Is Due to Inhibition of MHC Class I Antigen Presentation to CD8 T Cells
If these differences in neuroinvasiveness and neurovirulence are solely due to altered antigen presentation to CD8 T cells by MHC class I on infected cells, then 27gfp should be as neurovirulent as 27m152 or 27US11 in mice lacking MHC class I expression and in mice depleted of CD8 T cells. Consistent with this prediction and in sharp contrast to findings in wild-type mice (Figure 6), titers of 27gfp in the footpad, dorsal root ganglia, and spinal cord of β2m−/− mice at day 6 were equivalent to those for 27US11 (Figure 7A), and the frequency of paralysis in MHC class I-deficient β2m−/− mice infected with 27gfp, 27US11, or 27m152 was similar (Figure 7B). Moreover, the titers of 27gfp in the footpad, dorsal root ganglia, and spinal cord of wild-type mice depleted of CD8 T cells were equivalent to those for 27US11 and 27m152 on day 6 (Figure 7C). Similar findings were obtained at earlier time points—viral burdens in the spinal cord on day 4 were already significantly higher in mice infected with 27US11 compared to mice infected with 27gfp, and this difference was also abolished by CD8 depletion (data not shown). Conversely, depletion of CD4 T cells did not abolish the increase in viral load in the spinal cord of 27US11- or 27m152-infected BALB/c, as compared to 27gfp-infected mice (Figure S1). These findings indicate that HSV-specific CD8 T cells are controlling 27gfp, but not 27US11 or 27m152, and are consistent with the notion that the increased neurovirulence and neuroinvasiveness of 27US11 and 27m152 rHSV are attributable to MHC class I inhibition and evasion of CD8 T cells.
Figure 7 Differences in Neuroinvasiveness and Neurovirulence Are Dependent on MHC Class I and CD8 T Cells
β2m−/− BALB/c mice were infected in the hind footpads with 3.0 × 104 pfu of 27gfp, 27US11, or 27m152. Note that a lower inoculum was used in these experiments with β2m−/− mice than in wild-type BALB/c mice shown in other figures.
(A) The indicated tissues were isolated on day 6, homogenized, and viral titers were determined on vero cells.
(B) Nine mice per virus were monitored for paralysis induction for 14 d. Mice displaying ataxia or paralysis were euthanized.
(C) BALB/c mice were depleted of CD8 cells and infected with 2.5 × 105 pfu of the indicated virus in the hind footpads. Viral titers were determined on day 6.
MHC Class I Inhibition Does Not Affect the Numbers of HSV-Specific CD8 T Cells
The increased neurovirulence of 27US11 and 27m152 could result from impaired generation of HSV-specific CD8 T cells, impaired recognition of infected cells in neural tissues, or both. Recent work suggests that the generation of HSV-specific CD8 T cells in mice infected with wild-type HSV relies on cross-presentation of viral antigens by dendritic cells [29]. If this is the case, increased MHC class I inhibition by 27US11 and 27m152 should not affect the induction of HSV-specific CD8 T cells.
To address this, we infected BALB.B mice, which are congenic with BALB/c mice, and similar in their susceptibility to HSV, but of the H-2b haplotype. Since ~90% of HSV-specific CD8 T cells in H-2b mice recognize the peptide gB498–505 presented by H-2Kb, BALB.B mice allow accurate enumeration of antigen-specific CD8 T cells [30]. At the peak of infection, 27US11, 27m152, and 27gfp induced similar numbers of gB-specific IFN-γ-producing CD8 T cells (Figure 8). Thus, the difference in neuroinvasiveness and neurovirulence between these viruses was not due to altered generation of gB-specific CD8 T cells.
Figure 8 The Size of the CD8 T-Cell Response to rHSVs Is Not Altered by MHC Class I Inhibition
Lymphocytes from the draining popliteal lymph nodes were isolated from BALB.B mice on day 6 of infection with 27gfp, 27US11, or 27m152. Lymphocytes were stimulated with HSV gB498–505 for 5 h then stained for CD8 and intracellular IFN-γ. Of the unstimulated CD8+ cells, <0.1% were IFN-γ+.
Discussion
In this report, we show that two different rHSVs, 27US11 and 27m152, efficiently inhibited antigen presentation by MHC class I molecules on murine cells, as does wild-type HSV on human cells. BALB/c and BALB.B mice infected with either of these two rHSVs showed an increased incidence of paralysis induction in vivo compared to mice infected with the control 27gfp virus. Paralysis induction correlated with higher viral burden in the CNS, but not in the footpad or the peripheral nervous system. This increase in neurovirulence occurred despite the presence of a strong antigen-specific CD8 T-cell response, the size of which was not diminished in mice infected with 27US11 or 27m152 compared to 27gfp. By contrast, increased neurovirulence of 27US11 and 27m152 did correlate with the reduced sensitivity of target cells infected with these viruses to CD8 CTL lysis in vitro. The greater neurovirulence of 27US11 and 27m152 was not observed in MHC class I-deficient mice or in wild-type mice depleted of CD8 T cells. Taken together, these data indicate that inhibition of MHC class I antigen presentation by HSV is a neurovirulence factor, and that the primary mechanism for this increased virulence was the inhibition of target-cell recognition by antigen-specific CD8 T cells.
A previous study addressed the importance of ICP47-mediated inhibition of MHC class I antigen presentation by HSV in mice [23]. When HSV-susceptible A/J or BALB/c mice were challenged by ocular inoculation with a mutant HSV lacking ICP47, this strain induced a lower incidence of neurologic symptoms and death than the parental strain F. This difference was ablated when mice were depleted of CD8 cells and in mice that lacked CD8 T cells. This study suggested that although ICP47 has a much reduced impact on TAP-dependent antigenic peptide transport in murine cells versus human cells, this may be sufficient for functional inhibition in mice in vivo. However, the relatively weak effect of ICP47 in murine cells may underestimate its importance in viral pathogenesis in humans [21,22]. Using rHSVs that more closely approximate in mice the magnitude of MHC class I inhibition by wild-type HSV in humans, the impact of MHC class I inhibition on neurological outcome was clear—the titers of 27US11 and 27m152 were more than 100-fold higher, and the frequency of paralysis was significantly greater than 27gfp.
In the previous study, no attempt was made to determine whether the difference in neurological outcome was associated with differences in viral titers in peripheral tissues or the CNS or with differences in the generation of antigen-specific CD8 T cells [23]. We found that, despite robust inhibition of antigen presentation by murine MHC class I, the only significant difference in viral titers was found in the CNS, suggesting a focused role for this evasion strategy. This focused immune evasion may result from the greater impact of inhibition of MHC class I antigen presentation on cells that express low levels of MHC class I, such as neurons during acute infection, making the infected CNS more susceptible to this immune-evasion strategy [31].
Inhibition of antigen presentation on MHC class I may impact CD8 T-cell response to HSV infection at either or both of two distinct phases. First, inhibition could limit the size of the antigen-specific CD8 T-cell response. However, we found that the frequency of HSV-specific CD8 T cells was not diminished in response to infection with rHSVs that effectively inhibited MHC class I antigen presentation. Our findings provide strong support for the notion that CD8 T-cell priming in mice infected with HSV is carried out by cross-priming, as proposed by others from studies with wild-type HSV [29,32,33]. The second phase at which inhibition may impact outcome is at the site of productive infection, in the peripheral tissues, peripheral nervous system, or CNS. The selective increase in viral titers in the CNS in mice infected with 27US11 and 27m152 in the absence of differences in the magnitude of the CD8 T-cell response is compatible with the notion that the primary effect of MHC class I inhibition by HSV is to prevent recognition of infected cells by virus-specific CD8 T cells in the CNS. Thus, generation of an immune response does not always predict the functional relevance of that response. This distinction is important to the evaluation of vaccines targeted at pathogens that specialize in immune evasion such as herpesviruses and poxviruses.
Although depletion of CD8 T cells had a profound impact on the amount of 27gfp in the spinal cord, there was little difference in titers of any of the rHSVs in the footpad and dorsal root ganglia between intact and depleted mice (see Figures 6A and 7C). This result suggests that the primary impact of CD8 T-cell immunity to HSV is in the CNS. That depletion of CD8 T cells in mice infected with 27US11 or 27m152 did not affect CNS viral titers indicates that the effect of MHC class I inhibition is also manifest primarily at this site.
The effect of m152 expression in HSV on viral titer is different from the effect of m152 in MCMV. As reported by others, deletion of m152, together with the other MHC class I inhibitors expressed by MCMV, m04 and m06, has no effect on viral titers in the lung [34]. This difference may be due to the different tropisms of HSV and MCMV. Whereas MCMV infects cells with moderate-to-high expression of MHC class I, HSV targets neuronal cells which express very little MHC class I [31], and the impact of MHC class I inhibition would be more apparent in cells that normally express low levels of MHC class I. This difference may explain the significant increase in viral titer in the CNS of 27m152- and 27US11-infected mice, while lung titers of wild-type and Δ04+Δ06+Δ152 MCMV are similar.
While the outcome of infection with 27US11 and 27m152 viruses in vivo was similar, and these viruses inhibited CD8 T-cell-mediated killing of infected cells in vitro to a similar degree, cells infected with 27US11 but not 27m152 were more susceptible to killing by NK cells in vitro. Together, these findings suggest that evasion of NK cells does not substantially impact the outcome of acute HSV infection when CD8 T-cell recognition is impaired in mice of the susceptible BALB background. Furthermore, depletion of NK cells with anti-asialoGM1 antiserum did not abrogate the difference in viral titers in the CNS in mice infected with 27US11 and 27m152 compared to 27gfp—on day 3, titers of 27US11 and 27m152 were similar, and both were significantly greater than the titer of 27gfp (data not shown). It is possible that this finding may be mouse-strain-specific, as is the case for m152-mediated NK cell evasion for MCMV [13].
Another limitation of traditional murine models of HSV infection is the lack of spontaneous reactivation in vivo, which is a hallmark of human infection. This may reflect the fact that latently infected mice maintain HSV-specific CD8 T cells in the infected ganglia, which can prevent reactivation of wild-type HSV in vitro [35]. Given the greater efficiency with which they evade CD8 recognition, it is possible that 27US11 and 27m152 may display altered latency features in mice compared to wild-type HSV.
Materials and Methods
Cell lines and mice.
Vero cells were used for isolation of recombinant viruses. Viral stocks were prepared from infected vero cells at 90%–100% cytopathic effect. Stocks were sonicated, stored at −80 °C, and titered on vero cells. Plaque formation was visualized with crystal violet stain in 10% formaldehyde. Revertant virus was isolated from STO cells (ATCC, Manassas, Virginia, United States) that lack HPRT making them resistant to 6-thioguanine toxicity. Murine fibroblast cell lines (ATCC) MC57G from C57Bl/6 and K-BALB from BALB/c mice were used to assess MHC class I expression and CD8 T-cell- and NK-cell-mediated lysis.
Female BALB/c and Rag1−/− (BALB/c background) were purchased from Jackson Laboratory (Bar Harbor, Maine, United States) and used at 6–8 wk of age. BALB.B and β2m−/− (BALB/c background) breeder pairs were purchased from Jackson Laboratory and bred in-house. All mice were maintained in the University of Washington SPF facility. All studies were approved by the University of Washington Animal Care and Use Committee.
Generation of rHSV.
To determine the effect of MHC class I inhibition on HSV infection in mice, we generated rHSVs that express MCMV m152 or HCMV US11, denoted 27m152 and 27US11, respectively (see Figure 1). Additionally, a control virus termed 27gfp, which expresses the selection cassette used to isolate 27m152 and 27US11, was generated. A 2.5-kB EcoR I–Hind III fragment in the UL26–UL27 region of the HSV genome was isolated and cloned into pUC19. A unique Not I site in the non-coding region that separates the UL26 and UL27 open reading frames was mutated to a Spe I site and used for generation of the targeting vector. A selection cassette containing eGfp driven by the EF-1 promoter and E. coli guanosylphosoribosyltransferase (gpt) driven by the PGK-1 promoter was cloned into the Spe I site. The original Not I site is located in the polyA signal of UL26. To correct for this, the bi-directional SV40 polyA was also included. This targeting vector was used to generate 27gfp via homologous recombination.
The targeting vector was digested with EcoR I and Hind III, and 10 μg of DNA was electroporated into vero cells using a Bio-Rad (Hercules, California, United States) GenePulser Xcell (273 V, 1,100 μF, 186 Ω, and 0.4-mm cuvette). Electroporated cells were plated in six well plates and infected with 2.5 × 105 pfu KOS strain HSV-1 24 h after electroporation.
Similarly, 27US11 and 27m152 were generated by inserting HCMV US11 or MCMV m152 into the gfp/gpt-targeting vector under the control of the HCMV IE promoter. MCMV m152 was PCR-amplified from viral genomic DNA, while HCMV US11 was isolated from an expression vector (provided by Stan Riddell, Fred Hutchinson Cancer Research Center). Recombinant viruses were plaque-purified based on gfp expression, passed once through BALB/c mice, and re-isolated from the spinal cord to ensure neuroinvasiveness. A revertant virus for 27gfp, termed 27gfpR, was generated via homologous recombinant in STO cells and isolated by loss of gfp expression. Isolation of the revertant virus was supported by negative selection in the presence of 40 μg/ml 6-thioguanine, which is converted to the toxic 6-thioxanthine by GPT [36].
MHC class I and NKG2D surface expression.
MC57G (H-2b) or K-BALB (H-2d) fibroblast cell lines were infected at a multiplicity of infection (MOI) of 5:1 for 18 h in the presence of 200 μM gancyclovir. Cells were resuspended in 100 μl of phosphate-buffered saline + 1% bovine albumin and 0.09% sodium azide with α-CD16/32 (1:100) as an Fc-blocking reagent and phycoerythrin-conjugated α-H-2Kb (1:100), α-H-2Db (1:25), α-H-2Kd (1:100), α-H-2Dd (1:100), muIgG2aκ and muIgG2bκ isotype controls, NKG2D tetramer (1:1000), or irrelevant tetramer (1:1,000) for 30 min on ice. Phycoerythrin-conjugated tetramers were produced as previously described [37]. Analysis was performed with a Becton Dickenson (Palo Alto, California, United States) FACSscan. Antibodies were purchased from BD Biosciences Pharmingen (San Diego, California, United States).
CTL and NK lysis.
MC57G or K-BALB cells were infected at an MOI of 5:1 for 11 h (for CTL assays) or for 8 h (for NK cell assays) in the presence of 200 μM gancyclovir. Cells were resuspended in 300 μl of warm media with 30 μl of fresh 51Cr (PerkinElmer, las.perkinelmer.com). Cells were incubated at 37 °C for two 1-h periods and washed twice with warm media. Co-incubation of 105 cells/well took place for 5 h (CTL) or 8 h (NK) with effector cells. Eighteen hours after infection, 100 μl of media was collected and analyzed on a Wallac 1470 Wizard gamma counter (PerkinElmer). CD8 CTLs were derived from the draining lymph node of day-6 HSV-infected BALB.B or BALB/c mice. For 6 d, 106 lymphocytes were cultured in DMEM (Gibco, San Diego, California, United States) plus 10% FBS with 1:100 anti-CD3 and 25 μg/ml recombinant huIL-2. Media was changed on days 3 and 5 with fresh huIL-2. These CTL were added to target cells in graded numbers. Activated NK cells were derived from splenocytes from Rag1−/− BALB/c mice injected i.p. with 200 μg polyI:C 24 h prior to sacrifice. Total splenocytes were added to target cells in graded numbers. Specific lysis was determined: percentage specific lysis = (count − minimum)/(total lysis − minimum lysis) × 100.
Single-step growth in vitro.
Vero cells were infected at an MOI of 5:1 for 1 h at 4 °C to allow viral attachment. Cells were then washed thrice with cold PBS and warm RPMI media with 10% FBS added (t = 0). Cells were incubated at 37 °C for 1 h. Media was then removed and cells were quickly washed twice with sodium citrate buffer (pH 3.0) and rinsed thrice with warm PBS. Warm media was then replaced and cells were returned to 37 °C. At indicated time points, media from three wells per virus was collected; cells were then trypsinized and mixed with the corresponding media. Samples were stored at −80 °C and titered on vero cells.
Infection of mice.
Mice were infected in the hind footpads with the indicated inocula following dermal abrasion, as described previously [38]. In this model, virus travels anterograde up the enervating sciatic nerve to the dorsal root ganglia, replicates in the ganglion, and can then return to the site of infection via retrograde axonal transport resulting in a primary lesion of the footpad. Virus in the dorsal root ganglia can also bridge the synapse and enter the CNS at the spinal cord, from which it may ascend towards the brain. Infected mice were monitored twice daily for 14 d for ataxia and hind-limb paralysis. Previous findings indicated that more than 80% of mice displaying paralysis succumb to infection; thus paralyzed mice were euthanized in accordance with our IACUC protocol.
In vivo viral titers.
Infected mice were euthanized on day 6 of infection. Hind footpads, dorsal root ganglia with proximal sciatic nerve, and spinal cord were isolated and snap frozen on dry ice. Samples were stored at −80 °C. All samples were homogenized and titered in triplicate on vero cells.
Antibody depletion.
Mice were depleted of CD4 or CD8 T cells by i.p. injection of 200 μl of 1 mg/ml anti-CD4 (GK1.5) or anti-CD8 (clone 2.43) on two consecutive days, and were infected 2 d later. Mice were depleted of NK cells by i.p. injection of 100 μl of α-asialo-GM1 (Wako Biochemical, http://www.wako-chemicals.de) 1 d before infection.
Quantification of CD8 T-cell response.
BALB.B mice were infected with 2.5 × 105 pfu of HSV and sacrificed on day 6. Single-cell suspensions were prepared from their draining popliteal lymph nodes. For IFN-γ production, cells were stimulated with 5 μM of the immunodominant HSV peptide glycoprotein B498–505 (gB498–505) (United Biochemical Research, Seattle, Washington, United States) for 5 h in the presence of BD GolgiStop, followed by surface staining with anti-CD8-FITC (1:100), permeabilized as above and stained for intracellular IFN-γ with anti-IFN-γ-PE (1:200). Unstimulated lymphocytes were used as a negative control. Specific IFN-γ production = percentage CD8+ IFN-γ+ stimulated − unstimulated. All antibodies were purchased from BD Biosciences Pharmingen.
Supporting Information
Figure S1 Depletion of CD4 T Cells Does Not Equalize Spinal Cord Titers of 27 gfp Compared to 27US11 or 27m152, Whereas Depletion of CD8 T Cells Does
BALB/c mice were either (A) untreated; or (B) and (C) treated 3 and 2 days before the time of infection with 200 μg α-CD4 MAb (GK1.5) or α-CD8 MAb (53–6.7), respectively. Mice were infected with 2.5 × 105 pfu of 27 gfp, 27US11, or 27m152 on day 0, and footpads and spinal cord were harvested on day 6. This experiment was distinct from the experiments shown in Figures 6 and 7, but the findings for A and C are similar to those shown in Figures 6A and 7C, respectively.
(357 KB EPS)
Click here for additional data file.
Accession Numbers
The Swiss-Prot (http://www.ebi.ac.uk/swissprot) accession number for the MCMV m152 gene product is Q69G18, and for the HCMV US11 gene product is P09727. The Swiss-Prot accession number for vhs is Q69G18, and for the US12 gene product ICP47 is P03170. Swiss-Prot accession numbers for UL26, UL26.5, and UL27 are P10210 and P10211.
We thank Heidi Harowicz and Brooke Fallen for excellent animal husbandry and technical assistance. This work was supported in part by grant HD18184 (to CBW) and NIH grant R37 AI-42528 (to LC). MTO is supported by a predoctoral fellowship from the Howard Hughes Medical Institute.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MTO, KHE, JV, LC, DHR, and CBW conceived and designed the experiments. MTO and KHE performed the experiments. MTO and CBW analyzed the data. MTO, KHE, JV, LC, and DHR contributed reagents, materials, and analysis tools. MTO and CBW wrote the paper.
Abbreviations
CNScentral nervous system
CTLcytotoxic T lymphocyte
HCMVhuman cytomegalovirus
HSVherpes simplex virus
MCMVmurine cytomegalovirus
MHCmajor histocompatibility complex
MOImultiplicity of infection
NKnatural killer
rHSVrecombinant herpes simplex virus
TAPtransporter associated with antigen presentation
==== Refs
References
Alcami A Koszinowski UH 2000 Viral mechanisms of immune evasion Immunol Today 21 447 455 10953097
Tortorella D Gewurz BE Furman MH Schust DJ Ploegh HL 2000 Viral subversion of the immune system Annu Rev Immunol 18 861 926 10837078
Fruh K Gruhler A Krishna RM Schoenhals GJ 1999 A comparison of viral immune escape strategies targeting the MHC class I assembly pathway Immunol Rev 168 157 166 10399072
Fruh K Ahn K Peterson PA 1997 Inhibition of MHC class I antigen presentation by viral proteins J Mol Med 75 18 27 9020380
Yewdell JW Hill AB 2002 Viral interference with antigen presentation Nat Immunol 3 1019 1025 12407410
Ziegler H Thale R Lucin P Muranyi W Flohr T 1997 A mouse cytomegalovirus glycoprotein retains MHC class I complexes in the ERGIC/cis -Golgi compartments Immunity 6 57 66 9052837
Ziegler H Muranyi W Burgert HG Kremmer E Koszinowski UH 2000 The luminal part of the murine cytomegalovirus glycoprotein gp40 catalyzes the retention of MHC class I molecules EMBO J 19 870 881 10698929
Jones TR Hanson LK Sun L Slater JS Stenberg RM 1995 Multiple independent loci within the human cytomegalovirus unique short region down-regulate expression of major histocompatibility complex class I heavy chains J Virol 69 4830 4841 7609050
Wiertz EJ Jones TR Sun L Bogyo M Geuze HJ 1996 The human cytomegalovirus US11 gene product dislocates MHC class I heavy chains from the endoplasmic reticulum to the cytosol Cell 84 769 779 8625414
Orange JS Fassett MS Koopman LA Boyson JE Strominger JL 2002 Viral evasion of natural killer cells Nat Immunol 3 1006 1012 12407408
Arase H Mocarski ES Campbell AE Hill AB Lanier LL 2002 Direct recognition of cytomegalovirus by activating and inhibitory NK cell receptors Science 296 1323 1326 11950999
Lodoen M Ogasawara K Hamerman JA Arase H Houchins JP 2003 NKG2D-mediated natural killer cell protection against cytomegalovirus is impaired by viral gp40 modulation of retinoic acid early inducible 1 gene molecules J Exp Med 197 1245 1253 12756263
Krmpotic A Busch DH Bubic I Gebhardt F Hengel H 2002 MCMV glycoprotein gp40 confers virus resistance to CD8+ T cells and NK cells in vivo Nat Immunol 3 529 535 12021778
Koelle DM Corey L 2003 Recent progress in herpes simplex virus immunobiology and vaccine research Clin Microbiol Rev 16 96 113 12525427
Everly DN Jr Feng P Mian IS Read GS 2002 mRNA degradation by the virion host shutoff (Vhs) protein of herpes simplex virus: Genetic and biochemical evidence that Vhs is a nuclease J Virol 76 8560 8571 12163576
Kwong AD Frenkel N 1989 The herpes simplex virus virion host shutoff function J Virol 63 4834 4839 2552156
York IA Roop C Andrews DW Riddell SR Graham FL 1994 A cytosolic herpes simplex virus protein inhibits antigen presentation to CD8+ T lymphocytes Cell 77 525 535 8187174
Ahn K Meyer TH Uebel S Sempe P Djaballah H 1996 Molecular mechanism and species specificity of TAP inhibition by herpes simplex virus ICP47 EMBO J 15 3247 3255 8670825
Hill A Jugovic P York I Russ G Bennink J 1995 Herpes simplex virus turns off the TAP to evade host immunity Nature 375 411 415 7760935
Cook ML Stevens JG 1973 Pathogenesis of herpetic neuritis and ganglionitis in mice: Evidence for intra-axonal transport of infection Infect Immun 7 272 288 4348966
Tomazin R van Schoot NE Goldsmith K Jugovic P Sempe P 1998 Herpes simplex virus type 2 ICP47 inhibits human TAP but not mouse TAP J Virol 72 2560 2563 9499125
Jugovic P Hill AM Tomazin R Ploegh H Johnson DC 1998 Inhibition of major histocompatibility complex class I antigen presentation in pig and primate cells by herpes simplex virus type 1 and 2 ICP47 J Virol 72 5076 5084 9573278
Goldsmith K Chen W Johnson DC Hendricks RL 1998 Infected cell protein (ICP)47 enhances herpes simplex virus neurovirulence by blocking the CD8+ T cell response J Exp Med 187 341 348 9449714
Machold RP Wiertz EJ Jones TR Ploegh HL 1997 The HCMV gene products US11 and US2 differ in their ability to attack allelic forms of murine major histocompatibility complex (MHC) class I heavy chains J Exp Med 185 363 366 9016885
Wagner M Gutermann A Podlech J Reddehase MJ Koszinowski UH 2002 Major histocompatibility complex class I allele-specific cooperative and competitive interactions between immune evasion proteins of cytomegalovirus J Exp Med 196 805 816 12235213
Diefenbach A Raulet DH 2001 Strategies for target cell recognition by natural killer cells Immunol Rev 181 170 184 11513138
Roizman B Sears A 1996 Herpes simplex viruses and their replication Fields BN Fields virology Philadelphia Lippincott-Raven 2231 2295
Halford WP Balliet JW Gebhardt BM 2004 Re-evaluating natural resistance to herpes simplex virus type 1 J Virol 78 10086 10095 15331741
Mueller SN Jones CM Smith CM Heath WR Carbone FR 2002 Rapid cytotoxic T lymphocyte activation occurs in the draining lymph nodes after cutaneous herpes simplex virus infection as a result of early antigen presentation and not the presence of virus J Exp Med 195 651 656 11877488
Wallace ME Keating R Heath WR Carbone FR 1999 The cytotoxic T-cell response to herpes simplex virus type 1 infection of C57BL/6 mice is almost entirely directed against a single immunodominant determinant J Virol 73 7619 7626 10438852
Pereira RA Simmons A 1999 Cell surface expression of H2 antigens on primary sensory neurons in response to acute but not latent herpes simplex virus infection in vivo J Virol 73 6484 6489 10400743
Allan RS Smith CM Belz GT van Lint AL Wakim LM 2003 Epidermal viral immunity induced by CD8alpha+ dendritic cells but not by Langerhans cells Science 301 1925 1928 14512632
Smith CM Belz GT Wilson NS Villadangos JA Shortman K 2003 Cutting edge: Conventional CD8 alpha+ dendritic cells are preferentially involved in CTL priming after footpad infection with herpes simplex virus-1 J Immunol 170 4437 4440 12707318
Gold MC Munks MW Wagner M McMahon CW Kelly A 2004 Murine cytomegalovirus interference with antigen presentation has little effect on the size or the effector memory phenotype of the CD8 T cell response J Immunol 172 6944 6953 15153514
Khanna KM Bonneau RH Kinchington PR Hendricks RL 2003 Herpes simplex virus-specific memory CD8+ T cells are selectively activated and retained in latently infected sensory ganglia Immunity 18 593 603 12753737
Isaacs SN Kotwal GJ Moss B 1990 Reverse guanine phosphoribosyltransferase selection of recombinant vaccinia viruses Virology 178 626 630 2219714
Diefenbach A Jamieson AM Liu SD Shastri N Raulet DH 2000 Ligands for the murine NKG2D receptor: Expression by tumor cells and activation of NK cells and macrophages Nat Immunol 1 119 126 11248803
Holterman AX Rogers K Edelmann K Koelle DM Corey L 1999 An important role for major histocompatibility complex class I-restricted T cells, and a limited role for gamma interferon, in protection of mice against lethal herpes simplex virus infection J Virol 73 2058 2063 9971787
|
16201019
|
PMC1238742
|
CC BY
|
2021-01-05 12:11:27
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e7
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010007
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 1620102010.1371/journal.ppat.001000805-PLPA-RA-0012R3plpa-01-01-11Research ArticleEcologyEpidemiology - Public HealthGastroenterology - HepatologyInfectious DiseasesMicrobiologyGenetics/Genetics of DiseaseGenetics/Disease ModelsEubacteriaArthropodsInsectsDrosophila
Vibrio cholerae Infection of Drosophila
melanogaster Mimics the Human Disease Cholera V. cholerae Infects Flies
Blow Nathan S 1Salomon Robert N 2Garrity Kerry 3Reveillaud Isabelle 3Kopin Alan 3Jackson F. Rob 4Watnick Paula I 1*
1 Department of Geographic Medicine and Infectious Diseases, Tufts-New England Medical Center, Boston, Massachusetts, United States of America
2 Department of Pathology, Tufts-New England Medical Center, Boston, Massachusetts, United States of America
3 Molecular Cardiology Research Institute, Tufts-New England Medical Center, Boston, Massachusetts, United States of America
4 Department of Neurosciences, Tufts University School of Medicine, Boston, Massachusetts, United States of America
Schneider David Samuel EditorStanford University, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 1 e828 3 2005 8 8 2005 Copyright: © 2005 Blow et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Cholera, the pandemic diarrheal disease caused by the gram-negative bacterium Vibrio cholerae, continues to be a major public health challenge in the developing world. Cholera toxin, which is responsible for the voluminous stools of cholera, causes constitutive activation of adenylyl cyclase, resulting in the export of ions into the intestinal lumen. Environmental studies have demonstrated a close association between V. cholerae and many species of arthropods including insects. Here we report the susceptibility of the fruit fly, Drosophila melanogaster, to oral V. cholerae infection through a process that exhibits many of the hallmarks of human disease: (i) death of the fly is dependent on the presence of cholera toxin and is preceded by rapid weight loss; (ii) flies harboring mutant alleles of either adenylyl cyclase, Gsα, or the Gardos K+ channel homolog SK are resistant to V. cholerae infection; and (iii) ingestion of a K+ channel blocker along with V. cholerae protects wild-type flies against death. In mammals, ingestion of as little as 25 μg of cholera toxin results in massive diarrhea. In contrast, we found that ingestion of cholera toxin was not lethal to the fly. However, when cholera toxin was co-administered with a pathogenic strain of V. cholerae carrying a chromosomal deletion of the genes encoding cholera toxin, death of the fly ensued. These findings suggest that additional virulence factors are required for intoxication of the fly that may not be essential for intoxication of mammals. Furthermore, we demonstrate for the first time the mechanism of action of cholera toxin in a whole organism and the utility of D. melanogaster as an accurate, inexpensive model for elucidation of host susceptibility to cholera.
Synopsis
Cholera, the pandemic diarrheal disease caused by the gram-negative bacterium Vibrio cholerae, continues to be a major public health challenge in the developing world. Environmental studies have demonstrated a close association between V. cholerae and many species of arthropods, and insects have previously been implicated as vectors of this disease. Here researchers report the susceptibility of the fruit fly, Drosophila melanogaster, to oral V. cholerae infection through a process that exhibits many of the hallmarks of human disease. Furthermore, although ingestion of cholera toxin results in massive diarrhea in mammals, these researchers have found that ingestion of purified cholera toxin is not lethal to the fly. However, when co-ingested with a pathogenic strain of V. cholerae carrying a deletion of the cholera toxin genes, cholera toxin is lethal. These findings not only demonstrate the utility of D. melanogaster as an accurate, inexpensive model for elucidation of the host-pathogen interaction and identification of inhibitors of the action of cholera toxin; they also suggest that V. cholerae carries additional virulence factors that enable intoxication of an arthropod host. Based on these findings, the researchers suggest that the fly or a related arthropod may be a true host of V. cholerae in nature.
Citation:Blow NS, Salomon RN, Garrity K, Reveillaud I, Kopin A, et al. (2005) Vibrio cholerae infection of Drosophila melanogaster mimics the human disease cholera. PLoS Pathog 1(1): e8.
==== Body
Introduction
Cholera continues to be a major cause of morbidity and mortality in many parts of the world [1]. It is contracted through ingestion of contaminated food or water and is characterized by profuse diarrhea and vomiting. Cholera toxin, the primary determinant of this clinical syndrome, is an AB5-type exotoxin composed of an A subunit non-covalently bound to five B subunits, arranged in a rosette to form a lectin recognizing the GM1 ganglioside [2]. The mechanism by which cholera toxin enters intestinal epithelial cells and disrupts function has been studied extensively in cultured cells [3–7]. Prior to entry into the cell, the A subunit is proteolytically cleaved into a catalytic A1 subunit and an A2 subunit, whose role is to maintain the non-covalent association to the B subunit GM1 lectin. This lectin forms an association with GM1 gangliosides that are concentrated in lipid rafts within the cell membrane. Once bound to GM1, retrograde transport on lipid rafts delivers cholera toxin to the endoplasmic reticulum. The A1 subunit then dissociates from the toxin complex and exits the endoplasmic reticulum to ADP-ribosylate the stimulatory G protein subunit, Gsα. The modified Gsα constitutively activates adenylyl cyclase, and levels of cAMP in intestinal epithelial cells rise. The consequent secretory diarrhea depends on opening of cAMP-responsive Cl− channels and flow of Cl− and water through the apical surface of the epithelial cell into the intestinal lumen. KCNN4, an intermediate conductance Ca2+-activated K+ channel of mammals, maintains K+ export through the basolateral aspect of the intestinal epithelial cell. Clotrimazole, which blocks the KCNN4 channel, has been shown to decrease cholera toxin-induced Cl− secretion in both cultured mammalian cells and mice [8,9]. These results suggest that simultaneous basolateral export of K+ is required to maintain passage of Cl− through basolateral K+/Cl− cotransporters and apical Cl− channels into the intestinal lumen.
The utility of Drosophila melanogaster as a model host for human pathogens is well-established [10–18]. In the natural environment, Vibrio cholerae is closely associated with arthropods [19–21], and many have suggested that insects serve as vectors [22–26] or reservoirs [27–29] of V. cholerae. Thus, we hypothesized that insects or related arthropods might serve as excellent model hosts of V. cholerae. To test this, we subjected the model insect D. melanogaster to oral V. cholerae infection. Here we demonstrate that V. cholerae infection of D. melanogaster exhibits the following parallels to human disease: (i) ingestion of V. cholerae produces an intestinally-localized, lethal infection in the fly that is dependent on cholera toxin; (ii) host susceptibility is dependent on Gsα, adenylyl cyclase, and the Drosophila KCNN4 channel homolog; and (iii) clotrimazole, an inhibitor of the human KCNN4 channel, protects the fly against infection. However, we have also found differences between V. cholerae infection of mammals and flies. Ingestion of cholera toxin alone is sufficient to cause severe secretory diarrhea in humans and model mammals [30–33]. In contrast, in the fly, we have found that ingestion of cholera toxin is lethal only when pathogenic isolates of V. cholerae are ingested in tandem. Our findings not only demonstrate the utility of the fly as a model host for V. cholerae infection, but also suggest that the V. cholerae genome contains virulence factors specifically required for infection of non-mammalian hosts such as the fly.
Results/Discussion
Ingestion of V. cholerae Results in Lethal Infection of D. melanogaster
To test the utility of D. melanogaster as a model host for V. cholerae, flies were fed either Luria-Bertani (LB) broth alone or inoculated with V. cholerae. Consumption of this growth medium by the fly was documented on multiple occasions by addition of blue dye. Using this experimental design, wild-type flies fed LB broth alone survived for 5 d and could be maintained for up to 2 wk if a larger volume of LB broth was provided. In contrast, flies fed LB inoculated with V. cholerae expired after 3 d regardless of the amount of volume provided (Figure 1). Similar observations were made for the Canton-S wild-type strain of D. melanogaster and for several D. melanogaster strains carrying benign marker mutations (unpublished data).
Figure 1 The Genes Encoding Cholera Toxin Are Required for Lethal V. cholerae Infection of Drosophila
Fractional survival of wild-type Oregon R flies (wtDm) fed LB alone (LB), wild-type V. cholerae (wtVc), or a V. cholerae ΔctxB mutant (ctxB). Ten adult flies (five males and five females), 3–5 d following eclosion were used. Log-rank test analysis demonstrated a statistically significance difference in survival of wild-type V. cholerae infected flies and V. cholerae ΔctxB mutant infected flies (p < 0.0001).
V. cholerae Is Able to Multiply within the Fly
Once ingested by a model mammalian host, V. cholerae is able to multiply within the intestinal compartment [34]. In the experimental model presented above, flies were continuously fed V. cholerae. While this type of infection is rapidly lethal, it does not distinguish between bacterial accumulation and bacterial colonization and multiplication. To test whether V. cholerae was able to persist and multiply within the fly, we measured V. cholerae colony-forming unit (CFU)/fly over time in flies continuously fed LB inoculated with V. cholerae and in flies first fed LB inoculated with V. cholerae for 24 h and then transferred to a vial containing sterile LB broth. At 24 h, flies in both groups harbored equivalent numbers of V. cholerae. As shown in Figure 2A, flies exposed continuously to LB inoculated with V. cholerae expired after 3 d when the burden of V. cholerae reached 3.93 × 107 CFU/fly. Over the course of 4 d, numbers of V. cholerae also increased in flies removed from contaminated food, albeit at a slower rate than flies continuously exposed to V. cholerae. The number of V. cholerae required to bring about death was similar in both groups. These results suggest that V. cholerae is able to colonize and multiply within the fly in the absence of continued ingestion.
Figure 2
V. cholerae Multiplies within the Gut of the Fly following Infection
(A) Colony counts were assayed at 24-h time points from flies infected with V. cholerae. Grey bars indicate CFU per fly obtained from flies fed V. cholerae continuously, while black bars depict CFU per fly for flies fed V. cholerae for 24 h and then removed to a sterile, fresh LB solution.
(B) Section of the midgut of a fly harvested 48 h after introduction to medium containing V. cholerae. Arrows labeled with Vc point to clusters of slender, curved gram negative V. cholerae (pink) present in the lumen of the midgut of the infected fly. Occasional gram positive bacteria (violet), which represent the endogenous flora of the gut, are also present.
(C) Section of the midgut of a fly harvested 48 h after introduction to LB alone. Only endogenous gram positive bacteria (violet) could be observed in the intestines of flies fed sterile LB broth.
V. cholerae Remains Localized to the Fly Gut following Ingestion
During human infection, V. cholerae remains localized to the intestine, causing systemic disease through the action of cholera toxin. To determine whether V. cholerae also remained localized to the Drosophila gut, whole flies fed either sterile LB or the V. cholerae/LB mixture were processed into 5-μm thick histologic sections, stained, and examined. Many slender, comma-shaped, gram-negative rods were found within the midgut of V. cholerae-infected flies (Figure 2B). Although concentrated in the midgut, V. cholerae were also found in other regions of the gut. Careful histologic analysis of all tissues revealed no V. cholerae outside the fly alimentary tract. Interestingly, the intestines of flies fed both sterile LB, and LB inoculated with V. cholerae contained gram-positive rods (Figure 2C). These most likely represent the commensal flora of our laboratory flies.
Cholera Toxin Is a Virulence Factor in V. cholerae Infection of the Fly
We hypothesized that, as is the case in human disease, cholera toxin secreted from V. cholerae within the fly gut was responsible for death. To test this hypothesis, a V. cholerae mutant harboring a deletion in the ctxB gene was constructed and fed to wild-type flies [35]. The ΔctxB mutant was significantly less virulent in the fly model of cholera, demonstrating that cholera toxin is the primary virulence factor in V. cholerae infection of both flies and humans (Figure 1). Although flies fed a ΔctxB mutant survived several days longer than flies fed wild-type V. cholerae, they still died prematurely. Thus, we hypothesize that, in the absence of cholera toxin, other virulence factors contribute to death of the fly.
V. cholerae-Infected Flies Lose Weight Prior to Death
Cholera victims may lose 10% or more of their body weight due to dehydration as a result of secretory diarrhea [36]. If cholera toxin acts via a similar mechanism in the fly, weight loss should also occur during infection of the fly. To test this, flies fed either LB alone or LB inoculated with V. cholerae were weighed on a daily basis. Over the course of 3 d, flies fed V. cholerae lost approximately 25 % of their initial body weight, while flies fed LB alone showed a small weight gain (Figure 3). These results support the hypothesis that flies, like humans, become dehydrated during V. cholerae infection. However, we cannot exclude other causes of weight loss such as a decreased food intake or altered metabolic activity.
Figure 3 Ingestion of V. cholerae Induces Drosophila Weight Loss
Fraction of initial weight gained by wild-type flies (wt Dm) fed either LB alone (LB) or V. cholerae (wt Vc). Error bars represent the standard deviation based on three measurements.
G-sα60A, Adenylyl Cyclase, and SK Channel Mutants Are Resistant to Lethal V. cholerae Infection
Cell culture-based studies have shown that Gsα, adenylyl cyclase, and the KCNN4 channel play an important role in V. cholerae-induced Cl− secretion by intestinal epithelial cells [9,37,38]. We asked whether these same factors might be required for susceptibility of Drosophila to V. cholerae infection by examining the susceptibility of Drosophila strains bearing mutations in the genes encoding G-sα60A, the adenylyl cyclase rutabaga, or the SK channel, a Ca2+-sensitive K+ channel that is the closest Drosophila homolog of the human KCNN4 channel. As shown in Figures 4 and 5A, mutation of G-sα60A and rutabaga provided nearly complete protection against V. cholerae infection. Mutation of Sk provided only partial protection. This may be the result of persistent, albeit reduced expression of the SK channel in this mutant or of additional mechanisms that facilitate Cl− secretion in the fly (Figure 6). Importantly, we confirmed that the additional independently generated mutant alleles for G-sα60A, rut, or SK listed in Table 1 had similar effects on V. cholerae susceptibility, indicating that mutations in these genes, rather than other differences in genetic background, caused the observed phenotypes.
Figure 4 A G-sα60AR60 Mutant Strain Is Resistant to Lethal V. cholerae Infection
Fractional survival over time of wild-type flies (Oregon R; wt Dm) and G-sα60AR60 mutant flies [44] that were fed either LB alone or LB inoculated with wild-type V. cholerae (wt Vc). In these experiments and those illustrated in Figures 5 and 6, ten 3- to 5-d-old adult flies (five males and five females) were infected, and all experiments were performed in triplicate. Log-rank test analysis demonstrated a statistically significant difference in survival of wild-type flies fed wild-type V. cholerae and G-sα60AR60 mutant flies fed wild-type V. cholerae (p < 0.0001).
Figure 5 A rut2080 Mutant Strain Is Resistant to Lethal V. cholerae Infection
(A) Fractional survival over time of wild type flies, rut2080 mutant flies [47], and rut2080
UAS-rut+ fed LB inoculated with V. cholerae (wt Vc). Wild-type flies fed LB broth alone were included as a control. Log-rank test analysis demonstrated a statistically significant difference in the survival of wild-type flies fed wild-type V. cholerae and rut2080 mutant flies fed wild-type V. cholerae (p < 0.0001).
(B) RT-PCR amplification of rutabaga transcripts in wild-type (WT), rut2080, and rut2080 UAS-rut
+ flies. The ribosomal protein rp15a was used as a constitutively transcribed control.
Figure 6 SK Mutant Drosophila and Clotrimazole-Treated Wild-Type Flies Display Partial Resistance to Lethal V. cholerae Infection
Fractional survival over time of wild-type (wt Dm) or SK mutant ((WH}SKf07979) flies fed either wild-type V. cholerae alone or combined with 10 μg/ml clotrimazole (10 μg Clot). Log-rank test analysis demonstrated a statistically significant difference in survival of wild-type flies fed wild-type V. cholerae and SK mutant ((WH}SKf07979) flies fed wild-type V. cholerae (p < 0.0001). There was also a statistically significant difference in survival of wild-type flies fed wild-type V. cholerae and wild-type flies fed wild-type V. cholerae combined with 10 μg/ml clotrimazole (p < 0.0001).
Table 1
Drosophila Alleles Used in Mutant Studies
DOI: 10.1371/journal.ppat.0010008.t001
In preparation for genetic rescue of the rut mutant phenotype using the GAL4/UAS binary expression system, a rut2080 strain homozygous for a UAS-rut
+ transgene insertion on the second chromosome was obtained and assayed for susceptibility to V. cholerae infection [39]. Unexpectedly, these flies were susceptible (Figure 5A). To ascertain the basis of this susceptibility, we assayed levels of rut transcript in wild-type, rut2080, and rut2080;UAS-rut
+ flies by RT-PCR. As shown in Figure 5B, rut transcription was greatly reduced in the rut2080 mutant, but the rut2080;UAS-rut
+ flies had transcript levels comparable to those of wild-type flies. PCR analysis confirmed the presence of the rut2080 insertion in both strains. Thus, we conclude that the UAS-rut transgene is transcribed in the absence of Gal4, presumably by regulation from an adjacent genomic element. Furthermore, we conclude that susceptibility of rut mutant flies to V. cholerae infection is rescued by restoration of wild-type levels of the rutabaga transcript.
Clotrimazole Protects V. cholerae-Infected Flies against Death
Because clotrimazole abrogates the V. cholerae-induced secretory diarrhea in mammals by inhibiting K+ transport through KCNN4 channels, we postulated that co-administration of clotrimazole with V. cholerae might also block K+ transport through the Drosophila SK channel and, therefore, protect wild-type flies against death. Figure 6 shows that this was indeed the case. However, co-administration of clotrimazole had no effect on survival of SK mutant flies, suggesting that clotrimazole is, in fact, exerting its effect by interaction with the SK channel (Figure 6).
A Factor Carried by Pathogenic V. cholerae Is Required for Intoxication of the Fly by Cholera Toxin
Ingestion of cholera toxin is sufficient to cause massive intestinal fluid accumulation and diarrhea in mammals [30–33]. Thus, we predicted that ingestion of purified, active cholera toxin alone would result in death of the fly. Remarkably, ingestion of LB containing as much as 100 μg/ml of cholera toxin did not alter survival of the fly (unpublished data). We questioned whether the presence of V. cholerae itself might be required for intoxication of the fly by cholera toxin. To test this, we fed LB containing both cholera toxin and a V. cholerae ΔctxB mutant to flies. As shown in Figure 7, ingestion of cholera toxin in the presence of the ΔctxB mutant V. cholerae resulted in death of the flies at rates similar to those of flies infected with wild-type V. cholerae alone. This suggested to us that an unknown bacterial factor might be required for intoxication of the fly by cholera toxin. To determine whether this factor might be specific to pathogenic isolates of V. cholerae, we fed LB containing cholera toxin and one of several non-toxigenic environmental isolates of V. cholerae to flies. In each case, there was no significant difference in survival between flies fed V. cholerae alone and those fed V. cholerae combined with cholera toxin. To test whether this cholera toxin-potentiating factor was carried on the CTXΦ, we combined cholera toxin with Bengal2, a pathogenic strain of V. cholerae carrying a deletion of the CTXΦ. This mutant was also able to provide the fly-specific virulence factor (unpublished data). Thus, this factor is not carried on the CTXΦ. These experiments suggest that pathogenic V. cholerae possess a virulence factor or factors that are essential for intoxication of arthropods but not mammals by cholera toxin.
Figure 7 A Bacterial Factor Is Required for Intoxication of the Fly by Cholera Toxin
Fractional survival over time of wild-type flies fed LB alone, wild-type V. cholerae, or a V. choleraeΔctxB mutant (ctxB) either with or without 10 μg/ml purified cholera toxin. Log-rank test analysis demonstrated a statistically significant difference in the survival of wild-type flies fed a V. cholerae ΔctxB mutant (ctxB) alone and those fed a V. cholerae ΔctxB mutant (ctxB) combined with purified cholera toxin (p < 0.0001).
Implications of this Model for the Study, Treatment, and Ecology of Cholera
We have demonstrated surprising parallels in the mechanism of V. cholerae-mediated death of man and the model arthropod D. melanogaster. Cholera toxin is the primary virulence factor in both infections. While the mechanism of cholera toxin has previously been elucidated in cultured intestinal epithelial cells, we present the first evidence that this mechanism is also operative in whole organisms. Furthermore, this model system will have wide-ranging applications to the study of this devastating disease. Due to the expense and labor involved in mammalian genetic screens, little is known about the host factors that govern susceptibility to cholera. Because lethal oral infection of the fly requires no manipulation by the experimentalist and has an easily measured outcome, the fly provides a powerful tool to be used in large-scale genetic screens for host susceptibility factors and bacterial virulence factors. The current mainstay of cholera therapy consists of administration of oral or intravenous water and ions until the infection is overcome by antibiotics and /or the innate immune system. An inhibitor of the secretory diarrhea caused by cholera toxin would be a potentially life-saving adjuvant to this therapy. We have shown here that oral agents can block the action of cholera toxin in the fly. Thus, this model will also facilitate screens of combinatorial chemical libraries for inhibitors of cholera toxin and secretory diarrhea. Finally, these studies highlight a host-pathogen interaction that could easily occur in nature. Close contact between V. cholerae and arthropods has been documented and is likely more frequent than that between V. cholerae and humans [19,40–42]. In fact, environmental studies have demonstrated that common house flies carry V. cholerae in endemic areas [22–25]. In this work, we have presented evidence that pathogenic V. cholerae carry virulence factors that are essential for intoxication of the fly but not mammals. Thus, we present the provocative hypothesis that the pathogenic program of V. cholerae may have evolved for an arthropod rather than for us.
Materials and Methods
Bacterial strains, fly strains, and growth media.
MO10, a V. cholerae O139 clinical isolate, and mutants derived from this strain were used in all experiments [43]. All fly strains were reared at room temperature on standard Drosophila media. The wild-type OregonR fly strain was used for most studies. Gsα, rut, and Sk experiments utilized mutant fly lines harboring G-sα 60AR60, a loss-of-function allele that reduces the cAMP concentration 4- to 5-fold in larvae [44], rut2080, an enhancer trap element in the 5′ flanking region of the rut gene [45], and PBac(WH}SKf07979 , respectively (Table 1). The rut2080 and rut2080;UAS-rut+ fly lines were generously provided by Ron Davis. The presence of the rut2080 mutant allele was confirmed by PCR amplification of a portion of the insertion element for both lines. Additionally, fly lines carrying G-sα60AB19, P(EP}rutEP399 or P(GT1}rutBG00139, and P(SUPor-P}KG00471 or P(GT1}SKBG01378 were used to confirm the results of experiments with the G-sα60AR60 , rut2080, and PBac(WH}SKf07979 mutant fly strains, respectively (Table 1). Fly lines other than those noted were obtained from the Bloomington Drosophila Stock Center (Bloomington, Indiana).
V. cholerae mutant construction.
The V. cholerae ΔctxB mutant, harboring a 321 bp deletion in the ctxB gene (VC1456) was constructed by double homologous recombination according to previously described protocols [35]. The deletion removed all but 11 amino acids remaining at the amino-terminus of the protein and the terminal stop codon.
Survival of Drosophila following V. cholerae infection.
Ten wild-type Oregon-R adult flies were placed in each of three vials containing a cotton plug saturated with Luria-Bertani (LB) broth either alone or inoculated with 108 CFU/ml of V. cholerae O139 strain MO10 or another strain as noted in the text [46]. Viable flies were counted at 24-h intervals. Reproducibility of all survival curves was confirmed in at least three independent experiments, and log-rank tests were used to determine statistical significance.
Histological studies.
Flies were fed either LB inoculated with V. cholerae or LB alone for 48 h, and then anesthetized and fixed in formalin for 48 h prior to processing. Flies were processed on a tissue processor (Leica ASP 300, Wetzlar, Germany) and embedded in paraffin. The embedded flies were sectioned into 5-μm ribbons, which were placed on positively charged glass slides, baked at 65 °C overnight, and gram stained.
Weight loss measurements.
Sets of ten female flies were weighed and then transferred to fly vials containing either LB alone or LB inoculated with V. cholerae. Flies, housed in thin-walled Eppendorf tubes, were weighed 24 and 48 h after transfer, using a precision balance (Mettler Toledo AG204, Columbus, Ohio). All experiments were performed in triplicate, and the average ratios of final to initial weight were calculated.
Quantification of V. cholerae within flies.
To determine whether V. cholerae was able to colonize and multiply within the fly, flies fed either LB alone or LB inoculated with V. cholerae were anesthetized, removed from vials, and homogenized in LB broth at 24-h intervals. Particulates were pelleted, and dilutions of the resulting supernatants were plated on LB-agar supplemented with streptomycin (100 mg/ml). In all cases, no colonies were obtained from LB-fed flies.
RT-PCR.
Total RNA was extracted from five flies using the Trizol reagent (Gibco BRL, San Diego, California, United States). Prior to RT-PCR amplification, total RNA was DNAase I-treated (Ambion, Austin, Texas, United States) for 30 min at 37 °C. DNAse I was inactivated using the DNAse inactivation reagent (Ambion). RT-PCR was performed in two steps using Superscript II RT (Gibco BRL) to obtain cDNA and Taq to perform PCR. The following primer pairs were used: rut (5′-GATCCAGGATGAGAACGA-3′, 5′-CGGAGACACAATAGTAACAGTC-3′) and Drosophila ribosomal protein 15a (5′-CGTTTGCGTGACGGTCGTGT-3′, 5′-GCCGAGAATTTTGCCTCCCAA-3′).
Fly intoxication with purified cholera toxin.
Adult Oregon-R flies 3–5 d old were fed cholera toxin diluted to the specified concentrations in LB broth. Overnight cultures containing V. cholerae strains were also added to the mixture in a 1:10 dilution where specified. Flies were monitored at 24-h time intervals until death. Survival of flies was plotted against time using Kaplan-Meier plots, and a log-rank test was performed to determine statistical significance.
We thank Ron Davis for generous sharing of rutabaga mutants and constructs. We thank Dr. Anne Kane of the Tufts-NEMC GRASP Center for her careful reading of the manuscript, her staff for their expert preparation of many reagents, and Mr. Javier Mendez for expert technical assistance in the preparation of histologic specimens. This work was supported by a pilot project grant from the Tufts-NEMC GRASP Center NIH/NIDDK (P30 DK34928 and NIH R21 AI64800 to PIW).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. NSB, RNS, AK, FRJ, and PIW conceived and designed the experiments. NSB and RNS performed the experiments. NSB, RNS, and PIW analyzed the data. KG, IR, and AK contributed reagents/materials/analysis tools. NSB and PIW wrote the paper.
Abbreviations
CFUcolony-forming units
LBLuria-Bertani broth
==== Refs
References
Glass RI Black RE 1992 The epidemiology of cholera Barua D Greenough WBI Cholera New York Plenum 129 154
Walia K Ghosh S Singh H Nair GB Ghosh A 1999 Purification and characterization of novel toxin produced by Vibrio cholerae O1 Infect Immun 67 5215 5222 10496898
Tsai B Rodighiero C Lencer WI Rapoport TA 2001 Protein disulfide isomerase acts as a redox-dependent chaperone to unfold cholera toxin Cell 104 937 948 11290330
Lencer WI Constable C Moe S Jobling MG Webb HM 1995 Targeting of cholera toxin and Escherichia coli heat labile toxin in polarized epithelia: Role of COOH-terminal KDEL J Cell Biol 131 951 962 7490296
Lencer WI Reinhart FD Neutra MR 1990 Interaction of cholera toxin with cloned human goblet cells in monolayer culture Am J Physiol 258 96 102
Fujinaga Y Wolf AA Rodighiero C Wheeler H Tsai B 2003 Gangliosides that associate with lipid rafts mediate transport of cholera and related toxins from the plasma membrane to endoplasmic reticulum Mol Biol Cell 14 4783 4793 13679513
Lencer WI Tsai B 2003 The intracellular voyage of cholera toxin: Going retro Trends Biochem Sci 28 639 645 14659695
Rufo PA Jiang L Moe SJ Brugnara C Alper SL 1996 The antifungal antibiotic, clotrimazole, inhibits Cl− secretion by polarized monolayers of human colonic epithelial cells J Clin Invest 98 2066 2075 8903326
Rufo PA Merlin D Riegler M Ferguson-Maltzman MH Dickinson BL 1997 The antifungal antibiotic, clotrimazole, inhibits chloride secretion by human intestinal T84 cells via blockade of distinct basolateral K+ conductances. Demonstration of efficacy in intact rabbit colon and in an in vivo mouse model of cholera J Clin Invest 100 3111 3120 9399958
Mansfield BE Dionne MS Schneider DS Freitag NE 2003 Exploration of host-pathogen interactions using Listeria monocytogenes and Drosophila melanogaster
Cell Microbiol 5 901 911 14641175
Brandt SM Dionne MS Khush RS Pham LN Vigdal TJ 2004 Secreted bacterial effectors and host-produced Eiger/TNF drive death in a Salmonella -infected fruit fly PLoS Biol 2 e418 15562316
Garcia-Lara J Needham AJ Foster SJ 2005 Invertebrates as animal models for Staphylococcus aureus pathogenesis: A window into host-pathogen interaction FEMS Immunol Med Microbiol 43 311 323 15708304
Apidianakis Y Mindrinos MN Xiao W Lau GW Baldini RL 2005 Profiling early infection responses: Pseudomonas aeruginosa eludes host defenses by suppressing antimicrobial peptide gene expression Proc Natl Acad Sci U S A 102 2573 2578 15695583
Schneider D Shahabuddin M 2000 Malaria parasite development in a Drosophila model Science 288 2376 2379 10875925
Zambon RA Nandakumar M Vakharia VN Wu LP 2005 The Toll pathway is important for an antiviral response in Drosophila
Proc Natl Acad Sci U S A 102 7257 7262 15878994
Rutschmann S Jung AC Hetru C Reichhart JM Hoffmann JA 2000 The Rel protein DIF mediates the antifungal but not the antibacterial host defense in Drosophila
Immunity 12 569 580 10843389
Lionakis MS Lewis RE May GS Wiederhold NP Albert ND 2005 Toll-deficient Drosophila flies as a fast, high-throughput model for the study of antifungal drug efficacy against invasive aspergillosis and Aspergillus virulence J Infect Dis 191 1188 1195 15747256
Basset A Tzou P Lemaitre B Boccard F 2003 A single gene that promotes interaction of a phytopathogenic bacterium with its insect vector, Drosophila melanogaster
EMBO Rep 4 205 209 12612613
Huq A Huq SA Grimes DJ O'Brien M Chu KH 1986 Colonization of the gut of the blue crab (Callinectes sapidus) by Vibrio cholerae
Appl Environ Microbiol 52 586 588 3767362
Shukla BN Singh DV Sanyal SC 1995 Attachment of non-culturable toxigenic Vibrio cholerae O1 and non-O1 and Aeromonas spp. to the aquatic arthropod Gerris spinolae and plants in the River Ganga, Varanasi FEMS Immunol and Medical Microbiol 12 113 120
Tamplin ML Gauzens AL Huq A Sack DA Colwell RR 1990 Attachment of Vibrio cholerae serogroup O1 to zooplankton and phytoplankton of Bangladesh waters Appl Environ Microbiol 56 1977 1980 2383016
Fotedar R 2001 Vector potential of houseflies (Musca domestica) in the transmission of Vibrio cholerae in India Acta Trop 78 31 34 11164748
Echeverria P Harrison BA Tirapat C McFarland A 1983 Flies as a source of enteric pathogens in a rural village in Thailand Appl Environ Microbiol 46 32 36 6351748
Sukontason K Bunchoo M Khantawa B Piangjai S Methanitikorn R 2000 Mechanical carrier of bacterial enteric pathogens by Chrysomya megacephala (Diptera: Calliphoridae) in Chiang Mai, Thailand Southeast Asian J Trop Med Public Health 31 S157 S161
Khin Nwe O Sebastian AA Aye T 1989 Carriage of enteric bacterial pathogens by house flies in Yangon, Myanmar J Diarrhoeal Dis Res 7 81 84 2632639
Khan AR Huq F 1978 Disease agents carried by flies in Dacca city Bangladesh Med Res Counc Bull 4 86 93 757347
Broza M Halpern M 2001 Pathogen reservoirs. Chironomid egg masses and Vibrio cholerae
Nature 412 40
Halpern M Gancz H Broza M Kashi Y 2003
Vibrio cholerae hemagglutinin/protease degrades chironomid egg masses Appl Environ Microbiol 69 4200 4204 12839800
Halpern M Broza YB Mittler S Arakawa E Broza M 2004 Chironomid egg masses as a natural reservoir of Vibrio cholerae non-O1 and non-O139 in freshwater habitats Microb Ecol 47 341 349 14681736
Baselski V Briggs R Parker C 1977 Intestinal fluid accumulation induced by oral challenge with Vibrio cholerae or cholera toxin in infant mice Infect Immun 15 704 712 870427
Finkelstein RA Norris HT Dutta NK 1964 Pathogenesis experimental cholera in infant rabbits. I. Observations on the intraintestinal infection and experimental cholera produced with cell-free products J Infect Dis 114 203 216 14183392
Finkelstein RA LoSpalluto JJ 1969 Pathogenesis of experimental cholera. Preparation and isolation of choleragen and choleragenoid J Exp Med 130 185 202 4978880
Levine MM Kaper JB Black RE Clements ML 1983 New knowledge on pathogenesis of bacterial enteric infections as applied to vaccine development Microbiol Rev 47 510 550 6363898
Angelichio MJ Spector J Waldor MK Camilli A 1999
Vibrio cholerae intestinal population dynamics in the suckling mouse model of infection Infect Immun 67 3733 3739 10417131
Haugo AJ Watnick PI 2002
Vibrio cholerae CytR is a repressor of biofilm development Mol Microbiol 45 471 483 12123457
Sack DA Sack RB Nair GB Siddique AK 2004 Cholera Lancet 363 223 233 14738797
Moss J Vaughan M 1977 Mechanism of action of choleragen. Evidence for ADP-ribosyltransferase activity with arginine as an acceptor J Biol Chem 252 2455 2457 139409
Kassis S Hagmann J Fishman PH Chang PP Moss J 1982 Mechanism of action of cholera toxin on intact cells. Generation of A1 peptide and activation of adenylate cyclase J Biol Chem 257 12148 12152 6288709
Zars T Wolf R Davis R Heisenberg M 2000 Tissue-specific expression of a type I adenylyl cyclase rescues the rutabaga mutant memory defect: In search of the engram Learn Mem 7 18 31 10706599
Huo A Xu B Chowdhury MA Islam MS Montilla R 1996 A simple filtration method to remove plankton-associated Vibrio cholerae in raw water supplies in developing countries Appl Environ Microbiol 62 2508 2512 8779590
Huq A Small EB West PA Huq MI Rahman R 1983 Ecological relationships between Vibrio cholerae and planktonic crustacean copepods Appl Environ Microbiol 45 275 283 6337551
Chiavelli DA Marsh JW Taylor RK 2001 The mannose-sensitive hemagglutinin of Vibrio cholerae promotes adherence to zooplankton Appl Environ Microbiol 67 3220 3225 11425745
Waldor MK Mekalanos JJ 1994 Emergence of a new cholera pandemic: Molecular analysis of virulence determinants in Vibrio cholerae O139 and development of a live vaccine prototype J Infect Dis 170 278 283 8035010
Wolfgang WJ Hoskote A Roberts IJ Jackson S Forte M 2001 Genetic analysis of the Drosophila Gs(alpha) gene Genetics 158 1189 1201 11454767
Zars T 2000 Behavioral functions of the insect mushroom bodies Curr Opin Neurobiol 10 790 795 11240291
Waldor MK Colwell R Mekalanos JJ 1994 The Vibrio cholerae O139 serogroup antigen includes an O-polysaccharide capsule and lipopolysaccharide virulence determinant Proc Natl Acad Sci USA 91 11388 11392 7972070
Han PL Meller V Davis RL 1996 The Drosophila brain revisited by enhancer detection J Neurobiol 31 88 102 9120439
Bellen HJ Levis RW Liao G He Y Carlson JW 2004 The BDGP gene disruption project: Single transposon insertions associated with 40% of Drosophila genes Genetics 167 761 781 15238527
Thibault ST Singer MA Miyazaki WY Milash B Dompe NA 2004 A complementary transposon tool kit for Drosophila melanogaster using P and piggyBac Nat Genet 36 283 287 14981521
|
16201020
|
PMC1238743
|
CC BY
|
2021-01-05 12:11:27
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e8
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010008
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 1620102110.1371/journal.ppat.001000905-PLPA-RA-0042R3plpa-01-01-10Research ArticleGenetics/Gene FunctionParasitologyPlasmodiumExit of Plasmodium Sporozoites from Oocysts Is an Active Process That Involves the Circumsporozoite Protein Mutational Analysis of Region II+ of CS ProteinWang Qian 1*Fujioka Hisashi 2Nussenzweig Victor 1
1 Department of Pathology, Michael Heidelberger Division, New York University School of Medicine, New York, New York, United States of America
2 Institute of Pathology, Case Western Reserve University, Cleveland, Ohio, United States of America
Haldar Kasturi EditorNorthwestern University Medical School, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 1 e99 5 2005 3 8 2005 Copyright: © 2005 Wang et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Plasmodium sporozoites develop within oocysts residing in the mosquito midgut. Mature sporozoites exit the oocysts, enter the hemolymph, and invade the salivary glands. The circumsporozoite (CS) protein is the major surface protein of salivary gland and oocyst sporozoites. It is also found on the oocyst plasma membrane and on the inner surface of the oocyst capsule. CS protein contains a conserved motif of positively charged amino acids: region II-plus, which has been implicated in the initial stages of sporozoite invasion of hepatocytes. We investigated the function of region II-plus by generating mutant parasites in which the region had been substituted with alanines. Mutant parasites produced normal numbers of sporozoites in the oocysts, but the sporozoites were unable to exit the oocysts. In in vitro as well, there was a profound delay, upon trypsin treatment, in the release of mutant sporozoites from oocysts. We conclude that the exit of sporozoites from oocysts is an active process that involves the region II-plus of CS protein. In addition, the mutant sporozoites were not infective to young rats. These findings provide a new target for developing reagents that interfere with the transmission of malaria.
Synopsis
Malaria affects hundreds of millions of people, and kills at least 1 million children per year. The infective stages of the malaria parasites, named “sporozoites,” are found in the salivary gland of Anopheles mosquitoes, and are injected along with the saliva during blood feeding. From the skin, sporozoites enter the blood circulation and invade liver cells where the parasites multiply. When they exit the liver, these parasites infect blood cells and can cause severe symptoms. If ingested by mosquitoes, the blood-stage parasites continue their lifecycle in the insect stomach. Thousands of sporozoites are formed within a cyst-like structure (oocyst). The sporozoites come out of the oocyst and infect the salivary gland, where they remain until injected back into humans. Malaria parasites are increasingly resistant to drugs, mosquitoes are difficult to eliminate, and effective vaccines are not yet available. New tools to combat malaria are urgently needed. One exciting approach, although the application is in the distant future, is to release in endemic areas genetically modified mosquitoes that are resistant to parasite growth. This paper provides a new target for generating these “transmission-block” mosquitoes and shows that the exit of sporozoites from the oocysts is an active process that requires the enzymatic digestion of components of the oocyst wall. If these enzymes are inhibited in transgenic mosquitoes, sporozoites will never reach the salivary gland.
Citation:Wang Q, Fujioka H, Nussenzweig V (2005) Exit of Plasmodium sporozoites from oocysts is an active process that involves the circumsporozoite protein. PLoS Pathog 1(1): e9.
==== Body
Introduction
The Plasmodium life cycle in the Anopheles mosquitoes begins with the ingestion of a blood meal containing Plasmodium gametocytes. After fertilization of the resulting gametes, the zygotes transform into motile ookinetes that traverse the midgut epithelium, reach the basal lamina, and then transform into oocysts. The young oocyst, surrounded by a capsule and by the basal lamina, undergoes multiple mitotic nuclear divisions and progressively enlarges in size without cytokinesis. At the same time, the cytoplasm is subdivided by membrane clefts forming structures named “sporoblasts.” Later, uninucleate sporozoites bud from the sporoblast membrane. The mature oocyst is about 50 μm in diameter and contains thousands of sporozoites. Sporozoites leave the oocysts asynchronously, enter the hemolymph, and then invade the salivary glands where they remain until they are injected, with the saliva, into the skin of the mammalian host [1,2].
In order to reach the flowing hemolymph, sporozoites must traverse two physical barriers: the oocyst capsule and the mosquito basal lamina. Because oocyst sporozoites display limited movement [3], their egress from oocysts is generally thought to be a passive process. Early ultrastructural observations reveal the presence of small openings in the capsule of mature oocysts and the basal lamina. Occasionally, sporozoites are found “penetrating” these openings and entering the hemolymph [4]. The oocyst capsule contains laminin of mosquito origin and displays trans-glutaminase activity probably of parasite origin [5,6]. In addition, the inner surface of the capsule is covered with the Plasmodium circumsporozoite (CS) protein [7–9].
The development of sporozoites in oocysts is CS dependent. When the CS gene is deleted, the oocysts are devoid of mature parasites [10]. To investigate the mechanisms leading to this developmental arrest, we have generated Plasmodium berghei parasites bearing different mutations in the CS coding region. In one of the P. berghei CS mutants, we substituted the positively charged amino acids of the conserved region II-plus with alanines. Region II-plus is located at the 5′ end of the thrombospondin type 1 repeat (TSR) domain of CS protein. Several in vitro observations strongly suggest that region II-plus participates in the initial steps of sporozoite attachment and invasion of the host's hepatocytes via interaction with heparan sulfate proteoglycan (HSPGs) on the host cells [11,12]. Here we show for the first time that the mutation in region II-plus of CS protein prevents the exit of sporozoites from oocysts and progression of the Plasmodium lifecycle. In addition, the mutant sporozoites are unable to infect rats.
Results/Discussion
Construction of CS-RIImut and CS-WT
A P. berghei clone with mutated region II-plus of CS protein (R290A, K291A, R292A, and K293A) was obtained by homologous recombination. In order to avoid any potential defects in the locus associated with the recombination event, a control clone, CS-WT, which produces wild-type CS protein, was generated by the same method (pRCS-WT and pRCS-RIImut, Figure 1A and 1B). A PstI site was introduced in pRCS-RIImut in order to detect the presence of mutations by PCR and Southern blot analysis [13]. The schematic structure of CS and the sequence of region II-plus of wild-type and mutant CS are shown in Figure 1C.
Figure 1 Gene Targeting at the CS Locus of P. berghei
(A and B) Replacement plasmid pRCS-WT and pRCS-RIImut, wild-type (WT) CS locus and recombinant locus. ORFs are symbolized by boxes. Small black box in the CS ORF indicates the mutation in region II-plus (R290A, K291A, R292A, and K293A). A PstI site is introduced in the mutation site to differentiate CS-RIImut from CS-WT. Thick lines indicate 5′- and 3′-UTRs of DHFR-TS; thin lines, 5′- and 3′UTRs of CS; dashed lines, plasmid vector sequence. B, BamHI; K, KpnI; N, NotI; P, PstI; S, SacI; Xb, XbaI; Xh, XhoI.
Recombinant CS-WT or CS-RIImut was generated by double crossover occurring between the CS sequences in the KpnI-SacI fragment of plasmid pRCS-WT or pRCS-RIImut and their homologous sequence in the wild-type CS genomic locus. The CS probe used in the genomic Southern hybridization is symbolized by a thick dash-dot line. Restriction fragments of the wild type and of the expected recombinants are shown below the corresponding locus. Locations of primers used for PCR are indicated in (B).
(C) Schematic structure of the CS protein and sequences of the region II-plus of wild-type and mutant CS.
(D) Genomic Southern hybridization of the wild-type P. berghei strain (WT), the recombinant lines, control (CS-WT), and mutant (CS-RIImut) parasites upon digestion with XbaI and PstI, using a CS probe.
(E) PCR amplification with primers CS1 and PB103 of the expressed CS protein at the 5′ recombinant locus in CS-RIImut and CS-WT. The WT is used as a negative control (lack of recombination events). The amplicons (2.3 kb) were subjected to PstI restriction enzyme digestion (two fragments released, 1.6 and 0.7 kb) to examine the presence of mutation in CS-RIImut, which is absent in CS-WT.
(F) PCR amplification of the 3′ recombinant locus using primers PB106 and CS4.
Genomic DNA from WT, CS-WT, and CS-RIImut were digested with XbaI and PstI, and subjected to Southern blot hybridization. WT displays a 7.9-kb band, whereas CS-WT displays a 5.5-kb band and CS-RIImut a 2.2-kb band, indicating that the CS-RIImut and CS-WT have a correct recombination locus (Figure 1D). This was confirmed by PCR amplification specific for recombinants (Figure 1E and 1F), subsequent PstI digestion (Figure 1E), and by sequencing of PCR products. The sequences of coding regions of wild-type and mutant CS are as expected.
We cannot exclude the possibility that the substitution of the four positively charged amino acids of region II-plus led to secondary changes in the structure of CS protein.
CS-RIImut Sporozoites Do Not Exit from Oocysts
Groups of Anopheles stephensi mosquitoes were infected with CS-WT, a clone of wild-type P. berghei NK65 strain (WT), and two independent clones of CS-RIImut. CS-WT and WT are identical; therefore, CS-WT was used as a wild type control in all experiments. The numbers of oocysts and oocyst sporozoites at 14 d after blood meal (post-infection [PI]) were very similar in CS-WT and in the two mutant clones (Table 1). However, profound differences were observed at later time points. At day 16 and 18 PI, mutant infected mosquitoes contained many more oocyst sporozoites compared to wild-type infected ones (Figure 2A). In two other independent feeding experiments, similar results were obtained at day 16 PI: 50,000 and 96,000 oocyst sporozoites/mosquito for CS-RIImut, versus 35,000 and 75,000 oocyst sporozoites/mosquito for CS-WT, respectively. By contrast, the hemolymph of mosquitoes infected with CS-WT parasites contained many more sporozoites compared to mosquitoes infected with mutant parasites. CS-WT sporozoites entered the hemolymph beginning at day 12 PI, and the peak numbers were reached around day 18 PI. In contrast, even at 28 days PI, only minimal numbers of CS-RIImut sporozoites were found in the hemolymph (Figure 2B and 2C). Thus, the observed increase in the numbers of CS-RIImut oocyst sporozoites between days 14 and 18 is most likely a consequence of their inability to be released in to the hemolymph.
Table 1 Parasite Development in Mosquitoes
Figure 2 CS-RIImut Oocyst Sporozoites Are Not Released from the Midguts
Represented in each graph are the mean numbers of sporozoites per mosquito at different days PI. Each number is calculated based on an average of 20 mosquitoes.
(A) Numbers of oocyst sporozoites per mosquito from CS-WT– and CS-RIImut–infected mosquitoes. In CS-RIImut, the number of sporozoites increases progressively until day 18 PI, whereas in CS-WT a plateau is reached at day 14.
(B) Numbers of sporozoites in the hemolymph from CS-WT– and CS-RIImut–infected mosquitoes from the same lot as in (A). CS-WT oocysts release sporozoites into the hemolymph from day 12 to day 18 PI. The peak is around day 18 PI. In contrast, the hemolymph from CS-RIImut contains very few sporozoites.
(C) In another feeding experiment, the number of hemolymph sporozoites from CS-WT and CS-RIImut infected mosquitoes were calculated from day 14 to day 28 PI. CS-WT releases sporozoites into the hemolymph up to day 28 PI, whereas CS-RIImut does not.
CS-RIImut Sporozoites Display Normal Morphology and Motility
The morphology of CS-RIImut sporozoites was analyzed by immunofluorescence assays (Figure 3A), transmission electron microscopy, and immuno-electron microscopy using monoclonal antibodies (3D11) to the repeats of the CS protein. (Figure 3B–3F). CS-RIImut did not display any abnormalities in the sporozoite morphology during development (Figure 3B). The detailed structure of the mutant sporozoite surface is shown Figure 3C and 3D. The structures of the trimembrane pellicle (plasma membrane and inner membrane complex) and subpellicular microtubules are indistinguishable from those of wild type. Patterns of CS protein labeling in CS-WT and CS-RIImut sporozoites were indistinguishable. In the mutants, very similar to wild type, CS protein was detected on the surface of budding or fully developed sporozoites (Figure 3A and 3E), and on the inner surface of the capsule (Figure 3E and 3F) [2,14]. To compare the amounts of CS protein in mutant and CS-WT parasites, extracts of CS-WT and CS-RIImut oocyst sporozoites were analyzed by Western blot (Figure 3G). The intensity of both precursor (54 kDa) and mature (44 kDa) forms of CS [15], was very similar in the WT and the mutant. The two bands appear slightly smaller in the mutant as a result of the replacement of four basic residues (R290, K291, R292, and K293) with alanines. As a control we analyzed levels of TRAP (thrombospondin-related anonymous protein), another sporozoite surface protein [16], and found that it was not affected in the CS mutant (Figure 3G). We conclude that the recombination event and mutations did not grossly affect CS protein expression or stability.
Figure 3 Ultrastructure and CS Localization of CS-WT and CS-RIImut
(A) Immunofluorescence labeling of oocyst sporozoites from CS-WT and CS-RIImut at day 18 PI. Sporozoites were stained with anti-CS antibody and detected by FITC-conjugated anti-IgG antibodies, without prior permeabilization.
(B) Transmission electron micrographs showing a CS-RIImut oocyst at day 14 PI. The oocyst is surrounded by a capsule and the mosquito basal lamina, and contains sections of sporozoites with normal morphology. Fully developed sporozoites are found within the CS-RIImut oocyst. Sporozoites have homogenous size and morphology. Scale bar represents 1 μm.
(C) Longitudinal section of the CS-RIImut sporozoite pellicle shows the plasma membrane, inner membrane complex, and an associated microtubule. Scale bar represents 0.5 μm.
(D) Cross-section of a CS-RIImut sporozoite showing the trimembrane pellicle and subpellicular microtubules. Scale bar represents 0.5 μm.
(E) Immuno-electron micrographs showing CS localization in the CS-RIImut oocyst. CS protein is predominantly found on the surface of sporozoites and the residual body, and on the inner surface of the capsule. Scale bar represents 1 μm.
(F) Enlarged picture of the CS-RIImut oocyst capsule. CS protein is detected on the inner surface of the capsule. Scale bar represents 1 μm.
(G) Western blot analysis of extracts from CS-WT (WT) and CS-RIImut (RII) oocyst sporozoites. Numbers of sporozoites loaded are indicated on the top of each lane. Sporozoites were collected from oocysts in mosquito midguts at days 14 and 18 PI.
C, capsule; IMC, inner membrane complex; Mo, mosquito tissue; MT, microtubule; Oo, oocyst; PM, plasma membrane; Rb, residual body; SPZ, sporozoite.
It could be argued that the sporozoite exit from oocysts requires sporozoite motility and that the motility is impaired in the mutants. In fact, previous studies have shown that sporozoite motility is neither required nor does it ensure the exit of sporozoites from oocysts [14,17]. In contrast to the circular gliding observed in salivary gland sporozoites, movements of oocyst sporozoites are mostly limited to stretching and back-and-forth gliding [3]. We observed that approximately 3% CS-RIImut oocyst sporozoites display discontinuous gliding motility, and approximately 10%–15% display stretching and bending. The numbers are very similar to those of CS-WT.
CS-RIImut Oocysts Are More Resistant to Proteolytic Activity
Although CS-WT and CS-RIImut are morphologically indistinguishable, develop equally well in mosquitoes, move similarly, and contain equal levels of CS protein, sporozoite egress from mutant oocyst is profoundly defective. Little is known of the process of sporozoite exit from oocysts, but some information can be obtained from the erythrocytic stages of the parasite. During development in the red blood cells, malaria parasites reside inside a parasitophorous vacuole. Merozoite egress requires the rupture of the parasitophorous vacuole and the membrane of the red blood cells. Release of merozoites from infected erythrocytes requires proteases and is inhibited by inhibitors of proteolytic enzymes [18–20]. Plasmodium falciparum falcipain-2 (a cysteine protease) cleaves erythrocyte membrane skeletal proteins at late stages of parasite development [21], facilitating the merozoite egress.
To examine the possible role of a proteolytic event in the release of sporozoites from oocysts, we treated isolated midguts from CS-WT– or CS-RIImut–infected mosquitoes (14 days PI) with trypsin and measured the number of released sporozoites (Figure 4). The lower temperature (25 °C) was chosen to mimic natural conditions of oocyst development in the mosquito midguts. In the absence of trypsin, very few sporozoites were released from either CS-WT– or CS-RIImut–infected midguts even after 3 h of incubation. Treatment with trypsin for 40 min leads to a significant increase in the number of sporozoites released from CS-WT, but not from CS-RIImut. Release of sporozoites from CS-RIImut oocysts was achieved only after extended treatment with trypsin. In Figure 4 we show the release at 14 d PI, but identical results were observed with 18-d oocysts (data not shown). This effect is trypsin specific, because the release of sporozoites was abolished when the soybean trypsin inhibitor was included in the incubation (data not shown).
Figure 4 In Vitro Oocyst Sporozoite Release Assay
Release of oocyst sporozoites at 25 °C in vitro in the presence of trypsin (50 μg/ml) at day 14 PI. Y-axis represents released sporozoites as a percentage of total oocyst sporozoites. X-axis indicates the time point when the samples are collected. In the absence of the trypsin, very few sporozoites are released from the midguts. In the presence of the trypsin, oocyst sporozoites from both CS-WT (WT) and CS-RIImut (RII) are released in a time-dependent manner. Compared with CS-WT, CS-RIImut sporozoites are released more slowly.
These results indicated that the sporozoite egress from oocysts is a protease-dependent process and that CS-RIImut oocysts are more resistant to the trypsin treatment. CS is detected on the inner surface of the oocyst capsule [7–9]. Therefore, on their way out from the oocysts, sporozoites have to first traverse the CS protein layer beneath the capsule. The positively charged residues, arginines and lysines, of region II-plus are preferred substrates for certain cysteine proteases and serine proteases, such as trypsin. Therefore, it is possible that proteolysis of the CS protein layer underneath the capsule is required for sporozoite egress and is abolished in the mutant in which region II-plus has been substituted. Thus, a possible explanation for the defect of the mutant sporozoite is that the substitutions made in the region II-plus render CS protein more resistant to the putative protease. Trypsin treatment of the oocysts most likely cleaved the CS protein in many places, not only in region II-plus, since lysines and arginines are very abundant in the CS domains outside the repeats. Nevertheless, the subtle mutations introduced in CS region II-plus resulted in a clear difference in the kinetics of sporozoite release after treatment with the enzyme.
The exit of sporozoites from oocysts is most likely a stepwise process aimed at the sequential disruption of the capsule and the mosquito-derived basal lamina. In P. falciparum, PfCCp2 or PfCCp3, two secreted multidomain putative adhesive proteins, play an essential role in sporozoite release—in the absence of either protein, sporozoites were not released from the oocysts—but their localization and mechanism of action are unknown [22]. Our observations suggest that proteolysis of CS protein that lies beneath the capsule is likely to be an early event in sporozoite egress. Our hypothesis is supported by a recent finding that a papain-like cysteine protease egress cysteine protease 1 (ECP1) is required for sporozoite egress from oocysts [23]. Members of the papain family of cysteine proteases, similar to trypsin, consistently attack peptide bonds formed by lysine and arginine. We presume that ECP1 is only active when oocysts are “mature” and the sporozoites are ready to enter the hemolymph. At that particular time, only the CS protein that is beneath the capsule, but not CS on the sporozoite surface, is cleaved by ECP1, facilitating the egress of sporozoites from the oocysts. A possible explanation for this selectivity is that the enzyme that cleaves the capsule CS protein is also part of the capsule. We cannot exclude the possibility that the egress of sporozoites from oocysts is preceded by a proteolytic cascade and that ECP1 is only one of the participants.
CS-RIImut Oocysts Sporozoites Are Not Infective to Mammalian Hosts
As mentioned earlier, in vitro experiments strongly suggest that the region II-plus of CS protein plays an important role in the initial stages of sporozoite invasion of hepatocytes. Initial studies demonstrated that CS protein binds specifically to HSPGs in sections of human liver and that this binding is region II-plus dependent [24,25]. Synthetic peptides representing region II-plus specifically inhibit CS protein binding and sporozoite adhesion of HepG2 cells, the reference cell line that allows sporozoites to develop into mature exo-erythrocytic forms [26]. This inhibition is dependant on the downstream positively charged residues of region II-plus [26]. These and other findings (reviewed in [11,12]) give evidence that it is likely that the lysines and arginines (highly conserved in Plasmodium species) of region II-plus form ionic bonds with the negatively charged sulfate molecules of the HSPG glycosaminoglycan chains (GAGs). Our region II-plus mutant provided an opportunity to confirm the in vitro studies using a genetic approach. The wild-type and mutant sporozoites obtained by mechanical disruption of the midguts were incubated briefly with HepG2 cells to compare their binding to host cells. The sporozoites were obtained at a time when they are fully developed but not yet egressing from oocysts. There was a significant difference (~40%) in the binding of wild-type and mutant sporozoites (Figure 5). We emphasize that this assay was performed under static condition. The shear force generated by circulating blood in vivo should lead to more dramatic decrease in adhesion of mutant sporozoites to cells, as shown previously in vitro when the attachment assay was performed under rotating conditions [12]. Indeed, this is what we observed when we compared the infectivity of oocyst sporozoites from CS-WT and CS-RIImut to rats. The CS-WT and CS-RIImut sporozoites were injected intravenously into rats, and the pre-patent period of infection (time until the proportion of infected erythrocytes is less than 0.01%) measured. In two independent experiments, there was no infection in rats injected with 1–9 million CS-RIImut oocyst sporozoites (Table 2). In contrast, blood-stage parasites were detected in all rats injected with as few as 100,000 CS-WT oocyst sporozoites or 2,000 CS-WT salivary gland sporozoites [27].
Figure 5 Region II-plus is Important for Sporozoite Adhesion to HepG2 Cells
Midgut sporozoites of CS-WT and CS-RIImut (100,000 each) were added to confluent HepG2 cells. Adhesion is shown as the mean number of bound sporozoites in one microscopic field (400× magnification). Results are from three independent experiments.
Table 2 Infectivity of Sporozoites
Tewari et al. [28] also investigated the function of CS regions II-plus. In their study, the mutation was introduced in a P. berghei line in which the endogenous CS (PbCS) had been replaced by P. falciparum CS (PfCS), leading to a substantial decrease in parasite infectivity [29,30]. In their studies, Tewari et al. deleted the entire region II-plus of CS protein, including two of the four cysteines of the CS thrombospondin domain, and noted that the mutants did not enter the salivary glands. However, they did not measure the number of parasites in the mosquito hemocoel. It is possible therefore that the Tewari's mutant has the same phenotype as ours.
We conclude that the same CS protein motif participates in two different stages of the sporozoite lifecycle. Region II-plus is first required for sporozoite egress from oocysts. Later it is required for the invasion of the mammalian hepatocytes. It is tempting to speculate that receptors for region II-plus are identical in the mammalian host liver and in the oocysts capsule/basal lamina in the mosquito midgut, i.e., they are HSPGs. The presence of HSPGs has been documented in Drosophila [31], and HSPG core proteins are represented in the Anopheles genome (http://www.ensembl.org) but have not yet been characterized biochemically. Perhaps capsule/basal lamina HSPGs interact with the positively charged stretch of amino acids of CS region II-plus, and the disruption of those peptide bonds by the newly identified cysteine protease (ECP1), or other participating proteases in the same proteolytic cascade, is an early and necessary step for sporozoite egress from oocysts.
Materials and Methods
Parasite.
The parasite is a wild-type pyrimethamine-sensitive, gametocyte-producing clone of the P. berghei NK65 strain.
DNA construct and mutagenesis.
pQWCS-WT contains pUC19 backbone and 2.8 kb of CS cassette [32] cloned into XbaI and XhoI sites. Mutations were introduced into the CS coding region by using QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, California, United States). Primer1 (sense, 5′-GGTATAAGAGTTGCTGCAGCAGCAGGTTCAAATAAGAAAGC-3′) and its reverse and complement primer2 were used to mutate R290, K291, R292, and K293 to alanines. The resulting construct is named “pQWCS-RIImut.”
Targeting construct.
pMD205GFP is used as a backbone to generate targeting constructs, pRCS-WT and pRCS-RIImut. pMD205GFP contains the mutated copy of P. berghei DHFR-TS gene that confers resistance to pyrimethamine [33] and an Aequorea victoria green-fluorescent protein open reading frame (ORF) [34], and 2.2 kb and 0.55 kb of 5′- and 3′UTRs of P. berghei DHFR-TS. pQWCS-WT was digested with KpnI and XhoI to release the 2.1-kb fragment containing the CS cassette (0.6 kb of 5′UTR, 1-kb CS protein ORF and 0.5-kb 3′UTR). Subsequently, this fragment was cloned into pMD205GFP treated with same restriction enzymes to generate the intermediate construct pCS-WT. Targeting construct pRCS-WT was constructed by cloning a 0.6-kb BamHI-NotI fragment of CS 3′UTR (500–1,100 base pairs downstream of the stop codon) into the pCS-WT treated with the same restriction enzymes. Targeting construct pRCS-RIImut was constructed in the same way as pRCS-WT.
Parasite transfection and genotype analysis.
Schizonts were collected for transfection, and targeting constructs were introduced by electroporation as previous described [33]. Southern blotting was performed with the entire CS ORF and 0.6-kb 5′UTR as a probe. The probe was labeled with DIG-ddUTP by random priming, and the chemiluminescence was detected using CSPD (Roche, Basel, Switzerland). Specific amplification of the 5′ recombinant locus was preformed with a forward primer CS1 (sense: 5′-CTTTTTCACCCTCAAGTTGGG-3′, which hybridizes to the CS 5′UTR missing in the pRCS-WT/pRCS-RIImut), and a reverse primer PB103 (sense, 5′-TAATTATATGTTATTTTATTTCCAC-3′, which hybridizes to the 5′UTR of DHFR-TS). Specific amplification of the 3′ recombinant locus was preformed with a forward primer PB106 (sense, 5′-TGTGCATGCACATGCATGTA-3′, which hybridizes to the 3′UTR of DHFR-TS), and a reverse primer CS4 (sense, 5′-CGAAATAAGTTACTATTCGTGCCC-3′, which hybridizes to the CS 3′UTR missing in the pRCS-WT/pRCS-RIImut).
Mosquito infection and analysis of parasite development.
A. stephensi mosquitoes were fed on infected young Sprague-Dawley rats and dissected at various days PI. Midgut and salivary gland sporozoite populations were prepared from the various mosquito compartments and analyzed as previous described [27]. Hemolymph from each mosquito was perfused from the hemocoel with RPMI medium via air displacement from a micro-inoculation capillary inserted through the neck membrane and into the hemocoel. A small drop was made in the distal abdominal wall by gently removing the last two segments. The first three drops of perfusate (hemolymph and medium) from each mosquito were collected. Perfusate from at least 20 mosquitoes was collected. The number of sporozoites was determined using a haemocytometer.
Indirect immunofluorescence assays.
Oocyst sporozoites were collected at day 18 PI, centrifuged onto glass slides, and fixed with 4% paraformaldehyde for 20 min at room temperature. Sporozoites then were pre-incubated in PBS-3% BSA for 1 h at 37 °C followed by incubation of various anti-CS antibodies for 1 h at 37 °C. Bound anti-CS was detected with FITC-conjugated anti-mouse IgG.
Western blotting analysis of sporozoite lysates.
Protein samples were analyzed by SDS-PAGE and electrophoretically transferred to polyvinylidene difluoride membrane. CS-WT and CS-RIImut oocyst sporozoites on day 14 and 18 PI were collected, resuspended in SDS sample buffer, and incubated for 5 min at 70 °C prior to loading. The migrating bands were revealed with antibodies to P. berghei TRAP and CS protein, followed by horseradish peroxidase–coupled donkey anti-rabbit, and sheep anti-mouse IgG respectively, and visualized with enhanced chemiluminescence (ECL; Amersham Bioscience, Little Chalfont, United Kingdom).
Analysis of sporozoite infectivity.
To analyze sporozoite motility, sporozoites were incubated in 3% BSA-RPMI 1640 medium for 3 h prior to microscopic examination [3]. To determine the infectivity of sporozoites in vivo, young Sprague/Dawley rats were injected intravenously with sporozoite suspensions in RPMI 1640. The parasitemia of inoculated rodents was checked daily by a 10-min examination of a Giemsa-stained blood smear.
Sporozoite attachment assay.
A total of 100,000 midgut sporozoites were added and centrifuged down to confluent HepG2 cells. After 5-min incubation at 37 °C, cells were washed twice with PBS, and fixed with 4% formaldehyde. Adherent sporozoites were stained with a combination of anti-CS 3D11 and goat anti-mouse FITC antibodies. For each well, 25 microscopic fields were counted in duplicate using a 400× magnification.
In vitro assay of oocyst sporozoite release.
Intact midguts were dissected from either CS-WT– or CS-RIImut–infected mosquitoes. For each experiment, 10 midguts were incubated at 25 °C in 200-μl RPMI medium with or without trypsin (50 μg/ml; Sigma, St. Louis, Missouri, United States). At different time points, 10 μl was taken out after gentle shaking the tubes, and sporozoites were counted. At the end of the experiment, the midguts were ground in order to determine the mean number of remaining oocyst sporozoites per mosquito.
Transmission electron microscopy.
P. berghei (CS-WT and CS-RIImut) oocysts within mosquito midguts were fixed with 2.5% glutaraldehyde in 0.05 M phosphate buffer (pH 7.4) with 4% sucrose for 2 h. and then post-fixed in 1% osmium tetroxide for 1 h. After a 30-min en bloc stain with 1% aqueous uranyl acetate, the cells were dehydrated in ascending concentrations of ethanol and embedded in Epon 812. Ultrathin sections were stained with 2% uranyl acetate in 50% methanol and with lead citrate, and then examined in a Zeiss CEM902 electron microscope.
Immunoelectron microscopy.
P. berghei oocysts within mosquito midguts were fixed with 3% paraformaldehyde, 0.25% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4). Fixed samples were washed, dehydrated, and embedded in LR White resin (Polysciences, Warrington, Pennsylvania, United States) as described previously [7]. Thin sections were blocked in PBS containing 0.01% (v/v) Tween-20 and 5% (w/v) nonfat dry milk (PBTM). Grids were then incubated for 2 h at room temperature with the primary mouse anti-CS monoclonal antibody 3D11, and diluted 1:500 in PBTM. Normal mouse serum or PBTM were used as negative controls. After washing, grids were incubated for 1 h with 15-nm gold-conjugated goat anti-mouse IgG (Amersham Life Sciences), diluted 1:20 in PBS containing 1% (w/v) BSA and 0.01% (v/v) Tween-20, rinsed with Tween-20, and fixed with glutaraldehyde to stabilize the gold particles. Samples were stained with uranyl acetate and lead citrate, and then examined in a Zeiss CEM902 electron microscope.
This work was supported by a grant from the National Institutes of Health to Victor Nussenzweig. The authors wish to thank Dr. Elizabeth H. Nardin, Dr. Purnima Bhanot, and Dr. Jayne Raper for valuable comments on the manuscript. We also wish to thank Dr. Robert E. Sinden, Dr. Vandana Thathy, Dr. Purnima Bhanot, and members of Dr. Victor Nussenzweig's lab for discussions.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. QW and VN conceived and designed the experiments. QW and HF performed the experiments. QW, HF, and VN analyzed the data. QW and VN wrote the paper.
Abbreviations
CScircumsporozoite
ECP1egress cysteine protease 1
HSPGheparan sulfate proteoglycan
ORFopen reading frame
PIpostinfection
TSRthrombospondin type I repeats
==== Refs
References
Vanderberg J Rhodin J 1967 Differentiation of nuclear and cytoplasmic fine structure during sporogonic development of Plasmodium berghei
J Cell Biol 32 C7 C10 6034491
Sinden RE 2002 Molecular interactions between Plasmodium and its insect vectors Cell Microbiol 4 713 724 12427094
Vanderberg JP 1974 Studies on the motility of Plasmodium sporozoites J Protozool 21 527 537 4138523
Sinden RE 1974 Excystment by sporozoites of malaria parasites Nature 252 314
Adini A Krugliak M Ginsburg H Li L Lavie L 2001 Transglutaminase in Plasmodium parasites: Activity and putative role in oocysts and blood stages Mol Biochem Parasitol 117 161 168 11606226
Adini A Warburg A 1999 Interaction of Plasmodium gallinaceum ookinetes and oocysts with extracellular matrix proteins Parasitology 119 331 336 10581609
Aikawa M Atkinson CT Beaudoin LM Sedegah M Charoenvit Y 1990 Localization of CS and non-CS antigens in the sporogonic stages of Plasmodium yoelii
Bull World Health Organ 68 (Suppl) 165 171
Posthuma G Meis JF Verhave JP Hollingdale MR Ponnudurai T 1988 Immunogold localization of circumsporozoite protein of the malaria parasite Plasmodium falciparum during sporogony in Anopheles stephensi midguts Eur J Cell Biol 46 18 24 3294006
Hamilton AJ Suhrbier A Nicholas J Sinden RE 1988 Immunoelectron microscopic localization of circumsporozoite antigen in the differentiating exoerythrocytic trophozoite of Plasmodium berghei
Cell Biol Int Rep 12 123 129 3293804
Menard R Sultan AA Cortes C Altszuler R van Dijk MR 1997 Circumsporozoite protein is required for development of malaria sporozoites in mosquitoes Nature 385 336 340 9002517
Kappe SH Buscaglia CA Nussenzweig V 2004 Plasmodium sporozoite molecular cell biology Annu Rev Cell Dev Biol 20 29 59 15473834
Pinzon-Ortiz C Friedman J Esko J Sinnis P 2001 The binding of the circumsporozoite protein to cell surface heparan sulfate proteoglycans is required for Plasmodium sporozoite attachment to target cells J Biol Chem 276 26784 26791 11352923
Nunes A Thathy V Bruderer T Sultan AA Nussenzweig RS 1999 Subtle mutagenesis by ends-in recombination in malaria parasites Mol Cell Biol 19 2895 2902 10082556
Thathy V Fujioka H Gantt S Nussenzweig R Nussenzweig V 2002 Levels of circumsporozoite protein in the Plasmodium oocyst determine sporozoite morphology EMBO J 21 1586 1596 11927543
Yoshida N Potocnjak P Nussenzweig V Nussenzweig RS 1981 Biosynthesis of Pb44, the protective antigen of sporozoites of Plasmodium berghei J Exp Med 154 1225 1236 7026723
Sultan AA Thathy V Frevert U Robson KJ Crisanti A 1997 TRAP is necessary for gliding motility and infectivity of plasmodium sporozoites Cell 90 511 522 9267031
Sultan AA Briones MR Gerwin N Carroll MC Nussenzweig V 1997 Sporozoites of Plasmodium yoelii infect mice with targeted deletions in ICAM-1 and ICAM-2 or complement components C3 and C4 Mol Biochem Parasitol 88 263 266 9274888
Wickham ME Culvenor JG Cowman AF 2003 Selective inhibition of a two-step egress of malaria parasites from the host erythrocyte J Biol Chem 278 37658 37663 12857731
Salmon BL Oksman A Goldberg DE 2001 Malaria parasite exit from the host erythrocyte: A two-step process requiring extraerythrocytic proteolysis Proc Natl Acad Sci U S A 98 271 276 11114161
Rosenthal PJ 2004 Cysteine proteases of malaria parasites Int J Parasitol 34 1489 1499 15582526
Hanspal M Dua M Takakuwa Y Chishti AH Mizuno A 2002
Plasmodium falciparum cysteine protease falcipain-2 cleaves erythrocyte membrane skeletal proteins at late stages of parasite development Blood 100 1048 1054 12130521
Pradel G Hayton K Aravind L Iyer LM Abrahamsen MS 2004 A multidomain adhesion protein family expressed in Plasmodium falciparum is essential for transmission to the mosquito J Exp Med 199 1533 1544 15184503
Aly AS Matuschewski K 2005 A malarial cysteine protease is necessary for Plasmodium sporozoite egress from oocysts J Exp Med 202 225 230 16027235
Cerami C Frevert U Sinnis P Takacs B Clavijo P 1992 The basolateral domain of the hepatocyte plasma membrane bears receptors for the circumsporozoite protein of Plasmodium falciparum sporozoites Cell 70 1021 1033 1326407
Frevert U Sinnis P Cerami C Shreffler W Takacs B 1993 Malaria circumsporozoite protein binds to heparan sulfate proteoglycans associated with the surface membrane of hepatocytes J Exp Med 177 1287 1298 8478608
Sinnis P Clavijo P Fenyo D Chait BT Cerami C 1994 Structural and functional properties of region II-plus of the malaria circumsporozoite protein J Exp Med 180 297 306 8006589
Vanderberg JP 1975 Development of infectivity by the Plasmodium berghei sporozoite J Parasitol 61 43 50 1090717
Tewari R Spaccapelo R Bistoni F Holder AA Crisanti A 2002 Function of region I and II adhesive motifs of Plasmodium falciparum circumsporozoite protein in sporozoite motility and infectivity J Biol Chem 277 47613 47618 12244064
Nussenzweig V Menard R 2000 Analysis of a malaria sporozoite protein family required for gliding motility and cell invasion Trends Microbiol 8 94 97 10707054
Matuschewski K Nunes AC Nussenzweig V Menard R 2002
Plasmodium sporozoite invasion into insect and mammalian cells is directed by the same dual binding system EMBO J 21 1597 1606 11927544
Princivalle M de Agostini A 2002 Developmental roles of heparan sulfate proteoglycans: A comparative review in Drosophila, mouse and human Int J Dev Biol 46 267 278 12068947
Eichinger DJ Arnot DE Tam JP Nussenzweig V Enea V 1986 Circumsporozoite protein of Plasmodium berghei: Gene cloning and identification of the immunodominant epitopes Mol Cell Biol 6 3965 3972 2432395
de Koning-Ward TF Fidock DA Thathy V Menard R van Spaendonk RM 2000 The selectable marker human dihydrofolate reductase enables sequential genetic manipulation of the Plasmodium berghei genome Mol Biochem Parasitol 106 199 212 10699250
Sultan AA Thathy V Nussenzweig V Menard R 1999 Green fluorescent protein as a marker in Plasmodium berghei transformation Infect Immun 67 2602 2606 10225926
|
16201021
|
PMC1238744
|
CC BY
|
2021-01-05 12:11:28
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e9
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010009
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 1620101110.1371/journal.ppat.001001005-PLPA-RA-0038R1plpa-01-01-08Research ArticleMolecular Biology - Structural BiologyVirologyVirusesMus (Mouse)A Surface Groove Essential for Viral Bcl-2 Function During Chronic Infection In Vivo Role of Viral Bcl-2 BH1 Domain In VivoLoh Joy 1Huang Qiulong 2Petros Andrew M 2Nettesheim David 2van Dyk Linda F. 3Labrada Lucia 4Speck Samuel H 5Levine Beth 6Olejniczak Edward T 2Virgin Herbert W. IV1*
1 Departments of Pathology and Immunology and Molecular Microbiology, Washington University School of Medicine, St. Louis, Missouri, United States of America
2 Pharmaceutical Discovery Division, Abbott Laboratories, Abbott Park, Illinois, United States of America
3 Department of Microbiology, University of Colorado Health Science Center, Aurora, Colorado, United States of America
4 Department of Medicine, Columbia College of Physicians and Surgeons, New York, New York, United States of America
5 Division of Microbiology and Immunology, Yerkes Regional Primate Center, Emory University, Atlanta, Georgia, United States of America
6 Departments of Internal Medicine and Microbiology, University of Texas Southwestern Medical Center, Dallas, Texas, United States of America
McFadden Grant EditorUniversity of Western Ontario, Canada*To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 1 e105 5 2005 8 8 2005 Copyright: © 2005 Loh et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Antiapoptotic Bcl-2 family proteins inhibit apoptosis in cultured cells by binding BH3 domains of proapoptotic Bcl-2 family members via a hydrophobic BH3 binding groove on the protein surface. We investigated the physiological importance of the BH3 binding groove of an antiapoptotic Bcl-2 protein in mammals in vivo by analyzing a viral Bcl-2 family protein. We show that the γ-herpesvirus 68 (γHV68) Bcl-2 family protein (γHV68 v-Bcl-2), which is known to inhibit apoptosis in cultured cells, inhibits both apoptosis in primary lymphocytes and Bax toxicity in yeast. Nuclear magnetic resonance determination of the γHV68 v-Bcl-2 structure revealed a BH3 binding groove that binds BH3 domain peptides from proapoptotic Bcl-2 family members Bax and Bak via a molecular mechanism shared with host Bcl-2 family proteins, involving a conserved arginine in the BH3 peptide binding groove. Mutations of this conserved arginine and two adjacent amino acids to alanine (SGR to AAA) within the BH3 binding groove resulted in a properly folded protein that lacked the capacity of the wild-type γHV68 v-Bcl-2 to bind Bax BH3 peptide and to block Bax toxicity in yeast. We tested the physiological importance of this v-Bcl-2 domain during viral infection by engineering viral mutants encoding a v-Bcl-2 containing the SGR to AAA mutation. This mutation resulted in a virus defective for both efficient reactivation of γHV68 from latency and efficient persistent γHV68 replication. These studies demonstrate an essential functional role for amino acids in the BH3 peptide binding groove of a viral Bcl-2 family member during chronic infection.
Synopsis
Viruses can manipulate their hosts by expressing proteins that structurally and functionally resemble host cellular proteins. One important cellular process manipulated by viruses is apoptosis, a cell death program that is regulated by a family of Bcl-2-like proapoptotic and antiapoptotic proteins. Gammaherpesviruses encode Bcl-2 family proteins (v-Bcl-2) that may contribute to their ability to cause tumors and persist for the lifetime of their hosts. The authors solved the structure of the murine γ-herpesvirus 68 (γHV68) v-Bcl-2 and found that it is similar to cellular antiapoptotic proteins and that v-Bcl-2 uses the same mechanism as cellular Bcl-2 to bind to peptides from proapoptotic Bcl-2 family proteins. Furthermore, they found that a γHV68 virus expressing a mutated form of v-Bcl-2 that cannot bind to peptides from proapoptotic Bcl-2 family proteins is defective in its ability to cause chronic viral infection in mice. Thus, a specific structural feature and molecular mechanism of the v-Bcl-2 that is shared with host antiapoptotic Bcl-2 proteins is important for the function of this protein during viral infection. These findings enhance our understanding of the molecular mechanisms of chronic γ-herpesvirus infection, and suggest that targeting the functions of the v-Bcl-2 protein might have therapeutic benefit.
Citation:Loh J, Huang Q, Petros AM, Nettesheim D, van Dyk LF, et al. (2005) A surface groove essential for viral Bcl-2 function during chronic infection in vivo. PLoS Pathog 1(1): e10.
==== Body
Introduction
Bcl-2 family proteins are important regulators of cell death and other aspects of cell physiology such as glycolysis and calcium metabolism [1–5]. The Bcl-2 family can be divided into antiapoptotic proteins, which have in common four Bcl-2 homology domains (BH1–4), and proapoptotic proteins, which have either BH1–3 domains or only a BH3 domain [2–4,6–8]. A key mechanism by which this family regulates cell death in cultured cells involves a binding interaction between a hydrophobic groove on the surface of the antiapoptotic family protein and the BH3 domain of proapoptotic family members [2–4,9–13]. Furthermore, the importance of BH3 binding domains of antiapoptotic proteins has been shown in vivo in nematodes [14,15]. However, antiapoptotic Bcl-2 family proteins also interact with proteins outside of the Bcl-2 family, such as Aven, Apaf-1, Btf, Beclin1, Raf-1, calcineurin, tissue transglutaminase, FAST, and p53 [6,16–26]. The structural basis for interactions between these proteins and Bcl-2 family members, and the functional significance of these interactions, is less well understood than the structural basis and functional importance of interactions among Bcl-2 family members.
Certain viruses have acquired during evolution host genes that confer selective advantages. Viral proteins encoded by these genes often retain or even enhance advantageous functions of their host counterparts and lose functions that do not benefit the virus. Importantly, many viruses encode antiapoptotic Bcl-2 family proteins [7,27,28]. For example, viral Bcl-2 family proteins are encoded by all γ-herpesviruses, including the human viruses Epstein-Barr virus (EBV) and Kaposi's sarcoma-associated herpesvirus (KSHV), and the murine virus γ-herpesvirus 68 (γHV68) [7,27].
The conservation of v-Bcl-2 genes across the γ-herpesviruses suggests that these proteins play an important and evolutionarily conserved role in the pathogenesis of γ-herpesvirus infection. EBV, KSHV, and γHV68 all latently infect B cells [29–31], reactivate from latency, persistently replicate in both normal and immunocompromised hosts, and induce B cell lymphomas during chronic infection ([32] and unpublished data). The capacity to establish and reactivate from viral latency is essential for maintaining the lifelong infection that is characteristic of γ-herpesviruses. In addition, the capacity to persistently replicate at a low level despite the presence of active host immunity is a critically important aspect of γ-herpesvirus pathogenesis. Persistent replication contributes to viral spread from chronically infected hosts to new hosts, and may well contribute to tumorigenesis [33–35].
An in vivo role during infection has been demonstrated only for the γHV68 v-Bcl-2 [36]. The γHV68 v-Bcl-2 protein has been shown to inhibit apoptosis induced by Fas, TNFα, and Sindbis virus infection [37–39]. It is expressed during latency and has critical roles in reactivation from latency, the capacity to persistently replicate in immunocompromised mice such as those lacking interferon-γ (IFNγ), and in determining the number of latent cells early after infection [36,40,41]. Surprisingly, the γHV68 v-Bcl-2 has no role during acute infection of fibroblast cells in vitro or in vivo [36]. These data suggest that viral regulation of apoptosis is more important during chronic than acute γHV68 infection. Further support for this concept comes from studies showing that proapoptotic host molecules such as perforin, granzymes, and caspase 3 are unimportant during acute γHV68 infection but are critically important for limiting the amount of latent γHV68 infection [42,43]. Together, these data suggest that regulation of apoptosis by host and viral genes plays a critical role specific to chronic γHV68 infection.
These observations demonstrate a physiological role for a v-Bcl-2 family member in reactivation from viral latency and in persistent replication. The molecular mechanisms responsible for this role in vivo can be defined by evaluating the phenotypes of viruses expressing mutant forms of the v-Bcl-2 protein. We reasoned that analysis of a viral Bcl-2 protein using a combination of structural, biochemical, and pathogenesis studies would reveal whether viral and host Bcl-2 family members might function via the same mechanisms, and would clarify the importance of those mechanisms during infection. In this study, we identified the structure and biochemical function of a domain in γHV68 v-Bcl-2 predicted to be important for v-Bcl-2 function, and then tested the role of this domain in vivo. The γHV68 v-Bcl-2 shared with host antiapoptotic Bcl-2 family proteins the capacity to block apoptosis in primary lymphocytes induced by antigen receptor signaling, corticosteroids, and γ-irradiation, and to inhibit Bax toxicity in yeast. Solving the three-dimensional structure of the v-Bcl-2 revealed a functional hydrophobic surface groove that binds Bax and Bak BH3 peptides via a mechanism shared with host anti-apoptotic Bcl-2 family members. Mutations within this BH3 binding groove significantly decreased binding affinity for Bax peptide, abrogated inhibition of Bax toxicity in yeast, and ablated v-Bcl-2 function during chronic infection. These studies are the first to identify a specific domain of a viral Bcl-2 family protein that is essential for a physiological activity in vivo.
Results
γHV68 v-Bcl-2 Inhibits Apoptosis Induced by Diverse Apoptotic Stimuli in Primary Lymphocytes
The γHV68 v-Bcl-2 has been shown to block apoptosis in cultured cells in response to several proapoptotic stimuli [37–39]. We evaluated the antiapoptotic activity of the γHV68 v-Bcl-2 in vivo in primary cells by generating transgenic mice expressing v-Bcl-2 in thymocytes under the control of the lck proximal promoter (Figure 1A). While this approach can determine the effect of only ectopically expressed v-Bcl-2 in T lymphocytes, it has been successfully used to show the activity of both Bcl-2 and Bcl-xL in primary cells in vivo [44,45]. An added advantage of this system is the ability to study the activity of the v-Bcl-2 using multiple well-characterized lymphotoxic stimuli that have been shown to induce apoptotic cell death [46–51]. No antibodies for the γHV68 v-Bcl-2 are available, so we used quantitative RT-PCR to demonstrate specific γHV68 v-Bcl-2 mRNA expression in two transgenic mouse lines (v-Bcl-2A and v-Bcl-2B). v-Bcl-2A mice had 4- to 6-fold higher v-Bcl-2 expression in thymus and spleen than v-Bcl-2B mice (Figure 1B).
Figure 1 Transgenic Expression of v-Bcl-2 in Thymocytes
(A) Schematic illustration of the γHV68 genome, v-Bcl-2 genomic region, and transgene construct.
(B) Real-time RT-PCR quantitation of v-Bcl-2 expression in transgenic thymocytes (left) and splenocytes (right).
(C) Total number of thymocytes in v-Bcl-2 transgenic and nontransgenic mice.
(D) Percentage of CD4 or CD8 single-positive, double-positive (DP) and double-negative (DN) thymocytes in v-Bcl-2 transgenic and nontransgenic mice.
γHV68 v-Bcl-2 transgenic mice had higher numbers of total thymocytes than nontransgenic mice (Figure 1C), consistent with an effect of v-Bcl-2 on cell survival. However, v-Bcl-2 transgenic mice displayed no abnormalities in CD4 and CD8 T cell development in the thymus (Figures 1D and S1), nor were there any alterations in thymic architecture (unpublished data). Additionally, v-Bcl-2 transgenic thymocytes did not survive longer in explant culture than control cells (unpublished data). We next determined whether γHV68 v-Bcl-2 inhibits cell death induced by three diverse apoptotic stimuli in vivo. γHV68 v-Bcl-2 transgenic double-positive thymocytes were significantly more resistant than control thymocytes to apoptosis induced by dexamethasone, γ-irradiation, and CD3ɛ ligation (Figures 2 and S1). The increase in survival of transgenic double-positive thymocytes exceeds, and is therefore not attributable to, the slight increase in total thymocyte number in transgenic mice. Transgenic thymocytes from v-Bcl-2A mice were more resistant to dexamethasone than those from v-Bcl-2B mice, particularly at the highest doses of dexamethasone, consistent with higher levels of expression of v-Bcl-2 in these cells (see Figure 1B). No depletion of double-positive thymocytes was observed in transgenic or nontransgenic mice treated with vehicle or isotype antibody controls (unpublished data).
Figure 2 v-Bcl-2 Inhibits Cell Death in Thymocytes
(A) Survival of DP thymocytes from v-Bcl-2A (left) and v-Bcl-2B (right) mice 48 h following intraperitoneal injection with various doses of dexamethasone.
(B and C) Survival of DP thymocytes from v-Bcl-2A and v-Bcl-2B mice 48 h following (B) 250 rads of γ-irradiation and (C) intraperitoneal injection with 30 μg of anti-CD3ɛ antibody.
These results are similar to those previously found using host Bcl-2 and Bcl-xL transgenic mice [44,45,52] and show that the γHV68 v-Bcl-2 shares with its cellular counterparts the capacity to inhibit apoptosis induced by diverse proapoptotic stimuli in primary thymocytes. Together with previous studies in cultured cells [37–39], these data strongly support the concept that the γHV68 v-Bcl-2 blocks apoptosis by targeting a step or steps common to the death programs triggered by multiple cellular signals.
γHV68 v-Bcl-2 Is Structurally Homologous to Bcl-2 and Bcl-xL
The γHV68 v-Bcl-2 has limited amino acid homology to host Bcl-2 family proteins outside of the BH1 domain. Therefore, to understand the molecular basis of v-Bcl-2 function and the relationship between the γHV68 v-Bcl-2 and host Bcl-2 family members, we expressed and purified the protein and determined its structure by nuclear magnetic resonance (NMR) spectroscopy. For structural studies, we expressed amino acids 1–137 of the v-Bcl-2, removing the carboxy-terminal hydrophobic domain. The overall fold of v-Bcl-2 (Figure 3) was very similar to that of Bcl-xL (Figure 3B) [53] and other Bcl-2 family proteins [2]. The core of v-Bcl-2 was formed by two predominantly hydrophobic central helices, α5 and α6, which were flanked on one side by α3 and α4 and on the other side by α1, α2, and α7. The average minimized coordinates for the γHV68 v-Bcl-2 have been deposited with the Protein Data Bank (PDB [http://www.rcsb.org/pdb/] accession code 2ABO).
Figure 3 Solution Structure of γHV68 v-Bcl-2
(A and B) Ribbon representation of (A) γHV68 v-Bcl-2 and (B) Bcl-xL with BH1, BH2, BH3, and BH4 regions in magenta, red, green, and yellow, respectively. Helices are numbered with respect to Bcl-xL.
(C) Connolly surface for γHV68 v-Bcl-2 calculated using a probe radius of 1.4 Å. Residues are colored as follows: Leu, Val, Ile, Phe, Tyr, Trp, Met, and Ala in yellow; Arg, Lys, and His in blue; Asp and Glu in red; all others in gray. The hydrophobic groove is indicated by an arrow.
(D) Structural and sequence alignment of KSHV and γHV68 v-Bcl-2 proteins with Bcl-2 and Bcl-xL.
v-Bcl-2 clearly contained structural elements corresponding to the BH1, BH2, BH3, and BH4 regions of Bcl-xL, despite the fact that the v-Bcl-2 BH4 region was not readily apparent in sequence alignments with Bcl-xL (Figure 3D). The BH1 region (in magenta) was composed of the carboxy-terminal end of α4, the loop connecting α4 to α5, and the amino-terminal end of α5. The BH2 region (in red; Figure 3D) was composed of α7, and the BH3 (in green) and BH4 (in yellow) regions were composed of α2 and α1, respectively. The two side chain conformations of Trp10 in α1 generated two distinct sets of resonances for residues in this region, and the resonances corresponding to the predominant conformation were used for these analyses. One notable difference between the γHV68 and KSHV v-Bcl-2 proteins and host antiapoptotic Bcl-2 family proteins is the lack of an extended loop between α1 and α2, which in host Bcl-2 contains sites for caspase cleavage and regulatory phosphorylation [1,2,54].
An alignment revealing the structural and amino acid conservation between Bcl-2, Bcl-xL, the γHV68 v-Bcl-2, and the KSHV v-Bcl-2 proteins is shown in Figure 3D. Aside from the lack of the α1/α2 loop in the v-Bcl-2, this structural alignment reveals how a similar overall structure relates to relatively poor amino acid homology in regions outside of the BH1 domain. This striking conservation of structure between v-Bcl-2 and other Bcl-2 family members, despite only low-level amino acid homology outside of the BH1 domain, supports the concept that strong evolutionary pressure has led to retention of the overall structure, and suggests that the mechanisms of v-Bcl-2 action might therefore be similar to those of its cellular counterparts.
γHV68 v-Bcl-2, Like Antiapoptotic Host Bcl-2 Family Proteins, Has a Hydrophobic Surface Groove and Binds BH3 Peptides from Proapoptotic Bcl-2 Family Members
γHV68 v-Bcl-2 had an elongated surface hydrophobic groove composed of amino acid residues from α2, α3, α4, and α5, similar to the groove observed on Bcl-2 and Bcl-xL (Figures 3C and S2). The orientation of α3 and α4 defined the bottom of this groove, which is the predicted binding site for BH3 domains of proapoptotic Bcl-2 family proteins [2,53]. Conservation of this hydrophobic groove in the γHV68 v-Bcl-2 suggests that it may, like host Bcl-2, bind BH3 domains of proapoptotic Bcl-2 family proteins. One approach to defining the biochemical properties of the surface groove of an antiapoptotic Bcl-2 family member is to measure the affinity of the protein for BH3 peptides from proapoptotic Bcl-2 family members [2–4]. For example, the BH3 domain of Bak is necessary and sufficient for effecting cell death, and mutations in Bak BH3 peptide that reduce binding to Bcl-xL also abrogate interaction of full-length Bak with Bcl-xL [55,56].
We therefore used NMR to analyze binding of BH3 peptides from proapoptotic Bcl-2 family proteins Bax, Bak, and Bad to purified γHV68 v-Bcl-2. We focused on Bak and Bax, since γHV68 v-Bcl-2 blocks apoptosis induced by diverse stimuli (see Figure 2), and Bak and Bax are essential for cell death induced by many stimuli [57–59]. We selected the Bad BH3 peptide for analysis since Bad is a potently proapoptotic BH3-only Bcl-2 family member that induces cell death via interaction with antiapoptotic Bcl-2 family members [8]. Both Bax and Bak peptides were in slow exchange with γHV68 v-Bcl-2 on the NMR timescale, indicating Kd values of less than 5 μM. In contrast, Bad BH3 peptide bound weakly to γHV68 v-Bcl-2, with a Kd of greater than 300 μM. γHV68 v-Bcl-2 therefore differs significantly from Bcl-2 and Bcl-xL, which bind the Bad BH3 peptide with an affinity of 0.6–15 nM [9]. These results show that the γHV68-v-Bcl2 binds BH3 peptides from some, but not all, BH3 domain containing proapoptotic Bcl-2 family members.
A Common Mechanism for Binding of BH3 Peptides by γHV68 v-Bcl-2 and Bcl-xL
The retention of a hydrophobic groove and the capacity to bind BH3 peptides by γHV68 v-Bcl-2 suggested that γHV68 v-Bcl-2 shares with Bcl-2 and Bcl-xL a common capacity for binding BH3 peptides. However, the lack of Bad BH3 peptide binding led us to question whether the mechanism of BH3 peptide binding is shared between host Bcl-2 family proteins and the γHV68 v-Bcl-2. To directly address this question, we used nuclear Overhauser effect (NOE) experiments (Figure 4). We found that the Bak BH3 peptide bound to γHV68 v-Bcl-2 and Bcl-xL in the same orientation ([53] and unpublished data), with contact occurring primarily between hydrophobic residues of v-Bcl-2 and Bak peptide. γHV68 v-Bcl-2 residues with NOE contacts to the Bak peptide (Figure 4A) were structurally homologous to those in the complex of Bcl-xL and Bak peptide, in which Arg139 of Bcl-xL forms a key contact with Asp83 of the Bak peptide [53].
Figure 4 Mutagenesis of γHV68 v-Bcl-2 BH3 Binding Groove
(A) Ribbon representation of γHV68 v-Bcl-2 showing residues that contact Bak peptide (in green and R87) and the SGR residues that were mutated (in magenta).
(B) Growth of yeast following transformation with the indicated constructs (above, bar graph) and Western blot of SGR/AAA and wild-type v-Bcl-2 from transformed yeast (below, blot).
(C) Overlay of the 15N-HSQC spectra of SGR/AAA (red contours) and wild-type (black contours) v-Bcl-2 showing conservation of structure between mutant and wild-type proteins.
Arg139 is highly conserved among Bcl-2 family proteins, and mutation of Asp83 in the Bak peptide to alanine (D83A) reduces its affinity for Bcl-xL by more than 120-fold. The γHV68 v-Bcl-2 residue that is structurally homologous to Arg139 of Bcl-xL is Arg87, which is a conserved amino acid in the v-Bcl-2 BH1 domain (see Figure 3D). We tested the importance of this Arg:Asp interaction in the binding of γHV68 v-Bcl-2 to Bak peptide in an NMR-based titration of v-Bcl-2 with D83A mutant Bak peptide. We observed no binding of D83A mutant Bak peptide to γHV68 v-Bcl-2 over the concentration range of the titration, indicating a Kd of more than 300 μM. Thus, Arg87 of γHV68 v-Bcl-2 played a key role, similar to that of Bcl-xL Arg139, in binding to Asp83 of the Bak BH3 peptide. These data show that the molecular mechanisms of BH3 peptide binding are shared between γHV68 v-Bcl-2 and host antiapoptotic proteins.
Residues in the γHV68 v-Bcl-2 BH3 Binding Groove Are Required for v-Bcl-2 Binding to BH3 Peptides and Inhibition of Bax Toxicity in Yeast
The conservation of the mechanism by which γHV68 and host Bcl-2 family members bind BH3 peptides suggested that the γHV68 v-Bcl-2 would also inhibit the function of proapoptotic Bcl-2 family members. We tested this hypothesis using a Bax toxicity assay in yeast (Figure 4). Inhibition of Bax toxicity in yeast by Bcl-2 family proteins depends on amino acids within the BH3 binding groove of the antiapoptotic family member and, at least in part, on heterodimerization with Bax [60–63].
The γHV68 v-Bcl-2 inhibited Bax-mediated death as effectively as Bcl-2 (Figure 4B). No inhibition of Bax-mediated death was observed in empty vector control transformations (unpublished data). Because structural data suggested the importance of the v-Bcl-2 Arg87, we tested the effect of mutating Arg87 and two adjacent residues (i.e., Ser85-Gly86-Arg87; in magenta in Figure 4A) to alanine (hereafter termed SGR/AAA) on the capacity of the v-Bcl-2 to inhibit Bax-mediated death. Mutations in this region have been shown to alter the antiapoptotic activity of both host and other viral Bcl-2 proteins [3,4,7,27]. We mutated all three amino acids based on preliminary studies demonstrating partial effects of single mutations in this region (unpublished data). The SGR/AAA mutation abrogated γHV68 v-Bcl-2 inhibition of Bax-mediated toxicity in yeast despite equivalent expression levels of wild-type and SGR/AAA mutant v-Bcl-2 protein (Figure 4B). Moreover, SGR/AAA mutant γHV68 v-Bcl-2 bound Bax peptide at least 1500-fold less well than wild-type γHV68 v-Bcl-2, as measured by NMR (Kd of greater than 300 μM compared to less than 5 μM for wild-type γHV68 v-Bcl-2). Analogous mutations in Bcl-2 and Bcl-xL similarly abrogate antiapoptotic activity and binding to proapoptotic proteins [64,65]. Importantly, a comparison of the 15N- heteronuclear single-quantum coherence (HSQC) spectra of wild-type and SGR/AAA mutant γHV68 v-Bcl-2 confirmed that SGR/AAA mutant v-Bcl-2 retains the overall fold of wild-type v-Bcl-2 (Figure 4C), indicating that functional defects in SGR/AAA mutant γHV68 v-Bcl-2 are attributable specifically to alterations in the BH3 binding groove rather than to misfolding of the mutant protein. Thus, residues that lie in the γHV68 v-Bcl-2 BH3-binding groove were essential for both binding of Bax BH3 peptide and for inhibition of Bax-mediated toxicity in yeast.
A Functional BH3 Binding Groove Is Essential for Efficient Reactivation of γHV68 from Latency and for Persistent γHV68 Replication
The conservation of the structure and function of the γHV68 v-Bcl-2 BH3 binding groove led us to hypothesize that this groove, and the conserved Arg87 required for interaction with BH3 peptides from proapoptotic Bcl-2 proteins and for inhibition of Bax toxicity in yeast, would be essential for the function of the v-Bcl-2 in vivo during γHV68 infection. To test this hypothesis, we compared the phenotypes of wild-type γHV68 virus, a γHV68 virus lacking the entire v-Bcl-2 due to a null mutation in the M11 gene encoding v-Bcl-2 (Δv-Bcl-2) [36], and γHV68 viruses expressing the SGR/AAA mutant form of the v-Bcl-2.
We generated two independent isolates of γHV68 containing the SGR/AAA mutation in the M11 gene (SGR/AAA.1 and SGR/AAA.2; Figure 5A and 5B). SGR/AAA v-Bcl-2 mutant viruses displayed normal growth in fibroblast cells in vitro (Figure 5C), and replicated normally in the spleen (Figure 5D) and liver (unpublished data) of infected mice at 4 and 9 d post-infection (dpi). These results were expected, because the Δv-Bcl-2 γHV68 mutant virus, which lacks v-Bcl-2 expression, also exhibits completely normal replication during acute infection [36]. Together, these data indicate that the v-Bcl-2 does not have an essential function during γHV68 acute infection. The use of two independently generated mutants to analyze the phenotype attributable to the SGR/AAA mutation is an accepted standard in the field [36,66–68]. The equivalent phenotypes of SGR/AAA.1 and SGR/AAA.2 in multiple assays argues against the possibility that these phenotypes are due to chance mutations elsewhere in the viral genome.
Figure 5 SGR/AAA Mutant Viruses Replicate Normally In Vitro and In Vivo
(A) Shown are schematic illustrations of the genomes of γHV68, v-cyclin.LacZ, and SGR/AAA mutant viruses with the engineered PstI site underlined, and v-Bcl-2 containing the SGR/AAA mutation.
(B) Southern blot of γHV68, v-cyclin.LacZ, and SGR/AAA mutant viruses. Expected bands (kb) for PstI/BamHI digest: γHV68, 1.5, 1.3, and 1.2; v-cyclin.LacZ, 1.3, 1.2, 1.1, and 0.05; and SGR/AAA, 1.5, 1.2, 0.8, and 0.6. Expected bands (kb) for PstI/HindIII digest: γHV68, 7.5, 1.2, 0.9, and 0.4; v-cyclin.LacZ, 7.1, 4.3, 1.2, 0.9, 0.4, and 0.07; and SGR/AAA, 7.5, 1.2, 0.6, 0.39, and 0.36.
(C) Multistep growth curves of SGR/AAA mutants and γHV68.
(D) Acute splenic titers at 4 or 9 dpi from mice infected with SGR/AAA mutants.
As the Δv-Bcl-2 mutant γHV68 virus has defects in reactivation from latency and in persistent replication [36], we tested the effects of the SGR/AAA v-Bcl-2 mutation on chronic γHV68 infection of IFNγ−/− mice. IFNγ−/− mice clear acute γHV68 infection normally [42,69], but exhibit increases in persistent replication and in the efficiency with which latently infected cells reactivate virus ex vivo [36,42]. Both Δv-Bcl-2 and SGR/AAA mutant γHV68 viruses exhibited small but statistically significant decreases in the frequency of cells that reactivate from viral latency ex vivo compared to wild-type γHV68 (Figure 6, Table 1). Furthermore, SGR/AAA mutant γHV68 viruses, like the Δv-Bcl-2 mutant γHV68 virus, showed no persistent replication at 42 dpi (Figure 6B, Table 1). The effect of the SGR/AAA v-Bcl-2 mutation on persistent γHV68 virus replication was significant, but less than the effect of a null mutation in the gene at 16 dpi. This partial phenotype was not due to reversion of the mutation in vivo, since 20 individual SGR/AAA mutant γHV68 viruses present in these mice were confirmed to have the mutation by nucleotide sequencing (unpublished data). Together, these data demonstrate that amino acids that are critical for the capacity of γHV68 v-Bcl-2 to bind BH3 peptides via the BH3 binding groove are essential for optimal reactivation from latency and for the capacity to persistently replicate in tissues at a low level after acute viral infection is contained by the immune response.
Figure 6 SGR/AAA Mutant Viruses Exhibit Defects in Chronic Infection In Vivo
Ex vivo reactivation (A) and persistent replication (B) of γHV68, Δv-Bcl-2, or SGR/AAA mutant viruses at 16 or 42 dpi of IFNγ−/− mice. No significant differences were observed between the two independent isolates of SGR/AAA mutants, and data from these two groups were pooled.
Table 1 Summary of Chronic Infection Results
Discussion
In this study, we combined structural and biochemical approaches to identify a BH3 peptide binding groove on the surface of the γHV68 v-Bcl-2 protein and amino acids within this groove that are essential for the capacity of the v-Bcl-2 to bind BH3 peptides. These same amino acids are essential for the capacity of the γHV68 v-Bcl-2 to block cell death in yeast induced by expression of the proapoptotic Bcl-2 family protein Bax, and for the contribution of the v-Bcl-2 to two fundamentally important aspects of γHV68 pathogenesis: reactivation from viral latency and persistent replication. Together, these data show that this domain of the v-Bcl-2 is essential for function in vivo, and support the concept that the γHV68 v-Bcl-2 functions in vivo by binding BH3 domains of antiapoptotic Bcl-2 family members.
Presence of a Functional BH3 Binding Groove on γHV68 v-Bcl-2
We found that the γHV68 v-Bcl-2 shares with Bcl-2, Bcl-xL, and the KSHV v-Bcl-2 a BH3 binding groove [9]. Each of these proteins interacts with BH3 peptides from Bak and Bax, but the viral proteins differ from the host proteins in their inability to bind the Bad BH3 peptide [2,9]. Interestingly, the preference for Bak and Bax peptides over BH3 peptides from other proapoptotic family members is also seen for the adenovirus v-Bcl-2 [7].
These differences in BH3 peptide specificity between viral and host proteins have implications for how v-Bcl-2 proteins inhibit apoptosis. Some models of antiapoptotic Bcl-2 family protein function require that they heterodimerize with BH3-only proteins such as Bad in order to sequester these proapoptotic molecules away from Bak and Bax [3,4,57,59]. While we have not examined binding of BH3 peptides from BH3-only proteins other than Bad, the striking lack of binding of Bad BH3 peptide to γHV68 v-Bcl-2 suggests that the viral protein may not use this mechanism. Instead, these data support an alternative model that has also been proposed for cellular antiapoptotic Bcl-2 family proteins, but that is not generally accepted [3,4]. In this model, the antiapoptotic proteins directly target Bax and Bak rather than sequestering BH3-only proteins.
An interesting question is why KSHV and γHV68 v-Bcl-2 proteins bind the BH3 peptide from Bad with much lower affinity than is observed for Bcl-2 or Bcl-xL [9]. In this property, the γHV68 and KSHV v-Bcl-2 proteins resemble cellular Mcl-1, which also fails to bind Bad BH3 peptide and is important for development of B cells, one of the cell types in which γHV68 establishes latency [70,71]. One possible explanation for the lack of binding to the Bad BH3 peptide is that v-Bcl-2 binding to Bad is detrimental to the virus. Bad was recently found to be essential for assembly of a holoenzyme that regulates mitochondrial respiration, and Bad deficiency results in defects in glycolysis and in aberrant glucose homeostasis [5]. Thus, v-Bcl-2 proteins may bind poorly to Bad, and perhaps other BH3-only Bcl-2 family proteins, to avoid interfering with functions of BH3 domain containing proteins in cellular processes other than induction of apoptosis. This would be important for a virus, such as γHV68, that expresses the v-Bcl-2 in latently infected cells [40], and would presumably benefit by not having v-Bcl-2 interfere with cellular processes essential for cell viability.
γHV68 v-Bcl-2 Inhibits Apoptosis Induced by a Wide Array of Proapoptotic Stimuli
γ-herpesviruses maintain latent infection in lymphocytes for the life of the host, and must survive in the face of the immune response that effectively clears acute infection. These viruses then reactivate from latency and productively infect cells, yielding low levels of infectious virus in tissues, a process we refer to as persistent replication. It seems reasonable to speculate that latently infected cells may be exposed to multiple different proapoptotic stimuli over time. Furthermore, during the process of reactivation and persistent replication, infected cells are likely subjected to immune attack by lymphocytes that use proapoptotic cytokines or granule contents to kill infected cells. It is therefore not surprising that a protein such as the γHV68 v-Bcl-2, which is important during chronic infection, has the capacity to inhibit cell death induced by a wide variety of different stimuli, including antigen receptor crosslinking, corticosteroids, and γ-irradiation. In addition to the studies presented here, observations in cultured cells show that the γHV68 v-Bcl-2 inhibits apoptosis induced by TNFα, Fas, and Sindbis virus infection [37–39]. While it is possible that the γHV68 v-Bcl-2 utilizes different mechanisms to inhibit each of these death induction pathways, it is more reasonable to conclude that the viral protein inhibits a step in the death execution pathway common to all of these stimuli. In this regard, the capacity of the γHV68 v-Bcl-2 to bind BH3 peptides from Bax and Bak and to inhibit Bax-induced cell death in yeast is of particular interest. Bax and Bak are essential for induction of apoptosis through multiple pathways, and function therefore as a common gateway to the downstream steps in cell death execution [57–59]. Targeting these proteins would therefore have the advantage of protecting the infected cell from diverse proapoptotic insults.
Targeting Bax and Bak may be a common strategy for many viruses, since Bax and Bak are required for apoptosis during adenovirus infection, and the adenovirus v-Bcl-2 E1B 19K targets Bak and Bax for antiapoptotic function in cultured cells [7,72,73]. Furthermore, we believe that this is likely a general mechanism, since a role for BH1-domain amino acids in binding BH3 peptides has also been demonstrated for the adenovirus v-Bcl-2 and the African swine fever virus v-Bcl-2 family protein A179L [7]. Given the overall poor amino acid homology between viral Bcl-2 proteins and cellular Bcl-2 family members, the conservation of a mechanism for BH3 peptide binding between cellular and viral Bcl-2 family proteins strongly supports the hypothesis that this property of the v-Bcl-2 is important for function. This concept is supported by data provided here demonstrating the essential role of this domain in the function of the γHV68 v-Bcl-2 in vivo.
Is the Critical In Vivo Role of Amino Acids in the BH3 Binding Groove Explained by v-Bcl-2 Binding to Proapoptotic Bcl-2 Family Proteins?
We found that amino acids in the BH3 binding groove of the γHV68 v-Bcl-2 are essential for in vivo function of the v-Bcl-2 protein. The major known function of the BH3 binding groove is interaction with BH3 domains of proapoptotic Bcl-2 family proteins, providing the basis for our conclusion that our data are consistent with a role for this biochemical function of v-Bcl-2 during chronic infection. However, there are other potential explanations for our results. It is possible that binding of BH3 domains is important for function, but that binding Bcl-2 family members is not. For example, non-Bcl-2 family proteins can contain BH3-like domains [16,17], and it is possible that the γHV68 Bcl-2 functions by binding to these proteins rather than to Bcl-2 family proteins.
Alternatively, the portion of the BH3 peptide binding groove that we mutated in the studies presented here might interact with other proteins that do not contain BH3 domains. Support for this possibility comes from the structure of the EBV v-Bcl-2 BHRF1, which is antiapoptotic but does not have an intact BH3 binding groove and does not measurably bind BH3 peptides [10]. It is possible that this v-Bcl-2, which lacks a BH3 binding groove, inhibits apoptosis by a mechanism distinct from that used by the γHV68 and KSHV v-Bcl-2 proteins, both of which have functional BH3 binding grooves. For example, the EBV v-Bcl-2 might inhibit apoptosis via interactions with other proteins such as Aven, Apaf-1, Btf, Beclin1, Raf-1, calcineurin, tissue transglutaminase, FAST, or p53 [6,16–26], all of which have been shown or suggested to to interact with host antiapoptotic Bcl-2 family members. Perhaps a capacity to interact with one or more of these proteins despite the SGR/AAA mutation explains the partial phenotype of the SGR/AAA mutant γHV68 viruses in persistent replication observed 16 dpi. It will be interesting to compare the antiapoptotic mechanisms of the γHV68 and KSHV v-Bcl-2 proteins with that of BHRF1 in order to determine whether interaction with proteins that do not contain BH3 is important for v-Bcl-2 function.
Differences between Host and Viral Antiapoptotic Proteins
While γHV68 v-Bcl-2 shares an overall fold with its host counterparts, there are differences between the γHV68 v-Bcl-2 and host Bcl-2 family members, most notably the absence of a long loop between α1 and α2 [2]. Interestingly, the KSHV and EBV v-Bcl-2 proteins also lack this loop [9,10]. In cellular Bcl-2 family proteins, this loop contains sites for caspase cleavage and regulatory phosphorylation [1,2,54], and a portion of the loop has been predicted to be involved in p53 binding [19].
In contrast to host Bcl-2 family proteins [54,74], the KSHV and EBV v-Bcl-2 proteins are resistant to caspase cleavage [27,37], a process that generates proapoptotic products. Although γHV68 v-Bcl-2 also has a shortened α1/α2 loop, it does contain a caspase cleavage site (LDCV; see Figure 3D) and can be cleaved by caspases [37]. However, the predicted cleavage product does not have proapoptotic activity [37]. The absence of the α1/α2 loop in KSHV v-Bcl-2 also confers resistance to regulatory phosphorylation by the KSHV v-cyclin-CDK6 complex, which inactivates Bcl-2 [75,76]. Based on its structure, the γHV68 v-Bcl-2 is also expected to be refractory to regulatory phosphorylation. Thus, our structural analysis of the γHV68 v-Bcl-2 provides additional evidence that v-Bcl-2 proteins have evolved strategies to protect themselves from host cell regulatory mechanisms. The lack of an intact α1/α2 loop in v-Bcl-2 proteins decreases their susceptibility to inactivation via caspase cleavage or phosphorylation, an evolutionary strategy that we and others [2,6,7] speculate provides an advantage for the viruses by removing the antiapoptotic function of viral Bcl-2 family proteins from host cell control.
Implications for Control of Chronic γ-Herpesvirus Infection
Most of the disease burden of human γ-herpesvirus infection occurs during chronic infection. Studies here demonstrating that amino acids in the BH3 binding groove of the v-Bcl-2 are essential for chronic viral infection raise the possibility that pharmacologic inhibition of v-Bcl-2 BH3 binding groove function might inhibit chronic infection. There may be sufficient differences between viral and host proteins' BH3 binding grooves to allow specific targeting of v-Bcl-2 proteins. Therefore, v-Bcl-2 proteins may provide a suitable therapeutic target for preventing or ameliorating γ-herpesvirus disease.
Materials and Methods
Animals.
Mice were housed and bred in a specific pathogen-free environment at Washington University School of Medicine in accordance with all federal and university policies. v-Bcl-2 transgenic mice were generated as previously described [77]. Briefly, the v-Bcl-2 ORF (genome coordinates 103418–103930 bp) was PCR-amplified and ligated into the BamHI site of the p1017 vector containing the hGH enhancer and the lck proximal promoter [78,79]. The construct was confirmed by DNA sequencing, and the DNA fragment used for microinjection was isolated by SpeI restriction digest followed by gel purification. B6 embryo manipulations were performed as previously described [77]. Transgene-positive F1 mice were identified by PCR and confirmed by Southern blot (unpublished data). Mice were maintained as heterozygotes on the B6 background and were genotyped by PCR using primers specific for the hGH enhancer within the transgene construct [77]. Nontransgenic littermates were used as negative controls for v-Bcl-2 transgenic mice. IFNγ−/− mice on the B6 background were obtained from Jackson Laboratories (Bar Harbor, Maine, United States) and bred at Washington University. B6 mice were purchased from Jackson Laboratories. All studies were performed using age- and sex-matched mice between 8 and 12 wk of age.
Quantitative RT-PCR.
Thymi and spleens from euthanized mice were disrupted over a 100 μm Nytex filter to generate single-cell suspensions. Total RNA was prepared from dissociated cells, and 2 μg was used in reverse transcription reactions to generate cDNA template for quantitative real-time PCR. Six replicate reactions of real-time PCR were performed per sample using a BioRad iCycler (BioRad, Hercules, California, United States) [80]. v-Bcl-2 specific primers were 5′-TAACATTGACCCAGGAGTTTAG-3′ and 5′-CGAGGTGAAAAGTTTGGAC-3′, and control reactions utilized primers specific for 18S RNA. No positive signal was detected in RNA samples to which reverse transcriptase was not added (unpublished data). Three transgene-positive mice from each founder line were analyzed, and nontransgenic samples consisted of one transgene-negative mouse from each founder line. v-Bcl-2 message copy number was calculated from a standard curve generated using 101 to 105 copies of a plasmid containing v-Bcl-2, and was normalized to 18S RNA levels.
Flow cytometry.
Dissociated thymocytes were pre-incubated with anti-CD32/CD16 antibodies (Caltag Laboratories, Burlingame, California, United States; #MM7400) and stained with FITC-conjugated anti-CD4 (Pharmingen, San Diego, California, United States; #09424A) and tricolor-conjugated anti-CD8α (Caltag, #RM2206) antibodies. Cells were washed, fixed, and analyzed on a FACS Calibur. Data were analyzed using CellQuest (Becton Dickinson, Palo Alto, California, United States) and represent the mean ± SEM from at least two independent experiments. Statistical significance was determined using the Mann-Whitney test.
Bax toxicity in yeast.
Yeast strain W303 was a generous gift from S. Zheng. Murine Bax and Bcl-2 clones were generous gifts from S. Korsmeyer. Bax was cloned into p425 GPD (ATCC, Manassas, Virginia, United States; #87359) and Bcl-2 was cloned into p426 GPD (ATCC #87361). SGR/AAA mutant v-Bcl-2 was generated using the ExSite PCR-Based Site-Directed Mutagenesis Kit (Stratagene, La Jolla, California, United States). Wild-type and SGR/AAA v-Bcl-2 were amino-terminal HA-tagged and cloned into p426 GPD. All constructs were verified by DNA sequencing, and 1 μg of each construct was used to transform yeast using the lithium acetate method. The number of colonies were counted 3 d following plating on selection media and is expressed as a percentage of that for a control transformation with p425 GPD and p426 GPD empty vectors. Transformations of yeast with each single construct resulted in growth only on the appropriate selection media (unpublished data). Data represent the mean ± SEM from four independent experiments, and statistical significance was determined using the Mann-Whitney test. For Western blots, total protein from transformed yeast was prepared using Y-PER Yeast Protein Extraction Reagent (Pierce Biotechnology, Rockford, Illinois, United States), resolved on an SDS-PAGE gel, and detected with an anti-HA antibody (Covance, Princeton, New Jersey, United States; #MMS-101P).
Cell culture and viruses.
NIH 3T12 cells and MEFs were cultured in DMEM containing 10% fetal calf serum (D10). SGR/AAA mutant viruses were generated by homologous recombination as previously described [36]. Briefly, a targeting construct was generated by replacing wild-type v-Bcl-2 sequence with that for the SGR/AAA mutant in pL3700, which contains a 3,723-bp BamHI/BsrGI fragment of the γHV68 genome (genome coordinates 101,654–105,377 bp). The mutation generates a new PstI restriction site, which was used to distinguish wild-type from SGR/AAA viruses by Southern blot. The targeting construct was verified by DNA sequencing and cotransfected into 3T12 cells with viral DNA from v-cyclin.LacZ, which contains a LacZ expression cassette inserted into the adjacent v-cyclin ORF [81]. Recombinant “white” viruses were isolated following X-Gal staining and screened by Southern blot for the SGR/AAA mutation in v-Bcl-2. Two independent isolates were generated from separate cotransfections, and each isolate was subjected to three rounds of plaque purification. Fifteen individual plaques from each isolate were genotyped by Southern blot to confirm the absence of contamination with either wild-type or parental v-cyclin.LacZ viruses (unpublished data), and one of the tested plaques was used to generate virus stocks of each isolate. γHV68 clone WUMS (ATCC VR1465), Δv-Bcl-2 [36], and SGR/AAA mutants were subcultured and titered by plaque assay on 3T12 cells as previously described [82]. For in vitro multistep growth curves, 3T12 cells were infected at a MOI of 0.05. Samples were harvested at various time points postinfection, subjected to three freeze-thaw cycles, mechanically disrupted with 1 mm silica beads, and titered by plaque assay [36].
In vivo infection.
Mice were intraperitoneally injected with 106 PFU of virus in 0.5 ml of D10. For acute titers, half a spleen and half a lobe of liver per mouse were mechanically disrupted with 1 mm silica beads and titered by plaque assay [81]. Data represent the mean ± SEM from at least two independent experiments, and statistical significance was determined using the Mann-Whitney test. For ex vivo reactivation, peritoneal cells pooled from 3–5 mice per experimental group were plated in serial 2-fold dilutions onto MEF monolayers, which were assessed for cytopathic effect caused by reactivated virus after 21 d. Persistent replication was analyzed by quantitating preformed infectious virus in mechanically disrupted cells, which cannot reactivate virus [83]. Data represent the mean ± SEM from at least three independent experiments, and were analyzed by nonlinear regression using GraphPad Prism (GraphPad Software, San Diego, California, United States). The frequency of cells that reactivated virus ex vivo was determined by calculating the cell density at which 63.2% of wells were positive for cytopathic effect. Statistical significance was determined using a paired t-test.
v-Bcl-2 expression and purification.
Wild-type and SGR/AAA v-Bcl-2 used in NMR studies were expressed in E. coli BL21(DE3) strain grown on M9 media. For wild-type v-Bcl-2, uniformly 15N-labeled, uniformly 15N,13C-labeled, and uniformly 15N,13C-labeled, 75% 2H samples were prepared with media containing either 15NH4Cl, 15NH4Cl plus [U-13C]glucose or 15NH4Cl, [U-13C]glucose, and 75% 2H2O, respectively. For SGR/AAA v-Bcl-2, a 15N-labeled sample was prepared with media containing 15NH4Cl as the sole nitrogen source. Soluble protein was purified by Ni2+-affinity chromatography. NMR samples contained 0.5–1.0 mM protein in either 90% H2O with 10% 2H2O or 100% 2H2O, 20 mM 2H-TRIS (pH 7.8), and 5 mM 2H-dithiothreitol.
NMR spectroscopy.
All NMR experiments were acquired at 303 K on a Bruker DRX500, DRX600 or DRX800 NMR spectrometer. Backbone 1H, 13C, and 15N resonance assignments were achieved with (15N,13C,[75%]2H) γHV68 v-Bcl-2 using a suite of deuterium-decoupled, triple-resonance experiments (HNCA, HN[CO]CA, HN[CA]CB, HN[COCA]CB, HNCO and HN[CA]C) [84,85]. Side-chain 1H and 13C NMR signals were assigned from HCCH-TOCSY experiments [86]. Stereospecific assignments of valine and leucine methyl groups were obtained from an analysis of the 13C-13C coupling patterns observed for biosynthetically directed, fractionally 13C-labeled γHV68 v-Bcl-2 [87]. NOE distance restraints were obtained from three-dimensional 15N- and 13C-edited NOESY spectra [88,89] acquired with a mixing time of 80 ms. Slowly exchanging amide protons were identified in an 15N-HSQC spectrum recorded immediately after exchanging the protein into a buffer prepared with 2H2O.
Structure calculations.
v-Bcl-2 structures were calculated using a simulated annealing protocol [90] with the program CNX (MSI, San Diego, California, United States). A square-well potential (FNOE = 50 kcal mol−1) was employed to constrain NOE-derived distances. Based on cross-peak intensities, NOE-derived distance restraints were given upper bounds of 3.5, 4.5, or 6.0 H. In the refinement stage, additional ambiguous constraints were added, with an upper bound of 6.0 H, for unassigned cross peaks that were consistent with the chemical shift table (i.e., error bars of 0.07 ppm for protons and 0.7 ppm for hetero atoms) and the structure. Torsion angle restraints, φ and ψ, were generated from analysis of N, C′, Cα, and Hα chemical shifts using the TALOS program [91]. A force constant of 200 kcal mol−1rad−2 was applied to all torsional restraints. Explicit hydrogen bonds were included in α-helices only for residues observed to have slowly exchanging amide protons. The program PROCHECK was employed to analyze the geometric quality of the calculated structures in the ensemble [92].
Peptide binding.
Binding of the Bak 16-mer (GQVGRQLAIIGDDINR), the Bax 16-mer (KKLSECLKRIGDELDS), the Bad 25-mer (NLWAAQRYGRELRRMSDEFVSFKK), and a D83A mutant Bak 16-mer (GQVGRQLAIIGADINR) to wild-type v-Bcl-2, and binding of the Bax 16-mer to SGR/AAA v-Bcl-2 was assessed by NMR titration [93]. Each peptide was titrated from a concentrated stock solution into a sample of 13C-labeled protein, and binding was monitored from changes in a 13C methyl-HSQC spectrum [93]. For wild-type v-Bcl-2, 13C-HSQC spectra were recorded on a uniformly 13C,15N-labeled protein sample (120 μM) in the presence of increasing amounts of peptide (40, 80, 160, 240, and 320 μM). For SGR/AAA v-Bcl-2, 15N-HSQC spectra were recorded on a sample of uniformly 15N-labeled protein in the presence of increasing amounts of peptide.
Supporting Information
Figure S1 v-Bcl-2 Inhibits Cell Death in Thymocytes
Representative dot plots showing CD4 and CD8 profiles of nontransgenic or v-Bcl-2 transgenic thymocytes after no treatment or treatment with 0.3 mg of dexamethasone, 250 rads of γ-irradiation, or 30 μg of anti-CD3ɛ antibody.
(174 KB PDF)
Click here for additional data file.
Figure S2 Solution Structure of γHV68 v-Bcl-2
Ribbon representations of (A) γHV68 v-Bcl-2 and (B) Bcl-xL. Helices are numbered with respect to Bcl-xL. The BH1, BH2, BH3, and BH4 regions are colored magenta, red, green and yellow, respectively. Connolly surface for (C) γHV68 v-Bcl-2 and (D) Bcl-xL. The Connolly surface was calculated using a probe radius of 1.4 Å. Residues are colored as follows: Leu, Val, Ile, Phe, Tyr, Trp, Met, and Ala are yellow; Arg, Lys, and His are blue; Asp and Glu are red; and all other residues are gray. Black arrows indicate the hydrophobic grooves present on the surfaces of Bcl-xL and γHV68 v-Bcl-2.
(5.3 MB PDF)
Click here for additional data file.
HWV was supported by R01 CA74730 and HL60090; SHS by CA43143, CA52004, CA58524, and CA87650; JL by training grant T32 AI07163; BL by R01 AI40246, and LFV by PF-4379 from the American Cancer Society. We thank Mike White and Dr. Robert Schreiber for assistance in generating transgenic mice, Darren Kreamalmeyer for breeding of knock-out mice, Chaohong Sun and Jamey Mack for assistance in preparing proteins used in NMR studies, and the laboratories of HWV and SHS for helpful comments on this work. We thank Ute Moll for helpful comments.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. All authors participated in conception and design of the experiments. JL, QH, LFvD, and LL performed the experiments. All authors participated in analysis of the data. JL, QH, BL, ETO, and HWV wrote the paper.
Abbreviations
BHBcl-2 homology
dpidays post-infection
EBVEpstein-Barr virus
γHV68murine γ-herpesvirus 68
HSQCheteronuclear single-quantum coherence
IFNγinterferon-γ
KSHVKaposi's sarcoma-associated herpesvirus
NMRnuclear magnetic resonance
NOEnuclear Overhauser effect
v-Bcl-2viral Bcl-2 homolog
==== Refs
References
Bassik MC Scorrano L Oakes SA Pozzan T Korsmeyer SJ 2004 Phosphorylation of BCL-2 regulates ER Ca2+ homeostasis and apoptosis EMBO J 23 1207 1216 15010700
Petros AM Olejniczak ET Fesik SW 2004 Structural biology of the Bcl-2 family of proteins Biochim Biophys Acta 1644 83 94 14996493
Cory S Huang DC Adams JM 2003 The Bcl-2 family: Roles in cell survival and oncogenesis Oncogene 22 8590 8607 14634621
Danial NN Korsmeyer SJ 2004 Cell death: Critical control points Cell 116 205 219 14744432
Danial NN Gramm CF Scorrano L Zhang CY Krauss S 2003 BAD and glucokinase reside in a mitochondrial complex that integrates glycolysis and apoptosis Nature 424 952 956 12931191
Hardwick JM Bellows DS 2003 Viral versus cellular BCL-2 proteins Cell Death Differ 10 S68 S76 12655348
Cuconati A White E 2002 Viral homologs of BCL-2: Role of apoptosis in the regulation of virus infection Genes Dev 16 2465 2478 12368257
Letai A Bassik MC Walensky LD Sorcinelli MD Weiler S 2002 Distinct BH3 domains either sensitize or activate mitochondrial apoptosis, serving as prototype cancer therapeutics Cancer Cell 2 183 192 12242151
Huang Q Petros AM Virgin HW Fesik SW Olejniczak ET 2002 Solution structure of a Bcl-2 homolog from Kaposi sarcoma virus Proc Natl Acad Sci U S A 99 3428 3433 11904405
Huang Q Petros AM Virgin HW Fesik SW Olejniczak ET 2003 Solution structure of the BHRF1 protein from Epstein-Barr virus, a homolog of human Bcl-2 J Mol Biol 332 1123 1130 14499614
Muchmore SW Sattler M Liang H Meadows RP Harlan JE 1996 X-ray and NMR structure of human Bcl-xL, an inhibitor of programmed cell death Nature 381 335 341 8692274
Petros AM Medek A Nettesheim DG Kim DH Yoon HS 2001 Solution structure of the antiapoptotic protein bcl-2 Proc Natl Acad Sci U S A 98 3012 3017 11248023
Sattler M Liang H Nettesheim D Meadows RP Harlan JE 1997 Structure of Bcl-xL-Bak peptide complex: Recognition between regulators of apoptosis Science 275 983 986 9020082
Parrish J Metters H Chen L Xue D 2000 Demonstration of the in vivo interaction of key cell death regulators by structure-based design of second-site suppressors Proc Natl Acad Sci U S A 97 11916 11921 11027303
del Peso L Gonzalez VM Inohara N Ellis RE Nunez G 2000 Disruption of the CED-9 center dot CED-4 complex by EGL-1 is a critical step for programmed cell death in Caenorhabditis elegans
J Biol Chem 275 27205 27211 10846174
Li W Kedersha N Chen S Gilks N Lee G 2004 FAST is a BCL-X-L-associated mitochondrial protein Biochem Biophys Res Commun 318 95 102 15110758
Rodolfo C Mormone E Matarrese P Ciccosanti F Farrace MG 2004 Tissue transglutaminase is a multifunctional BH3-only protein J Biol Chem 279 54783 54792 15485857
Kasof GM Goyal L White E 1999 Btf, a novel death-promoting transcriptional repressor that interacts with Bcl-2-related proteins Mol Cell Biol 19 4390 4404 10330179
Mihara M Erster S Zaika A Petrenko O Chittenden T 2003 p53 has a direct apoptogenic role at the mitochondria Mol Cell 11 577 590 12667443
Wang HG Rapp UR Reed JC 1996 Bcl-2 targets the protein kinase Raf-1 to mitochondria Cell 87 629 638 8929532
Chau BN Cheng EH Kerr DA Hardwick JM 2000 Aven, a novel inhibitor of caspase activation, binds Bcl-xL and Apaf-1 Mol Cell 6 31 40 10949025
Pan G O'Rourke K Dixit VM 1998 Caspase-9, Bcl-XL, and Apaf-1 form a ternary complex J Biol Chem 273 5841 5845 9488720
Shibasaki F Kondo E Akagi T McKeon F 1997 Suppression of signalling through transcription factor NF-AT by interactions between calcineurin and Bcl-2 Nature 386 728 731 9109491
Liang XH Kleeman LK Jiang HH Gordon G Goldman JE 1998 Protection against fatal Sindbis virus encephalitis by beclin, a novel Bcl-2-interacting protein J Virol 72 8586 8596 9765397
Reed JC 1998 Bcl-2 family proteins Oncogene 17 3225 3236 9916985
Hu Y Benedict MA Wu D Inohara N Nunez G 1998 Bcl-XL interacts with Apaf-1 and inhibits Apaf-1-dependent caspase-9 activation Proc Natl Acad Sci U S A 95 4386 4391 9539746
Polster BM Pevsner J Hardwick JM 2004 Viral Bcl-2 homologs and their role in virus replication and associated diseases Biochim Biophys Acta 1644 211 227 14996505
Benedict CA Norris PS Ware CF 2002 To kill or be killed: Viral evasion of apoptosis Nat Immunol 3 1013 1018 12407409
Sunil-Chandra NP Efstathiou S Nash AA 1992 Murine gammaherpesvirus 68 establishes a latent infection in mouse B lymphocytes in vivo J Gen Virol 73 3275 3279 1469366
Flano E Husain SM Sample JT Woodland DL Blackman MA 2000 Latent murine gamma-herpesvirus infection is established in activated B cells, dendritic cells, and macrophages J Immunol 165 1074 1081 10878386
Weck KE Kim SS Virgin HW Speck SH 1999 Macrophages are the major reservoir of latent murine gammaherpesvirus 68 in peritoneal cells J Virol 73 3273 3283 10074181
Sunil- Chandra NP Arno J Fazakerley J Nash AA 1994 Lymphoproliferative disease in mice infected with murine gammaherpesvirus 68 Am J Pathol 145 818 826 7943173
Grundhoff A Ganem D 2004 Inefficient establishment of KSHV latency suggests an additional role for continued lytic replication in Kaposi sarcoma pathogenesis J Clin Invest 113 124 136 14702116
Kieff E Rickinson AB 2001 Epstein-Barr virus and its replication Knipe DM Howley P Fields virology, 4th ed Hagerstown (Maryland) Lippincott Williams and Wilkins 2511 2573
Rickinson AB Kieff E 2001 Epstein-Barr virus Knipe DM Howley P Fields virology, 4th ed Hagerstown (Maryland) Lippincott Williams and Wilkins 2575 2627
Gangappa S Van Dyk LF Jewett TJ Speck SH Virgin HW 2002 Identification of the in vivo role of a viral bcl-2 J Exp Med 195 931 940 11927636
Bellows DS Chau BN Lee P Lazebnik Y Burns WH 2000 Antiapoptotic herpesvirus bcl-2 homologs escape caspase-mediated conversion to proapoptotic proteins J Virol 74 5024 5031 10799576
Roy DJ Ebrahimi BC Dutia BM Nash AA Stewart JP 2000 Murine gammaherpesvirus M11 gene product inhibits apoptosis and is expressed during virus persistence Arch Virol 145 2411 2420 11205127
Wang GH Garvey TL Cohen JI 1999 The murine gammaherpesvirus-68 M11 protein inhibits Fas- and TNF- induced apoptosis J Gen Virol 80 2737 2740 10573168
Virgin HW Presti RM Li X-Y Liu C Speck SH 1999 Three distinct regions of the murine gammaherpesvirus 68 genome are transcriptionally active in latently infected mice J Virol 73 2321 2332 9971815
de Lima BD May JS Marques S Simas JP Stevenson PG 2004 Murine gammaherpesviruis 68 bcl-2 homologue contributes to latency establishment in vivo J Gen Virol 86 31 40
Tibbetts SA Van Dyk L Speck SH Virgin HW 2002 Immune control of the number and reactivation phenotype of cells latently infected with a gamma-herpesvirus J Virol 76 7125 7132 12072512
Loh J Thomas DA Revell PA Ley TJ Virgin HW 2004 Granzymes and caspase 3 play important roles in control of gammaherpesvirus latency J Virol 78 12519 12528 15507639
Sentman CL Shutter JR Hockenbery D Kanagawa O Korsmeyer SJ 1991 Bcl-2 inhibits multiple forms of apoptosis but not negative selection in thymocytes Cell 67 879 888 1835668
Chao DT Linette GP Boise LH White LS Thompson CB 1995 Bcl-XL and Bcl-2 repress a common pathway of cell death J Exp Med 182 821 828 7650488
Cohen JJ Duke RC 1984 Glucocorticoid activation of a calcium-dependent endonuclease in thymocyte nuclei leads to cell death J Immunol 132 38 42 6317746
Wyllie AH 1980 Glucocorticoid-induced thymocyte apoptosis is associated with endogenous endonuclease activation Nature 284 555 556 6245367
Clarke AR Purdie CA Harrison DJ Morris RG Bird CC 1993 Thymocyte apoptosis induced by p53-dependent and independent pathways Nature 362 849 852 8479523
Lowe SW Schmitt EM Smith SW Osborne BA Jacks T 1993 p53 is required for radiation-induced apoptosis in mouse thymocytes Nature 362 847 849 8479522
Shi YF Bissonnette RP Parfrey N Szalay M Kubo RT 1991 In vivo administration of monoclonal antibodies to the CD3 T cell receptor complex induces cell death (apoptosis) in immature thymocytes J Immunol 146 3340 3346 1827483
Smith CA Williams GT Kingston R Jenkinson EJ Owen JJ 1989 Antibodies to CD3/T-cell receptor complex induce death by apoptosis in immature T cells in thymic cultures Nature 337 181 184 2521375
Strasser A Harris AW Cory S 1991 Bcl-2 transgene inhibits T cell death and perturbs thymic self-censorship Cell 67 889 899 1959134
Sattler M Liang H Nettesheim D Meadows RP Harlan JE 1997 Structure of Bcl-x(L)-Bak peptide complex: Recognition between regulators of apoptosis Science 275 983 986 9020082
Cheng EH Kirsch DG Clem RJ Ravi R Kastan MB 1997 Conversion of Bcl-2 to a Bax-like death effector by caspases Science 278 1966 1968 9395403
Chittenden T Flemington C Houghton AB Ebb RG Gallo GJ 1995 A conserved domain in Bak, distinct from BH1 and BH2, mediates cell death and protein binding functions EMBO J 14 5589 5596 8521816
Petros AM Nettesheim DG Wang Y Olejniczak ET Meadows RP 2000 Rationale for Bcl-xL/Bad peptide complex formation from structure, mutagenesis, and biophysical studies Protein Sci 9 2528 2534 11206074
Cheng EH Wei MC Weiler S Flavell RA Mak TW 2001 BCL-2, BCLXL sequester BH3 domain-only molecules preventing BAX- and BAK-mediated mitochondrial apoptosis Mol Cell 8 705 711 11583631
Wei MC Zong WX Cheng EH Lindsten T Panoutsakopoulou V 2001 Proapoptotic BAX and BAK: A requisite gateway to mitochondrial dysfunction and death Science 292 727 730 11326099
Zong WX Lindsten T Ross AJ MacGregor GR Thompson CB 2001 BH3-only proteins that bind pro-survival Bcl-2 family members fail to induce apoptosis in the absence of Bax and Bak Genes Dev 15 1481 1486 11410528
Jurgensmeier JM Krajewski S Armstrong RC Wilson GM Oltersdorf T 1997 Bax- and Bak-induced cell death in the fission yeast Schizosaccharomyces pombe
Mol Biol Cell 8 325 339 9190211
Zha HB Fisk HA Yaffe MP Mahajan N Herman B 1996 Structure-function comparisons of the proapoptotic protein Bax in yeast and mammalian cells Mol Cell Biol 16 6494 6508 8887678
Minn AJ Kettlun CS Liang H Kelekar A Vander Heiden MG 1999 Bcl-x(L) regulates apoptosis by heterodimerization-dependent and -independent mechanisms EMBO J 18 632 643 9927423
Hanada M Aime-Sempe C Sato T Reed JC 1995 Structure-function analysis of Bcl-2 protein. Identification of conserved domains important for homodimerization with Bcl-2 and heterodimerization with Bax J Biol Chem 270 11962 11969 7744846
Yin XM Oltvai ZN Korsmeyer SJ 1994 BH1 and BH2 domains of Bcl-2 are required for inhibition of apoptosis and heterodimerization with Bax Nature 369 321 323 8183370
Cheng EH Levine B Boise LH Thompson CB Hardwick JM 1996 Bax-independent inhibition of apoptosis by Bcl-XL Nature 379 554 556 8596636
Bodaghi B Jones TR Zipeto D Vita C Sun L 1998 Chemokine sequestration by viral chemoreceptors as a novel viral escape strategy: withdrawal of chemokines from the environment of cytomegalovirus-infected cells J Exp Med 188 855 866 9730887
Patterson CE Shenk T 1999 Human cytomegalovirus UL36 protein is dispensable for viral replication in cultured cells J Virol 73 7126 7131 10438798
Penfold ME Dairaghi DJ Duke GM Saederup N 1999 Cytomegalovirus encodes a potent alpha chemokine Proc Natl Acad Sci U S A 96 9839 9844 10449781
Sarawar SR Cardin RD Brooks JW Mehrpooya M Hamilton-Easton AM 1997 Gamma interferon is not essential for recovery from acute infection with murine gammaherpesvirus 68 J Virol 71 3916 3921 9094668
Opferman JT Letai A Beard C Sorcinelli MD Ong CC 2003 Development and maintenance of B and T lymphocytes requires antiapoptotic MCL-1 Nature 426 671 676 14668867
Day CL Chen L Richardson SJ Harrison PJ Huang DCS 2004 Solution structure of pro-survival Mcl-1 and characterization of its binding by pro-apoptotic BH3-only ligands J Biol Chem 280 4738 4744 15550399
Han J Modha D White E 1998 Interaction of E1B 19K with Bax is required to block Bax-induced loss of mitochondrial membrane potential and apoptosis Oncogene 17 2993 3005 9881701
Cuconati A Degenhardt K Sundararajan R Anschel A White E 2002 Bak and Bax function to limit adenovirus replication through apoptosis induction J Virol 76 4547 4558 11932420
Clem RJ Cheng EH Karp CL Kirsch DG Ueno K 1998 Modulation of cell death by Bcl-XL through caspase interaction Proc Natl Acad Sci U S A 95 554 559 9435230
Ojala PM Tiainen M Salven P Veikkola T Castanos-Velez E 1999 Kaposi's sarcoma-associated herpesvirus-encoded v-cyclin triggers apoptosis in cells with high levels of cyclin-dependent kinase 6 Cancer Res 59 4984 4989 10519412
Ojala PM Yamamoto K Castanos-Velez E Biberfeld P Korsmeyer SJ 2000 The apoptotic v-cyclin-CDK6 complex phosphorylates and inactivates Bcl-2 Nat Cell Biol 2 819 825 11056537
Van Dyk LF Hess JL Katz JD Jacoby M Speck SH 1999 The murine gammaherpesvirus 68 v-cyclin is an oncogene that promotes cell cycle progression in primary lymphocytes J Virol 73 5110 5122 10233974
Chaffin KE Beals CR Wilkie TM Forbush KA Simon MI 1990 Dissection of thymocyte signaling pathways by in vivo expression of pertussis toxin ADP-ribosyltransferase EMBO J 9 3821 3829 2123451
Garvin AM Abraham KM Forbush KA Farr AG Davison BL 1990 Disruption of thymocyte development and lymphomagenesis induced by SV40 T-antigen Int Immunol 2 173 180 1965144
Morrison TB Weis JJ Wittwer CT 1998 Quantification of low-copy transcripts by continuous SYBR Green I monitoring during amplification BioTechniques 24 954 958 960 962 9631186
Van Dyk LF Virgin HW Speck SH 2000 The murine gammaherpesvirus 68 v-cyclin is a critical regulator of reactivation from latency J Virol 74 7451 7461 10906198
Weck KE Barkon ML Yoo LI Speck SH Virgin HW 1996 Mature B cells are required for acute splenic infection, but not for establishment of latency, by murine gammaherpesvirus 68 J Virol 70 6775 6780 8794315
Weck KE Kim SS Virgin HW Speck SH 1999 B cells regulate murine gammaherpesvirus 68 latency J Virol 73 4651 4661 10233924
Yamazaki T Lee W Arrowsmith CH Muhandiram DR Kay LE 1994 A suite of triple-resonance NMR experiments for the backbone assignment of N-15, C-13, H-2 labeled proteins with high-sensitivity J Am Chem Soc 116 11655 11666
Muhandiram DR Kay LE 1994 Gradient-enhanced triple-resonance 3-dimensional NMR experiments with improved sensitivity J Magn Reson Series B 103 203 216
Clore GM Gronenborn AM 1994 Multidimensional heteronuclear nuclear-magnetic-resonance of proteins Methods Enzymol 239 349 363 7830590
Neri D Szyperski T Otting G Senn H Wuthrich K 1989 Stereospecific nuclear magnetic-resonance assignments of the methyl-groups of valine and leucine in the DNA-binding domain of the 434-repressor by biosynthetically directed fractional C-13 labeling Biochemistry 28 7510 7516 2692701
Fesik SW Zuiderweg ERP 1988 Heteronuclear 3-dimensional NMR-spectroscopy—A strategy for the simplification of homonuclear two-dimensional NMR-spectra J Magn Reson 78 588 593
Ikura M Kay LE Tschudin R Bax A 1990 3-Dimensional NOESY-HMQC spectroscopy of a C-13-labeled protein J Magn Reson 86 204 209
Brunger AT 1992 X-PLOR, version 3.1 [computer program] New Haven and London Yale University Press
Cornilescu G Delaglio F Bax A 1999 Protein backbone angle restraints from searching a database for chemical shift and sequence homology J Biomol NMR 13 289 302 10212987
Laskowski RA MacArthur MW Moss DS Thornton JM 1993 Procheck—A program to check the stereochemical quality of protein structures J Appl Cryst 26 283 291
Hajduk PJ Augeri DJ Mack J Mendoza R Yang JG 2000 NMR-based screening of proteins containing C-13-labeled methyl groups J Am Chem Soc 122 7898 7904
|
16201011
|
PMC1238745
|
CC BY
|
2021-01-05 12:11:27
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e10
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010010
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science 10.1371/journal.ppat.001001205-PLPA-E-0145plpa-01-01-09Editorial
PLoS Pathogens—A High-Impact Journal for Pathogen Research EditorialYoung John A T John A. T. Young is Editor-in-Chief of PLoS Pathogens. E-mail: [email protected]
9 2005 30 9 2005 1 1 e12Copyright: © 2005 John A. T. Young.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Citation:Young JAT (2005) PLoS Pathogens—Top-tier journal for pathogen research. PLoS Pathog 1(1): e12.
==== Body
On behalf of the editorial board and the PLoS staff, I welcome you to the inaugural issue of PLoS Pathogens. We hope that you enjoy reading the exciting articles in this first issue. These papers were selected because each represents a breakthrough in understanding the biology of pathogens and pathogen–host interactions. For example, in a highly innovative study (DOI: 10.1371/journal.ppat.0010007), Orr and colleagues have taken advantage of recombinant herpes viruses, using the mouse model system to demonstrate the importance of major histocompatibility complex (MHC) Class I–driven antigen cross presentation for the induction of virus-specific CD8+ T cells. They show that viral inhibition of MHC class I function is important for viral entry, replication, and survival in the central nervous system. Also, Wang et al. (DOI: 10.1371/journal.ppat.0010009) elegantly describe a critical role for a specific region of the Plasmodium circumsporozoite protein in sporozoite exit from oocysts, a process that might represent a new therapeutic target for treating malaria. These are precisely the type of high-impact research articles that we envisage appearing in PLoS Pathogens.
From the beginning, the editorial board committed to publishing groundbreaking findings on bacterial, fungal, parasitic, prionic, and viral pathogens. By doing so, we intend to showcase the most important new ideas and results, particularly those of broad interdisciplinary interest. Our goal, achieved in this inaugural issue, is to bring to you articles detailing significant and substantive scientific advances that merit your attention and that of the broad readership of PLoS Pathogens.
The readership of this journal is, indeed, as varied as it is vast. As an open-access journal, PLoS Pathogens makes new findings that can shape and guide research efforts and scientific progress immediately and freely available to the broadest possible global audience. The free availability of key, new information and findings encourages the active exchange of ideas within and throughout the pathogen research community worldwide. The addition of a short, nontechnical summary accompanying each article in PLoS Pathogens adds another dimension to our accessibility. These synopses represent an important mechanism for sharing scientific advances with nonspecialists, and facilitating communication between scientists and members of the general public.
Research articles are the primary focus of PLoS Pathogens because the editorial board has recognized the critical need for a high-impact journal in our field. Consequently, we have set a very high bar for manuscript acceptance. Our review process is extremely rigorous, but timely, and is receiving high marks from our authors. Indeed, many have told us that they find the detailed and constructive feedback they receive to be extremely useful.
In addition to research articles, PLoS Pathogens' readers will find Opinions and Reviews in most monthly issues. Opinions Editor Marianne Manchester outlines her vision for this section in an accompanying article in this issue; I encourage you to read and respond to her ideas and plans. Brett Finlay, PLoS Pathogens' Reviews Editor, has solicited a number of high-quality, cutting-edge Reviews, and is also capitalizing on the great expertise of the editorial board to invigorate and develop this section of the journal. But he is also keen to hear your suggestions and ideas about important, new topics that would capture your and your colleagues' interest.
PLoS Pathogens is launching at an important time, when the pathogen research community has much to learn from each other and to share with the global community. We invite you to be a part of it—as readers, learners, authors, or reviewers—your contributions will enrich us all. Send us your presubmission inquiries and manuscript submissions at http://plospathogens.org—your best ideas and your best work!
|
16201012
|
PMC1238746
|
CC BY
|
2021-01-05 12:11:21
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e12
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010012
|
oa_comm
|
==== Front
PLoS PathogPLoS PathogppatplpaplospathPLoS Pathogens1553-73661553-7374Public Library of Science San Francisco, USA 1620101310.1371/journal.ppat.001001305-PLPA-OP-0142plpa-01-01-12OpinionsWhy Provide an Opinions Section in PLoS Pathogens? OpinionManchester Marianne Manchester Marianne Opinions Editor
Marianne Manchester is at the Scripps Research Institute, La Jolla, California, United States of America. E-mail: [email protected]
9 2005 30 9 2005 1 1 e13Copyright: © 2005 Marianne Manchester.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Citation:Manchester M (2005) Why provide an opinions section in PLoS Pathogens? PLoS Pathog 1(1): e13.
==== Body
“There were never in the world two opinions alike, any more than two hairs or two grains. Their most universal quality is diversity.”—Michel de Montaigne, French essayist
The global scope of pathogen research—a dynamic field that spans basic science, public health, epidemiology, vaccines, infectious diseases, genetics, and population dynamics—ensures that a great diversity of opinions and perspectives exist within its community of scientists. The verbal expression of such points of view, shared expertise, and opinion for its own sake may be readily witnessed at any scientific or faculty meeting. But in research articles as well as in reviews, it seems scientists are increasingly discouraged from broad speculation or from making statements that border on the provocative. This trend risks dampening the rigor and enthusiasm found in the vibrant pathogen research field.
Thus, it was the opinion of the editorial board of this new open-access journal that it was important to provide an Opinions forum. We recognize that there are important questions we can raise, and points of view to put forth—especially at a time when the field is advancing in many and key directions. The Opinions section of PLoS Pathogens is intended to provide a place for pathogen researchers to express their views on topics ranging from issues in experimental science to those involving science and public-health policy, education and training, and funding. And it is also a forum in which colleagues can bring another view in response to a stated opinion or observation, thus enriching the dialogue.
Our view for the Opinions section is to include original, editorial-length pieces that convey a concise and well-stated point of view on a timely subject. We will encourage authors to rely heavily on their unique perspective as scientists in the field, and will strive to allow plenty of leeway for speculation and model building, more than for a regular research article or a review article. An important goal of Opinions is to reach the collective PLoS Pathogens community; therefore, it will be important that each piece provides specific examples supporting the points made that relate to pathogenesis in the broader sense, rather than focusing exclusively on a single pathogen.
Opinions published in PLoS Pathogens can have an immediate and powerful effect on your colleagues in the pathogen research community and in the broader community of scientists. The open-access format of PLoS Pathogens ensures that Opinions are immediately accessible to the widest possible global audience to read, digest, and respond to.
While Opinion articles are typically solicited directly from potential authors by the Opinions editor and the editorial board, suggestions for topics of interest and import are welcomed, and should be directed to the Opinions editor, Marianne Manchester ([email protected]).
About Opinion Articles
Published monthly online
Invited and submitted articles are considered
Target length of 1,000 words
Subject to limited peer review
Addressing topics of interest to the wider pathogen research community
|
16201013
|
PMC1238747
|
CC BY
|
2021-01-05 12:11:21
|
no
|
PLoS Pathog. 2005 Sep 30; 1(1):e13
|
utf-8
|
PLoS Pathog
| 2,005 |
10.1371/journal.ppat.0010013
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1620100710.1371/journal.pcbi.001004205-PLCB-RV-0136R1plcb-01-04-08ReviewBioinformatics - Computational BiologyNeurosciencePsychologySystems BiologyHomo (Human)PrimatesMammalsRattus (Rat)Mus (Mouse)The Human Connectome: A Structural Description of the Human Brain ReviewSporns Olaf *Tononi Giulio Kötter Rolf Olaf Sporns is in the Department of Psychology and Programs in Neuroscience and Cognitive Science, Indiana University, Bloomington, Indiana, United States of America. Giulio Tononi is in the Department of Psychiatry, University of Wisconsin, Madison, Wisconsin, United States of America. Rolf Kötter is at C. and O. Vogt Brain Research Institute and Institute of Anatomy II, Heinrich Heine University, Düsseldorf, Germany.
*To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 4 e42Copyright: © 2005 Sporns et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.ABSTRACT
The connection matrix of the human brain (the human “connectome”) represents an indispensable foundation for basic and applied neurobiological research. However, the network of anatomical connections linking the neuronal elements of the human brain is still largely unknown. While some databases or collations of large-scale anatomical connection patterns exist for other mammalian species, there is currently no connection matrix of the human brain, nor is there a coordinated research effort to collect, archive, and disseminate this important information. We propose a research strategy to achieve this goal, and discuss its potential impact.
Citation:Sporns O, Tononi G, Kötter R (2005) The human connectome: A structural description of the human brain. PLoS Comput Biol 1(4): e42.
==== Body
Introduction
To understand the functioning of a network, one must know its elements and their interconnections. The purpose of this article is to discuss research strategies aimed at a comprehensive structural description of the network of elements and connections forming the human brain. We propose to call this dataset the human “connectome,” and we argue that it is fundamentally important in cognitive neuroscience and neuropsychology. The connectome will significantly increase our understanding of how functional brain states emerge from their underlying structural substrate, and will provide new mechanistic insights into how brain function is affected if this structural substrate is disrupted. It will provide a unified, time-invariant, and readily available neuroinformatics resource that could be used in virtually all areas of experimental and theoretical neuroscience.
Recent research in neuroscience has resulted in a rapid proliferation of neuroscience datasets and in the arrival of a new discipline, neuroinformatics [1–4]. Despite considerable advances in experimental techniques and computational paradigms, we still have an incomplete understanding of how human cognitive function emerges from neuronal structure and dynamics. Here, we focus on the relative lack of information about the different types of neural elements and their neural connections in the human brain. While a larger number of anatomical studies of the human brain have been carried out at the macroscopic (cerebral lobes, surface landmarks, and white matter tracts) or microscopic (cytoarchitectonics, myeloarchitectonics, chemoarchitectonics, etc.) anatomical level, there is virtually no information on the finer connectivity patterns, including neuronal connection densities or laminar projection patterns in relation to anatomically segregated cortical areas or intraregional differentiation. Furthermore, none of the available information is deposited in a single standardized data format, nor can it be accessed through a public database.
Experimental approaches to human cognition have been significantly enhanced by the arrival of functional neuroimaging [5], a set of techniques that can be applied to study a broad range of cognitive functions, with ever-increasing spatial and temporal resolution. But the mechanistic interpretation of neuroimaging data is limited, in part due to a severe lack of information on the structure and dynamics of the networks that generate the observed activation patterns. A potential theoretical framework for conceptualizing cognition as a network phenomenon is based on two main organizational principles found in the cerebral cortex, functional segregation, and functional integration [6,7]. Emerging network theories of cognition emphasize the contextual [8], distributed [9], dynamic [10], and degenerate [11,12] nature of structure–function mappings in the brain. To successfully map structure to function in the human brain, we urgently need a comprehensive, detailed structural model of neuronal units and their connections. Connectional models of the human brain are scarce and poorly defined [13], and they are largely based on extrapolating anatomical information from other primate species such as the macaque monkey, an approach that is problematic [14], in part, because of our incomplete understanding of evolutionary changes.
Our proposal to assemble the human connectome has several components. First, we attempt to define a level of scale at which a first draft of the human connectome might be assembled. We consider several potential experimental and neuroinformatics approaches for creating this first draft. We then discuss potential problems and limitations of the connectome, including the central issues of individual variability and development. Finally, we sketch out the potential impact of the connectome in computational and cognitive neuroscience.
Scales and Levels of Structural Description
The human genome is composed of approximately 3 × 109 base pairs, containing around 20,000–30,000 genes [15]. The compilation of the genome was aided by the fact that base pairs and genes are relatively straightforward choices as basic structural elements. The connectome will consist of two main descriptors defining its network architecture: neural elements and neural connections. Data fields for these elements specify superordinate or subordinate structures, a normalized position within a standard coordinate system, and additional parameters such as physiological/biophysical metadata that are crucial for specifying neural dynamics. The set of all N elements constitutes the columns (targets) and rows (sources) of an N2–N connection matrix A, whose aij entries represent connections between individual elements i and j. In keeping with conventions adopted by other authors [16,17], confirmed absence of a connection is denoted by aij = 0, while confirmed presence of a connection (irrespective of its strength or physiological characteristics) results in aij = 1. Once a connection is confirmed to be present, its nonzero matrix element receives additional data entries cataloguing a range of structural and physiological parameters, such as fiber direction, connection density, strength, sign (excitatory/inhibitory), conduction delay, potential modulatory effects (voltage dependence), etc. The union of the binary connection matrix and connection-specific physiological data then results in a structural description that combines connection topology and biophysics.
Compiling the connectome faces two significant challenges not shared by other natural or technological networks. First, the human brain is a highly complex organ with a great number of structurally distinct, heterogeneous, yet interconnected components. Because a primary application of the connectome will likely be a structural substrate for understanding human cognitive function and interpreting neuroimaging studies, a first draft of the connectome may focus on the structural description of the corticothalamic system, including all regions of the cortex and their associated thalamic nuclei. Extensions of this first draft might include additional connected regions (striatum, cerebellum, etc.), with the ultimate aim of compiling the connectome of the whole brain.
A second challenge is that basic structural elements of the human brain, in terms of network nodes and connections, are difficult to define. Different kinds of structural descriptions could target at least three rather distinct levels of organization. At opposite ends of the scale are the level of single neurons and synapses (microscale) and the level of anatomically distinct brain regions and inter-regional pathways (macroscale). Between these two levels is the level of neuronal groups or populations (mesoscale). It is important to determine which level of description is the most appropriate for a first draft of the human connectome.
Microscale: Single Neurons and Synapses
Attempting to assemble the human connectome at the level of single neurons is unrealistic and will remain infeasible at least in the near future. With single neurons as the basic element, the size of the connectome would be several orders of magnitude larger than that of the genome, comprising an estimated 1011 neurons, with 1015 connections between them (approximately 1010 neurons and 1013 connections in the cortex alone) [18,19]. Recording or tracing 1015 connections is not only technically impossible, it may also be unnecessary. While a genomic mutation in a single base pair can have dramatic consequences, alterations of single synapses or cells have not been shown to have similar macroscopic effects. Instead, there is overwhelming evidence that human cognitive functions depend on the activity and coactivity of large populations of neurons in distributed networks, including the corticothalamic system [20]. Furthermore, individual neurons and connections are subject to rapid plastic changes. These changes include synaptic weights as well as structural remodeling of dendritic spines and presynaptic boutons [21], possibly switching synaptic connections between large numbers of potential synaptic sites [22,23]. We suggest that the vast number, high variability, and fast dynamics of individual neurons and synapses render them inappropriate as basic elements for an initial draft of the connectome.
Macroscale: Brain Regions and Pathways
An advantage of single neurons is that the elements themselves are relatively easily demarcated and well defined. In contrast, brain areas and neuronal populations are more difficult to delineate. No single universally accepted parcellation scheme currently exists for human brain regions (e.g., areas of the cerebral cortex), posing a significant obstacle to creating a unified resource such as the connectome. In the human cerebral cortex, neurons are arranged in an unknown number of anatomically distinct regions and areas, perhaps on the order of 100 [24] or more. Different subdivisions of the human brain (e.g., brain stem, thalamus, cerebellum, or cortex) may require different criteria for parcellation.
Nonetheless, anatomically distinct brain regions and inter-regional pathways represent perhaps the most feasible organizational level for compiling a first draft of the human connectome. Several neuroinformatics resources recording large-scale connection patterns in the cerebral cortex of various mammalian species already exist, for example, for most cortical regions of the macaque monkey [16,17,25], cat [26], and rat [27]. Computational analyses of these datasets have revealed a broad range of network characteristics [28], including the existence of clusters of brain regions [29], hierarchical organization [30,31], small-world attributes [32,33], distinct functional streams [34], motifs [35], and areal contributions to global network measures [36].
A broad range of experimental approaches exist at the macroscale. Cerebral white matter has traditionally been taken as a marker of the amount of connectivity within a cortical system. The relative contribution of cerebral white matter has increased throughout phylogeny to such an extent that its volume and metabolic requirements may present a limitation to further increases in connectional complexity [37]. The structural organization of white matter has been investigated by dissection, histological staining, degeneration methods, and axonal tracing [38]. Axonal tracing methods are the main basis for existing mammalian large-scale connection matrices [16,25–27], and their systematic compilation is currently being refined [17,39] to extract more sophisticated data, perhaps with the help of automated text analysis. In the human brain, postmortem tracing with carbocyanine dyes has provided details of connectivity within and between adjacent areas [40,41]. However, because of the slow transport and the length of fibers in the human brain, this method fails to reveal more remote connections. Another tracing method employs the in vivo detection by magnetic resonance imaging of high-contrast rare earth ions (e.g., manganese [42]) that have been injected into fiber tracts or inhaled, and taken up by neurons. The invasive nature and potential toxicity of the procedure makes it an unlikely candidate for human connectivity analyses.
Recent noninvasive imaging methods (diffusion tensor imaging [DTI] in its several variants, commonly followed by computational tractography) have been shown to produce results that are consistent with known pathways formed by major fiber tracts in the human brain [43–45], although there continues to be some limitations in data acquisition and processing algorithms [46]. To disambiguate signals produced by crossing or intersecting fibers, advances in diffusion imaging technology may allow the resolution of multiple axonal pathways within a single image voxel [47,48]. Despite the promise of diffusion imaging, a systematic atlas of DTI-based neuroanatomy, including probabilistic data gathered from individual brains, has not yet been produced, and the relationship among tensor fields, fiber tracts, and neuronal connections remains controversial. This controversy is likely to be resolved only through a combination of anatomical tracing techniques with noninvasive diffusion and functional imaging.
Perhaps the most promising avenue for compiling the human connectome originates from the notion that individual brain regions maintain individual connection profiles. What defines a segregated brain region is that all its structural elements share highly similar long-range connectivity patterns, and that these patterns are dissimilar between brain regions. These connectivity patterns determine the region's functional properties [49], and also allow their anatomical delineation and mapping. Diffusion imaging can be used to identify borders between cortical areas [50], most clearly on the basis of differences in long-range connections to the thalamus [51,52]. The idea that patterns of connectivity can be used to identify areal boundaries has also been tested in a combination of functional imaging and DTI/tractography in the human medial frontal cortex [53]. First, connection probabilities from voxels within the medial frontal cortex to all other voxels in the rest of the brain were obtained. Binarized connection patterns were then used to calculate a cross-correlation matrix, which was examined for the existence of distinct clusters of voxels with shared connection patterns. Such clusters were taken to represent anatomically segregated areas corresponding to human supplementary motor area and pre- supplementary motor area, respectively. Functional mapping revealed close correspondence between DTI and functional activation patterns. While this combined structural–functional approach is computationally intensive, nothing prevents its application to the entire corticothalamic system. We suggest that the correlated use of noninvasive structural and functional imaging methods offers the most promising experimental route toward the human connectome.
Structural connection data obtained using this combined methodology can in principle be validated by histological or tracing methods. Likely, no single method will turn out to be sufficiently powerful or comprehensive. A systematic application of sophisticated diffusion-weighted imaging combined with spatially registered high-resolution anatomical or spectroscopic imaging, regional activation, and coactivation data (e.g., electroencephalography, magnetoencephalography) obtained within the same individual subject, may offer the most feasible strategy for mapping the human connectome at the macroscale. This first draft of the human connectome would take the form of a probabilistic map of voxel-by-voxel connectivity embedded within standard coordinates containing approximately 104–105 elements and approximately 105–107 structural connections. It would not only provide information on the large-scale connection patterns within the corticothalamic system, but also on parcellation of human cortex into distinct areas based on a combination of structural and functional data in the same individual. Since this connectivity matrix is voxel-based, it can be cross-referenced with existing reference templates and with population-based brain atlases.
Mesoscale: Minicolumns and Their Connection Patterns
A first draft of the corticothalamic connectome at the macroscale might provide a comprehensive dataset comprising several hundred brain regions and thousands of pathways, but it does not incorporate information on functional subdivisions or segregated subcircuits within each brain region. While such a macroscale description will provide many fundamental insights into the large-scale organization of human cortex, it is an insufficient basis for a complete understanding of the human brain's functional dynamics and information processing capacities. A further step in this direction involves the characterization of connection patterns among elementary processing units, corresponding to local populations of neurons such as cortical minicolumns. Mountcastle [54,55] originally proposed the cortical minicolumn as a basic functional unit of mammalian cerebral cortex. While details of minicolumn architecture are likely to vary across different brain regions [56–58], minicolumns generally contain approximately 80–100 neurons, spanning all cortical layers, with a diameter of approximately 30–50 μm [55]. Minicolumns may possess relatively stereotypic internal processing, and maintain generic patterns of inputs and outputs with minicolumns in other regions [56–59].
Recent studies have provided evidence for functionally specialized and precisely wired subnetworks of neurons coexisting within single cortical columns [60,61]. These studies have shown that cortical columns may contain segregated subnetworks, corresponding to minicolumns, which promote greater intracolumnar functional independence and informational heterogeneity. The members of such subnetworks are selectively interconnected with each other, indicating that connections within and between minicolumns follow more complex rules than simple random patterns or Gaussian (distance-dependent) connection profiles.
Minicolumns may be a sensible choice for neural elements at the mesoscale of the connectome because they may represent basic functional elements that are crucial for cortical information processing. While tracing or recording all minicolumns in an individual brain is still impossible, it may be feasible to collect data on minicolumn anatomy for selected brain regions, which can then be “fit” into the appropriate positions within the macroscale connection matrix. Functional imaging at columnar resolution has been carried out in animal experiments using high field strength [62], and may be facilitated in the future through selective imaging of fast and spatially precise capillary cerebral blood flow response components [63]. Especially important for determining functional responses of brain regions are connection patterns between each region's constituent elements. Axonal tracing methods have revealed specific patterns of horizontal connections between individual cells and cell populations within a brain region, which are often found to preferentially link cells with similar response characteristics [64,65], resulting in clustered intra-areal connectivity. Such patterns might also be accessible in the human brain [40], and regional variations in such patterns may provide important clues regarding the way in which information is processed within each region.
The axonal tracing approach delivers minicolumn maps for each distinct brain region, including information on their functional segregation and local (intra-areal) interconnectivity. While parcellated brain regions at the macroscale can be identified across individuals, we have no means to resolve the locations of corresponding minicolumns across different brains. The mapping of smaller-than-macroscopic units, therefore, requires coordinate-independent mapping approaches [66], which preserve the anatomical relationships and basic physiological properties of these units. Thus, macroscale and minicolumn descriptions deliver complementary datasets that need cross-level integration to achieve a single mesoscale version of the connectome. Effectively, minicolumn maps need to be mapped onto brain region–specific voxel sets rather than individual voxels, with voxel sets providing spatial embedding and probabilistic long-range connections, and minicolumn maps providing local connectivity, processing, and coding information. The minicolumn description provides a functionally heterogeneous architecture that is unique to each parcellated brain region, with specific (probabilistic) patterns of intra-areal and interareal minicolumn connections. A crucial task will be to convert long-range, nondirected voxel-by-voxel connectivity into directed, functionally heterogeneous long-range minicolumn interconnectivity. This requires intermediate descriptive levels such as stripes, bands, and blobs in the early visual system, for which specific connection patterns have been demonstrated across areal boundaries. When accomplished, this cross-level integration will result in a mesoscale connection matrix of the human brain that might comprise as many as 107–108 structural elements (comparable in size to the 2005 World Wide Web), with minicolumn elements directly cross-referenced to voxels in the macroscale map.
Individual Variability and Development
The large-scale connectivity structure of the brain above the synaptic level represents a relatively invariant characteristic of our species. Once the elements and connections in the human brain are recorded, this dataset will remain stable, essentially forever. However, as in the case of the genome, the precise combinations and patterns of elements and their connections exhibit significant variations between individuals, at all levels of scale. Some individual variations may be due to genetic differences, others may be the result of developmental and experiential history, gender differences, pathologies, or responses to injury. To complicate matters further, the human connectome undergoes development through time, from early stages of the embryo to adolescence to adult age. Incorporating individual variations and developmental stages is absolutely crucial in rendering the connectome an effective resource.
Anatomical and imaging studies have revealed significant interindividual variability in the size and location of brain areas, as well as in the relationship between macrostructures (e.g., the cortical gyrification pattern) and microstructures (e.g., cytoarchitectonics and cytochemistry). Statistical analysis of variations in macroscopic surface features of the human brain, for example its sulcal geometry [67], demonstrates the extent to which even large-scale features of cortical morphology vary between individuals, possibly as a result of genetic differences [68]. An anatomical study of Broca's area in ten postmortem human brains revealed significant variations in size as well as in the area's relation to sulcal landmarks [69]. Applying structural magnetic resonance imaging to map the boundaries of the planum temporale has demonstrated significant variations in its size and position across 50 individuals [70]. Functional neuroimaging studies have also revealed significant interindividual differences, for example, in the location and extent of area MT/V5 [71] and other visual cortical areas [72]. These functional differences are presumably due to variations in underlying structural (cytoarchitectonic and connectional) substrates.
The presence of significant interindividual variability in structural connection patterns, even at the macroscale level, and the fundamentally probabilistic nature of connectivity datasets provided in the connectome may be viewed as fundamental weaknesses of the proposal, undermining its comprehensive goal of a definitive structural description of the human brain. However, we should consider the fact that there is also clear interindividual variability in the human genome. Nevertheless, the first draft with a DNA sequence obtained from cells from only a few individuals [15] has proven immensely useful for gaining insights into general organizational features of the human genome. Mapping of interindividual variability in the connectome is a necessary further step, but does not detract from the potential insights gained from a first draft that does not yet systematically incorporate these differences.
Steps Toward the Human Connectome: Its Compilation, Assembly, and Integration
Based on a combination of functional and diffusion-weighted imaging, we outline the following five-step strategy for compiling a first draft of the human connectome at the macroscale.
Step 1 is to perform diffusion-weighted imaging followed by probabilistic tractography of thalamocortical tracts as well as corticocortical interareal pathways, using correlations in connectivity profiles to assist in parcellating human cortical regions. The end result is a voxel-wise probabilistic all-to-all structural connectivity matrix for the human brain. Step 2 is to perform a correlation analysis of spatially registered, equally resolved resting activity and/or multistimulus/multitask activation data (functional magnetic resonance imaging and/or magnetoencephalography) recorded in the same person [73], emphasizing strong functional relationships that are consistent across tasks [74]. The end result is a voxel-wise all-to-all functional connectivity matrix for the human brain. Step 3 is to perform a cluster analysis of correspondences between the structural and functional connectivity matrix obtained under steps 1 and 2, with the goal of identifying regions of consistent structure–function relationships in the human brain, possibly involving indirect projections [75]. Step 4 is to compare the results obtained by cluster analysis (step 3) with macaque data in order to identify correspondences (e.g., in visuomotor pathways) and deviations (e.g., in structures such as the fasciculus arcuatus). Step 5 is to validate the strongest predictions generated by assembling the final combined structural–functional connectivity matrix using custom-designed stimuli and perturbational techniques such as transcranial magnetic stimulation.
The following three steps represent additional stages designed to further test and verify the human connectome, including population analyses of individual variability and pathology. Step 6 is to perform a population analysis of healthy subjects and spatial registration to standard brain coordinates for probabilistic statements about data from steps 1–5. Step 7 is to compare population data on clustered brain regions to histologically identified regions in a probabilistic human brain atlas to assess correspondence. Step 8 is to compare population data between healthy individuals and patient groups with specific pathologies, to identify connectional differences.
Ultimately, the connectome will likely describe connectivity patterns at multiple levels of scale, for example, by incorporating linkages between the macroscale of brain regions and pathways in more elementary mesoscale functional units such as minicolumns and their patterns of interconnectivity. As experimental techniques mature, the connectome will gradually evolve through different stages of assembly as it is refined, updated, cross-validated, and “filled in.” Standardization of parcellation schemes, elimination of unreliable data, and incorporation of additional structural levels above and below the one chosen for the initial draft will drive this effort. An additional driving force is the continued innovation in data acquisition and analysis techniques, particularly in diffusion-weighted imaging and tractography, which will result in progressive revision, refinement, and extension of the connectome. To become a useful research tool, the connectome must be linked to other databases (compiled in parallel efforts) that record additional information mapped across the human brain, such as receptor distributions or gene expression patterns. We expect that assembling even the first draft of the connectome will require significant experimental and computational resources over an extended period of time. Compilation, assembly, and integration efforts are likely to be extensive tasks, requiring large-scale collaboration, coordinated data collection and dissemination, and the establishment of new computational methods, data standards, and mechanisms for controlled validation and quality assurance.
Conclusions: The Potential Impact of the Connectome
How can the connectome be used to map brain structure to function? The step from structure to function is essential for understanding how cognitive processes emerge from their morphological substrates. Our central motivating hypothesis is that the pattern of elements and connections as captured in the connectome places specific constraints on brain dynamics, and thus shapes the operations and processes of human cognition. In turn, data recording the activity of the human brain in combination with the structural model provided by the connectome will help to discern causal interactions in large-scale brain networks (e.g., [76–78]). We emphasize that structure–function relationships are not directly evident from the connectional dataset itself. Rather, their elucidation will require further intense empirical and computational study. Depending on sensory input, global brain state, or learning, the same structural network can support a wide range of dynamic and cognitive states. This should not, however, discourage the effort to assemble the connectome. Similar difficulties in mapping structure to function exist in the case of the genome, although they generally were not foreseen when the project was initiated. Highly complex transcriptional regulatory networks, signaling pathways, mechanical forces, and elaborate mechanisms of gene expression all play essential roles in translating base sequences into functioning cells, tissues, and organisms. Both genome and connectome constitute complex networks [79], whose functions are not fully understood even if their structural substrates are fully catalogued.
An obvious and related use of the human connectome would be providing structural information that can be implemented as part of large-scale computational models [80,81]. If the connectome is sufficiently comprehensive and accessible, it could also provide a set of structural benchmarks that might facilitate the comparison and integration of the numerous specialized and structurally based models that have already been proposed in computational neuroscience. Drawing upon human connectional datasets would help ground modeling efforts aimed specifically at brain mechanisms of human cognitive function (e.g., language). Other computational applications of the connectome are topological analyses of network structure [28], perturbational studies aiming at mapping structural disruption to functional defects, and synthetic brain imaging [82–84].
The human connectome could potentially have a major impact on our understanding of brain damage and subsequent recovery. The effects of developmental variations or abnormalities, traumatic brain injury, or neurodegenerative disease can all be captured as specific structural variants of the human connectome. The functional consequences of network perturbations will allow a better understanding of structural causes of dysfunction, and may permit the design of strategies for recovery based on network analysis. Understanding the basic network causes of brain diseases may also open new avenues for therapy and prevention by harnessing inherent network mechanisms that ensure robustness and compensation.
There are many structural and functional aspects that the human connectome, as envisioned in this article, does not contain or address. For example, it does not explicitly capture or catalogue the rich variety of neuronal morphologies, the diversity of physiological and biochemical neural subtypes, glial cells, or brain vascularization. Its first draft does not capture synaptic plasticity and remodeling, nonsynaptic communication, hormonal regulation, or degenerative processes. We do not view these issues as shortcomings that fundamentally undermine the usefulness or potential impact of the connectome. Rather, they illustrate the open-ended, integrative nature of the proposed research effort. The connectome will represent a foundational resource and a central reference point for a broad range of specialized databases [85], thus making federating these databases more effective. Most importantly, the connectome will provide an important tool for mechanistic modeling and interpretation of human functional brain data.
OS was supported by National Institutes of Health National Institute on Drug Abuse grant 1R21DA15647–01. RK was supported by the Deutsche Forschungsgemeinschaft (Graduate School 320 and KO 1560/6–2). Core ideas were developed at a workshop sponsored by the James S. McDonnell Foundation. The authors thank M. Arbib, K. Friston, R. McIntosh, B. Horwitz, and T. Behrens for comments and suggestions.
Abbreviation
DTIdiffusion tensor imaging
==== Refs
References
Kötter R 2001 Neuroscience databases: Tools for exploring brain structure function relationships Phil Trans R Soc Lond B Biol Sci 356 1111 1120 11545693
Amari S Arbib MA Kötter R 2003 Special issue: Neuroinformatics Neural Netw 16 1235 1398
Koslow SH Subramanian S 2005 Databasing the brain: From data to knowledge Hoboken (New Jersey) Wiley 466 p.
Insel TR Volkow ND Li TK Battey JF Landis SC 2003 Neuroscience networks: Data sharing in an information age PLoS Biol 1 e17. DOI: 10.1371/journal.pbio.0000017 14551914
Frackowiak RSJ Friston KJ Frith C Dolan R Price CJ 2003 Human brain function, 2nd ed New York Academic Press 1,144 p.
Tononi G Edelman GM Sporns O 1998 Complexity and the integration of information in the brain Trends Cogn Sci 2 44 52
Friston KJ 2002 Beyond phrenology: What can neuroimaging tell us about distributed circuitry? Annu Rev Neurosci 25 221 250 12052909
McIntosh AR 2000 Towards a network theory of cognition Neural Netw 13 861 870 11156197
Bressler SL 1995 Large-scale cortical networks and cognition Brain Res Brain Res Rev 20 288 304 7550362
Bressler SL Kelso JAS 2001 Cortical coordination dynamics and cognition Trends Cogn Sci 5 26 36 11164733
Tononi G Sporns O Edelman GM 1999 Measures of degeneracy and redundancy in biological networks Proc Natl Acad Sci U S A 96 3257 3262 10077671
Price CJ Friston KJ 2002 Degeneracy and cognitive anatomy Trends Cogn Sci 6 416 421 12413574
Crick F Jones E 1993 The backwardness of human neuroanatomy Nature 361 109 110 8421513
Tootell RBH Tsao D Vanduffel W 2003 Neuroimaging weighs in: Humans meet macaques in “primate” visual cortex J Neurosci 23 3981 3989 12764082
Venter JC Adams MD Myers EW Li PW Mural RJ 2001 The sequence of the human genome Science 1304 1351
Young MP 1993 The organization of neural systems in the primate cerebral cortex Proc R Soc Lond B Biol Sci 252 13 18
Stephan KE Kamper L Bozkurt A Burns GA Young MP 2001 Advanced database methodology for the collation of the connectivity data on the macaque brain (CoCoMac) Philos Trans R Soc Lond B Biol Sci 356 1159 1186 11545697
Murre JMJ Sturdy DPF 1995 The connectivity of the brain: Multi-level quantitative analysis Biol Cybern 73 529 545 8527499
Braitenberg V Schüz A 1991 Anatomy of the cortex: Statistics and geometry Berlin Springer 249 p.
Mountcastle VB 1998 Perceptual neuroscience: The cerebral cortex Boston Harvard University Press 512 p.
Engert F Bonhoeffer T 1999 Dendritic spine changes associated with hippocampal long-term synaptic plasticity Nature 399 66 70 10331391
Stepanyants A Hof PR Chklovskii DB 2002 Geometry and structural plasticity of synaptic connectivity Neuron 34 275 288 11970869
Kalisman N Silberberg G Markram H 2005 The neocortical microcircuit as a tabula rasa Proc Natl Acad Sci U S A 102 880 885 15630093
Van Essen DC Drury HA Joshi S Miller MI 1998 Functional and structural mapping of human cerebral cortex: Solutions are in the surfaces Proc Natl Acad U S A 95 788 795
Felleman DJ Van Essen DC 1991 Distributed hierarchical processing in the primate cerebral cortex Cereb Cortex 1 1 47 1822724
Scannell JW Burns GAPC Hilgetag CC O'Neil MA Young MP 1999 The connectional organization of the cortico-thalamic system of the cat Cereb Cortex 9 277 299 10355908
Burns GAPC Young MP 2000 Analysis of the connectional organisation of neural systems associated with the hippocampus in rats Phil Trans R Soc Lond B Biol Sci 355 55 70 10703044
Sporns O Chialvo D Kaiser M Hilgetag CC 2004 Organization, development and function of complex brain networks Trends Cogn Sci 8 418 425 15350243
Hilgetag CC Burns GAPC O'Neill MA Scannell JW Young MP 2000 Anatomical connectivity defines the organization of clusters of cortical areas in the macaque monkey and the cat Phil Trans R Soc Lond B Biol Sci 355 91 110 10703046
Hilgetag CC O'Neill MA Young MP 2000 Hierarchical organisation of macaque and cat cortical sensory systems explored with a novel network processor Phil Trans R Soc Lond B Biol Sci 355 71 90 10703045
Hilgetag CC O'Neill MA Young MP 1996 Indeterminate organization of the visual system Science 271 776 777 8628990
Sporns O Tononi G Edelman GM 2000 Theoretical neuroanatomy: Relating anatomical and functional connectivity in graphs and cortical connection matrices Cereb Cortex 10 127 141 10667981
Sporns O Zwi J 2004 The small world of the cerebral cortex Neuroinformatics 2 145 162 15319512
Young MP 1992 Objective analysis of the topological organization of the primate cortical visual system Nature 358 152 155 1614547
Sporns O Kötter R 2004 Motifs in brain networks PLoS Biol 2 e369. DOI: 10.1371/journal.pbio.0020369 15510229
Kötter R Stephan KE 2003 Network participation indices: Characterizing component roles for information processing in neural networks Neural Netw 16 1261 1275 14622883
Chklovskii DB Schikorski T Stevens CF 2002 Wiring optimization in cortical circuits Neuron 34 341 347 11988166
Köbbert C Apps R Bechmann I Lanciego JL Mey J 2000 Current concepts in neuroanatomical tracing Prog Neurobiol 62 327 351 10856608
Kötter R 2004 Online retrieval, processing, and visualization of primate connectivity data from the CoCoMac database Neuroinformatics 2 127 144 15319511
Galuske RA Schlote W Bratzke H Singer W 2000 Interhemispheric asymmetries of the modular structure in human temporal cortex Science 289 1946 1949 10988077
Tardif E Clarke S 2001 Intrinsic connectivity of human auditory areas: A tracing study with DiI Eur J Neurosci 13 1045 1050 11264678
Saleem KS Pauls JM Augath M Trinath T Prause BA 2002 Magnetic resonance imaging of neuronal connections in the macaque monkey Neuron 34 685 700 12062017
Conturo TE Lori NF Cull TS Akbudak E Snyder AZ 1999 Tracking neuronal fiber pathways in the living human brain Proc Natl Acad Sci U S A 96 10422 10427 10468624
LeBihan D 2003 Looking into the functional architecture of the brain with diffusion MRI Nature Rev Neurosci 4 469 480 12778119
Catani M Jones DK Donato R Ffytche DH 2003 Occipito-temporal connections in the human brain Brain 126 2093 2107 12821517
Dougherty RF Ben-Shachar M Bammer R Brewer AA Wandell BA 2005 Functional organization of human occipital-callosal fiber tracts Proc Natl Acad Sci USA 102 7350 7355 15883384
Tuch DS Reese TG Wiegell MR Wedeen VJ 2003 Diffusion MRI of complex neural architecture Neuron 40 885 895 14659088
Tuch DS Wisco JJ Khatchaturian MH Ekstrom LB Kötter R 2005 Q-Ball imaging of macaque white matter architecture Phil Trans R Soc B Biol Sci 360 869 879
Passingham RE Stephan KE Kötter R 2002 The anatomical basis of functional localization in the cortex Nat Rev Neurosci 3 606 616 12154362
Behrens TEJ Johansen-Berg H 2005 Relating connectional architecture to grey matter function using diffusion imaging Phil Trans R Soc B Biol Sci 360 903 911
Jones EG Adelman G Smith B 1998 Brodmann's areas and cerebral cortex Encyclopedia of neuroscience New York Elsevier pp. 294–296, 339–341.
Behrens TEJ Johansen-Berg H Woolrich MW Smith SM Wheeler-Kingshott CAM 2003 Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging Nat Neurosci 6 750 757 12808459
Johansen-Berg H Behrens TEJ Robson MD Drobnjak I Rushworth MFS 2004 Changes in connectivity profiles define functionally distinct regions in human medial frontal cortex Proc Natl Acad Sci U S A 101 13335 13340 15340158
Mountcastle VB 1978 An organizing principle for cerebral function Edelman GM Mountcastle VB The mindful brain Cambridge MIT Press 7 50
Mountcastle VB 1997 The columnar organization of the neocortex Brain 120 701 722 9153131
Buxhoeveden DP Casanova MF 2002 The minicolumn hypothesis in neuroscience Brain 125 935 951 11960884
Rockland KS Ichinohe N 2004 Some thoughts on cortical minicolumns Exp Brain Res 158 265 277 15365664
Jones EG 2000 Microcolumns in the cerebral cortex Proc Natl Acad Sci U S A 97 5019 5021 10805761
Douglas RJ Martin KA 1991 A functional microcircuit for cat visual cortex J Physiol 440 735 769 1666655
Yoshimura Y Dantzker JLM Callaway EM 2005 Excitatory cortical neurons form fine-scale functional networks Nature 433 868 873 15729343
Ohki K Chung S Ch'ng YH Kara P Reid C 2005 Functional imaging with cellular resolution reveals precise microarchitecture in visual cortex Nature 433 597 603 15660108
Zhao F Wang P Hendrich K Kim SG 2005 Spatial specificity of cerebral blood volume-weighted fMRI responses at columnar resolution Neuroimage 27 416 424 15923128
Vanzetta I Slovin H Omer DB Grinvald A 2004 Columnar resolution of blood volume and oximetry functional maps in the behaving monkey: Implications for fMRI Neuron 42 843 854 15182722
Callaway EM Katz LC 1990 Emergence and refinement of clustered horizontal connections in cat striate cortex J Neurosci 10 1134 1153 2329372
Amirikian B Georgopoulos AP 2003 Modular organization of directionally tuned cells in the motor cortex: Is there a short-range order? Proc Natl Acad Sci U S A 100 12474 12479 14523235
Kötter R Wanke E 2005 Mapping brains without coordinates Phil Trans R Soc Lond B Biol Sci 360 751 766 15971361
Thompson PM Schwartz C Lin RT Khan AA Toga AW 1996 Three-dimensional statistical analysis of sulcal variability in the human brain J Neurosci 16 4261 4274 8753887
Thompson P Cannon TD Narr KL van Erp T Poutanen VP 2001 Genetic influences on brain structure Nat Neurosci 4 1 6 11135631
Amunts K Schleicher A Bürgel U Mohlberg H Uylings HBM 1999 Broca's region revisited: Cytoarchitecture and intersubject variability J Comp Neurol 412 319 341 10441759
Westbury CF Zatorre RJ Evans AC 1999 Quantifying variability in the planum temporale: A probability map Cereb Cortex 9 392 405 10426418
Watson JD Myers R Frackowiak RS Hajnal JV Woods RP 1993 Area V5 of the human brain: Evidence from a combined study using positron emission tomography and magnetic resonance imaging Cereb Cortex 3 79 94 8490322
Hasnain MK Fox PT Woldorff MG 1998 Intersubject variability of functional areas in the human visual cortex Hum Brain Mapp 6 301 315 9704267
Lahaye PJ Poline JB Flandin G Dodel S Garnero L 2003 Functional connectivity: Studying nonlinear, delayed interactions between BOLD signals Neuroimage 20 962 974 14568466
Bartels A Zeki S 2005 Brain dynamics during natural viewing conditions—A new guide for mapping connectivity in vivo Neuroimage 24 339 349 15627577
Stephan KE Hilgetag CC Burns GAPC O'Neill MA Young MP 2000 Computational analysis of functional connectivity between areas of primate cerebral cortex Phil Trans R Soc Lond B Biol Sci 355 111 126 10703047
Kötter R Sommer F 2000 Global relationship between anatomical connectivity and activity propagation in the cerebral cortex Phil Trans R Soc Lond B Biol Sci 355 127 134 10703048
McIntosh AR Gonzalez-Lima F 1994 Structural equation modeling and its application to network analysis in functional brain imaging Hum Brain Mapp 2 2 22
Friston KJ Harrison L Penny W 2003 Dynamic causal modeling Neuroimage 19 1273 1302 12948688
Bota M Dong HW Swanson LW 2003 From gene networks to brain networks Nat Neuroscience 6 795 799 12886225
Horwitz B Sporns O 1994 Neural modeling and functional neuroimaging Hum Brain Mapp 1 269 283 24591196
Horwitz B Tagamets MA McIntosh AR 1999 Neural modeling, functional brain imaging, and cognition Trends Cogn Sci 3 91 98 10322460
Arbib MA, Bischoff A, Fagg AH, Grafton S T 1994 Synthetic PET: Analyzing large-scale properties of neural networks Hum Brain Mapp 2 225 233
Horwitz B 2004 Relating fMRI and PET signals to neural activity by means of large-scale neural models Neuroinformatics 2 251 266 15319520
Chadderdon G Sporns O 2005 A large-scale neurocomputational model of task-oriented behavior selection and working memory in prefrontal cortex J Cogn Neurosci In press.
Bota M Arbib MA 2004 Integrating databases and expert systems for the analysis of brain structures: connections, similarities, and homologies Neuroinformatics 2 19 58 15067167
|
16201007
|
PMC1239902
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Sep 30; 1(4):e42
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010042
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1620100810.1371/journal.pcbi.001004405-PLCB-RA-0077R2plcb-01-04-07Research ArticleBioinformatics - Computational BiologyEvolutionGenetics/EvolutionRetrotransposonsAluPrimatesHuman EvolutionModeling the Amplification Dynamics of Human Alu Retrotransposons Alu Retrotransposition Dynamics
Hedges Dale J 1Cordaux Richard 1Xing Jinchuan 1Witherspoon David J 2Rogers Alan R 3Jorde Lynn B 2Batzer Mark A 1*1 Department of Biological Sciences, Biological Computation and Visualization Center, Center for Bio-Modular Microsystems, Louisiana State University, Baton Rouge, Louisiana, United States of America
2 Department of Human Genetics, University of Utah Health Sciences Center, Salt Lake City, Utah, United States of America
3 Department of Anthropology, University of Utah, Salt Lake City, Utah, United States of America
Holmes Eddie EditorPennsylvania State University, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 4 e4415 4 2005 24 8 2005 Copyright: © 2005 Hedges et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Retrotransposons have had a considerable impact on the overall architecture of the human genome. Currently, there are three lineages of retrotransposons (Alu, L1, and SVA) that are believed to be actively replicating in humans. While estimates of their copy number, sequence diversity, and levels of insertion polymorphism can readily be obtained from existing genomic sequence data and population sampling, a detailed understanding of the temporal pattern of retrotransposon amplification remains elusive. Here we pose the question of whether, using genomic sequence and population frequency data from extant taxa, one can adequately reconstruct historical amplification patterns. To this end, we developed a computer simulation that incorporates several known aspects of primate Alu retrotransposon biology and accommodates sampling effects resulting from the methods by which mobile elements are typically discovered and characterized. By modeling a number of amplification scenarios and comparing simulation-generated expectations to empirical data gathered from existing Alu subfamilies, we were able to statistically reject a number of amplification scenarios for individual subfamilies, including that of a rapid expansion or explosion of Alu amplification at the time of human–chimpanzee divergence.
Synopsis
Nearly 50% of the human genome is composed of mobile elements. While much of this sequence consists of inactive “fossil” elements that are no longer actively moving or generating new copies, three families are currently proliferating in human genomes. Among these, the Alu lineage has reached a copy number of over 1 million and alone accounts for approximately 10% of the genome. While considerable evidence has been gathered concerning the underlying biological mechanisms of Alu mobilization and proliferation, a detailed understanding of Alu amplification history is currently lacking. Researchers are aware, for example, that several thousand Alu elements have inserted within the human genome since the divergence of humans and chimpanzees, but how those insertions were distributed over this ~6-million-year time period is currently unknown. In this work, the authors introduce a simulation framework that seeks to incorporate both sequence diversity and empirically gathered population data from human Alu elements, in order to provide a better understanding of the last several million years of human Alu evolution. The results suggest that a rapid explosion of Alu amplification at the time of the human–chimpanzee divergence is unlikely. Therefore, it is improbable that an increase in Alu retrotransposition activity was involved in the speciation of humans and chimpanzees.
Citation:Hedges DJ, Cordaux R, Xing J, Witherspoon DJ, Rogers AR, et al. (2005) Modeling the amplification dynamics of human Alu retrotransposons. PLoS Comput Biol 1(4): e44.
==== Body
Introduction
A collection of evolutionarily recent and older “fossil” mobile element sequences compose more than 45% of the human genome [1–5]. Along with the recently characterized SVA family, Alu and L1 have the distinction of being the only mobile element lineages to be actively proliferating in modern humans [3,6,7]. All three of these lineages belong to the retrotransposon class of mobile elements, replicating themselves via an RNA intermediate [6,8]. They differ, however, in that L1 retrotransposons are ~6-kb-long autonomous elements that encode the proteins required for their retrotransposition [2] while Alu and SVA retrotransposons are shorter, non-autonomous elements that are trans-mobilized by the L1 protein machinery [9]. L1 elements have been active in mammalian genomes for the past 150 million years (myrs) and have reached a copy number of ~0.5 million in the human genome, and Alu retrotransposons have reached a copy number of ~1.1 million within the past 65 myrs [1]. By comparison, the SVA lineage is a relative newcomer to the primate lineage, having achieved a copy number of approximately 5,000 copies in the human genome over the last 15 myrs [7]. Together, the amplification activity of these retrotransposon families has had a substantial impact on their host genomes. In addition to contributing to genome size expansion, they have shaped the architecture of the human genome by mediating genetic exchanges such as duplications, deletions, inversions, transductions, and translocations [6,8,10–17]. L1 and Alu have also been implicated in DNA repair [18] and alteration of gene expression [2,19–21]. As they are still actively retrotransposing and thus acting as insertional mutagens, Alu, SVA, and L1 elements are responsible for more than 0.5% of all human genetic disorders [2,22,23].
While much attention has been given to the underlying biology driving retrotransposon expansion in primates, little attempt has been made to assess what can broadly be described as “amplification dynamics.” Under this heading we include the evolutionary window during which lineages were actively retrotransposing, the intensity of retrotransposition, and the degree of rate fluctuation during this period. Notable exceptions to this general dearth of information concerning mobile element amplification dynamics include data for mobile element activity in Drosophila species [24–26]. While a considerable body of theoretical work exists concerning mobile element expansion (e.g., [27–33]), these models generally focus on element copy number behavior under equilibrium conditions and do not address the impact of diverse amplification histories on sequence composition. The observation of divergent mobile element retrotransposition levels among closely related host species [24,34], however, suggests that the assumption of equilibrium may often be unrealistic, as noted in [35]. A more complete understanding of how mobile element sequence structure and frequencies are influenced by diverse nonequilibrium expansion scenarios would be invaluable for developing realistic models of how transposable elements spread through given taxa.
The problem we are faced with is how to reconstruct the evolutionary amplification history of a mobile element lineage given only a static snapshot of sequence and polymorphism data from present-day genomes. Previously, efforts used the phylogenetic distribution of mobile element lineages to bound their period of activity in time by the divergence dates of their host taxa (e.g., [36–38]). While such analyses can provide useful information, particularly where allele frequency information is unavailable, they nevertheless cannot yield the kind of temporal resolution that would be most helpful in understanding the amplification process. For example, we know that some 6,000–7,000 Alu elements have fixed in the human genome since Pan troglodytes and Homo sapiens last shared a common ancestor 5–8 myrs ago [39–41], but the temporal pattern of expansion giving rise to these elements remains unknown. Age estimates of individual retrotransposon insertions based on sequence divergence from a consensus typically possess a great degree of uncertainty because of the relatively short sequence lengths of many retrotransposons, particularly among short interspersed elements, as well as because of uncertainty over the accuracy of the consensus “source” sequence used for comparison [42–45]. In younger, recently active retrotransposon lineages, an additional piece of evidence is at our disposal to aid in reconstructing their amplification history. For these families, we are able to obtain population frequency data for insertions at given loci, which allow estimation of the percentage of polymorphic loci for presence/absence in the corresponding subfamilies (termed in the following text “insertion polymorphism level” [IPL]).
Alone, sequence diversity and IPL prove insufficient to reconstruct the historical amplification pattern of a mobile element family with any degree of accuracy. When effectively combined, however, we hypothesized that they may serve to narrow the alternative scenarios. We tested this hypothesis by focusing on the Alu family of retrotransposons, for which subfamily structure is well characterized and population frequency data are available for a number of distinct subfamilies [3,39–41,46]. Furthermore, Alu retrotransposons presented an attractive target for this initial study because, as detailed below, they possess several features that make the process of modeling their retrotransposition more tractable. It was first necessary to determine what set of Alu sequence and IPL observations might be expected under various evolutionary amplification patterns. To generate quantitative expectations for these parameters under diverse patterns of expansion, we developed a computer simulation that incorporates established aspects of Alu retrotransposon biology (see Materials and Methods). Our simulation also accommodates the effect of significant sampling biases inherent in the way Alu elements have been characterized in the human genome. By comparing existing Alu sequence diversity and polymorphism levels, we were able to statistically reject multiple amplification scenarios for individual Alu subfamilies, resulting in a more refined understanding of the retrotransposition dynamics of human-specific Alu subfamilies.
Results/Discussion
The Alu Simulation Framework
Two fundamental processes underlie the various descriptive statistics that can be tabulated from genomic Alu sequences, namely the post-insertion evolution of Alu nucleotide sequences and the associated evolution of insertion polymorphism allele frequencies. To make the modeling process more straightforward, we divided these processes into distinct core simulator programs, one to model the behavior of nucleotide sequence and one to model the behavior of inserted retrotransposon allele frequencies. Several of the known properties of Alu subfamily structure and sequence mutation patterns were incorporated within the programs (see Materials and Methods). Both programs implement a strict “master gene” model of Alu retrotransposition under which a single source element produces inert, non-retrotransposing copies [47]. While it has been demonstrated that most Alu subfamilies deviate from the strict master gene model, this scenario can nevertheless explain the majority of overall subfamily sequence structure [48]. More importantly, implementing deviations from the master gene model (i.e., secondary and tertiary retrotransposition sources and so on) can lead to exponential copy number increase when limiting factors such as negative selection do not constrain numbers, a scenario which is clearly not historically accurate. In this analysis, we have restricted our simulation to neutrally evolving loci within a panmictic population of constant size. In our model we also assume that retrotransposition rates (RRs) do not fluctuate during the window of time during which retrotransposition occurs.
The above assumptions are almost certainly oversimplifications, but are necessary to keep the number of amplification scenarios at a manageable level in this initial investigation. We believe the existence of secondary source genes would have a limited impact on our analysis because any secondary source that is active enough to produce appreciable copy numbers would be classified as a separate subfamily under current naming conventions, and it would be analyzed separately in our approach. The effect of population substructure is more difficult to anticipate because the nature of population substructure during the time period in question is largely unknown. Significant population structure would impact the behavior of polymorphisms by extending their average persistence time, affecting the rate of insertion polymorphism decay both during and after transposition. The nature and magnitude of these effects will be the subject of future investigations.
For both sequence mutation and frequency drift simulations, retrotransposition started at time t
0 and proceeded at a constant rate for a time window T
retro. Thus, given a subfamily copy number n, T
retro defines the RR of the simulation (i.e., RR = n/T
retro). For the sequence mutation simulations, elements were allowed to mutate neutrally from their initial time of retrotransposition until the end of the simulation. T
mut represents the total elapsed simulation time, which is also the amount of time the oldest element in the subfamily has been evolving. We have chosen a maximum T
mut of 6 myrs, as this roughly corresponds to the human–chimpanzee divergence time and, thus, is suitable for investigating the amplification dynamics of recent human Alu subfamilies. During the course of each run, sequence variation and allele frequency statistics (described in detail below) were calculated at 100,000-y observation intervals. One thousand replicates were performed under each of seven basic amplification models (M0 to M6), ranging from M0, which has an instantaneous burst of insertion activity generating all subfamily members, to M6, in which new retrotransposition events occur at a uniform rate from the beginning of the subfamily throughout the entire simulation of 6 myrs. Intermediate models (M1 to M5), in which amplification occurred for 1 to 5 myrs and then ceased, were also evaluated. Simulations were performed using a human effective population size (N
e) of 10,000 individuals and a generation time of 25 y. To assess the impact of alternative values for N
e and generation time, we also performed simulations using N
e values of 5,000, 15,000, and 20,000 individuals as well as generation times of 20 and 30 y.
Amplification History and Sequence Variation
As an estimator of Alu subfamily sequence variation, we used the parameter π, which is defined as the mean number of nucleotide differences observed among all pairs of Alu sequences in the subfamily [49]. For example, a π value of three means that there are, on average, three nucleotide differences between any two Alu sequences in the subfamily. Means, modes, and standard deviations for π were calculated across all replicates (available at http://batzerlab.lsu.edu). In addition to π, we evaluated the use of the mismatch distribution raggedness index as a metric of sequence diversity [50], but its informativeness proved limited for our purposes, and it was excluded from subsequent analyses.
As expected, mean π values increased linearly with time in our simulation (Figure 1A). The effect of retrotransposition during T
retro is to slow the rate of increase in π. In scenarios M1 through M6, where retrotransposition occurs for a period of time then ceases, the rate of π increase becomes steeper (though still linear) immediately following the cessation of retrotransposition (Figure 1A). A clear relationship exists between sequence diversity and the particular amplification model of the family. For example, a scenario with a burst of retrotransposition followed by dormancy leads to higher π values than scenarios where RR has been uniform over long periods of time. This result is intuitive, as any scenario resulting in an increased element insertion number earlier in a subfamily's history will result in additional opportunity for mutation and consequently higher π values. The problem, however, is that when evaluating real mobile element data, the time of onset of retrotransposition (i.e., the beginning of T
retro) is typically unknown. From examining Figure 1A, it is evident that any value of π can be obtained by any model, provided that an appropriate amount of time (T
mut) has elapsed prior to the point of observation. Thus, although π is directly influenced by the particular amplification history, it cannot be used to infer that history without additional information.
Figure 1 Temporal Variation of Subfamily Sequence Variation π and IPL
Results for three expansion models are shown, in which retrotransposition activity was instantaneous (M0) or lasted for 3 (M3) or 6 (M6) myrs. Variation in π (A) is slowed during retrotransposition, but increases immediately upon the cessation of retrotransposition. Rate of IPL decay (B) is attenuated during retrotransposition activity but increases once retrotransposition ends.
Alu Insertion Polymorphism
In addition to π, we also modeled the behavior of IPL, which indicates the percentage of polymorphic insertion loci in a subfamily. Like π, IPL is expected to be influenced by both the age of the subfamily and its historical pattern of retrotransposition. Figure 1B illustrates the decay of IPL over time under models M0, M3, and M6. As might be expected, ongoing retrotransposition in a mobile element family slows the rate of IPL decay by providing an influx of new polymorphisms. When retrotransposition ceases, IPL falls relatively rapidly over the course of approximately 1 myrs. This rate of IPL breakdown is consistent with the expected on average 1-myr coalescence time (4N
e generations, where N
e is the effective population size) for our simulated human effective population size of 10,000 individuals. As with π, there is clearly an effect of amplification history on IPL values, with IPL values remaining higher for those families whose T
retro windows extended closer to the present day. But also, as is the case with π, any scenario can yield a given IPL value depending on what time point (T
mut value) is being sampled. A researcher examining empirical Alu frequency data does not know what position his or her data occupy on the timeline of the model of retrotransposition being considered (Figure 1B). Yet, as we demonstrate below, for a given model there exists a set of IPL and π parameters that are mutually exclusive across a range of time points. As a consequence, by combining the π and IPL statistics, one can effectively narrow the possible range of amplification histories for a given Alu subfamily.
Inferring Amplification Scenarios from Genomic Alu Data
Plots of IPL versus π for equivalent time points over the course of seven amplification scenarios (i.e., models M0–M6) are shown in Figure 2, based on a generation time of 25 y and an effective population size of 10,000 individuals. The 95% confidence intervals, generated from 1,000 simulation replicates, are represented as the bounded area in each graph (see Materials and Methods). In addition, π values were estimated for ten human Alu subfamilies for which IPL data are available (Table 1). These data were collected from subsets of elements from the respective polymorphic subfamilies for which population data were available. For each of these subfamilies, the relationship between IPL and π is indicated in Figure 2. In our analysis, if a subfamily's IPL versus π point falls within the 95% confidence interval of a given model's results, the model cannot be excluded as a possible amplification pattern (see Materials and Methods for details). Conversely, when a subfamily's data point falls outside the bounded area, that model can be excluded for the subfamily in question.
Figure 2 Distribution of Subfamily Sequence Variation π (x-Axis) versus IPL (y-Axis)
Expectations based on 1,000 replicates of expansion models M0–M6. Shaded area in each plot indicates 95% of resulting values for each model. Observed (π and IPL) values for ten recent human Alu subfamilies are shown as black diamonds. These results are based on a generation time of 25 y and an effective population size of 10,000 individuals.
Table 1
Alu Subfamily Diversity (π) and IPL Parameters and Their Age under Different Models of Amplification
Impact of Effective Population Size and Generation Time Parameters
Our initial simulation replicates were conducted under the conditions of a 25-y generation time and effective population size of 10,000 breeding individuals. While these represent commonly accepted values for these parameters, we also investigated the impact of a broader range of generation times (20 and 30 y) and N
e (5,000, 15,000, and 20,000) on the simulations. For N
e = 5,000, our models fail to encompass most of the observed data for extant Alu subfamilies (Figure S1; Table S1). This is not unexpected, as this N
e value is approximately half that of most literature estimates. Likewise, N
e = 20,000 yields IPL and π values that are largely not concordant with observed Alu subfamily data (Figure S2; Table S1). N
e values of 10,000 and 15,000 individuals manage to encompass the majority of observed Alu data points (Figures 2 and S3; Tables 1 and S1). In this respect, the behavior of our simulations is congruent with current literature estimates, which place the human N
e on the order of 10,000 to 15,000 individuals [51–53].
Altering the generation time also had an appreciable effect on simulation behavior by shifting the timescale of the simulated data. While a generation time of 20 y did not perform very well (Figure S4; Table S2), our models were generally able to encompass more observed Alu subfamily data under generation time parameters of 25 and 30 y (Figures 2 and S5; Tables 1 and S2). Such values lie within the range of currently estimated values for ancestral generation times spanning the relevant period ([54] and references therein). Also, as discussed below, a higher generation time parameter of 30 y has the effect of bringing Alu subfamily age estimates derived from our simulation closer in line with previous literature values determined by other methods.
Estimating the Age of Alu Subfamilies
Once improbable amplification scenarios are excluded (Tables 1, S1, and S2), it is possible to determine time periods of amplification for subfamilies that are compatible with both their π and IPL values. By using the present time as a point of reference (i.e., T
mut = present), it is further possible to infer the age of the subfamilies. Figure 3 illustrates this process. In this example, the Ya5a2 subfamily has a π value (0.65) that is consistent with an age ranging from 0.6 to 1.0 myrs before present under M4 (N
e = 10,000, generation time = 25 y). Within that range, the Ya5a2 IPL value is only compatible with 0.7 to 1.0 myrs before present. Estimated age ranges that are consistent with both π and IPL for all Alu subfamilies analyzed in this study under a generation time of 25 y and N
e of 10,000 are given in Table 1. We note here that Alu subfamily age estimates derived in this study are generally higher than those reported in previous literature [42]. However, the age estimates obtained from sequence diversity alone typically have large standard deviations [42] that overlap with our estimates derived from both sequence diversity and IPL. This might indicate that time estimates derived from sequence diversity alone may underestimate the true ages of the subfamilies. Nevertheless, alternative values of N
e and generation time also have an impact on the potential age of the subfamily as estimated by our method. For example, when age calculations are made using a generation time of 30 y, our age estimates more closely approximate those of previous literature.
Figure 3 Estimation of the Age of the Ya5a2 Alu Subfamily under Simulation M4
In M4, N
e is 10,000 and generation time is 25 y. Data are based on observed subfamily sequence variation π and IPL parameters. Time estimates consistent with π and IPL values are indicated in boxes. The bold double arrow indicates age estimates concordant with both parameters.
Our results suggest that, while a range of retrotransposition scenarios remain possible for each subfamily, some alternatives can be statistically rejected. Notably, when using standard values for effective population size (N
e =10,000) and generation time (25 y), our results exclude the possibility that the majority of human Alu insertions occurred during a brief, intense burst of retrotransposition activity after the divergence between humans and chimpanzees. Such a scenario, intermediate between M1 and M0 (instantaneous) results in an IPL versus π distribution well outside the observed data points (see Figure 2). This result is well supported because variation in effective population size and generation time leads to the same conclusion (Figures S1–S5). Thus, these analyses provide evidence against the notion of a burst of retrotransposition at or near the human–chimpanzee divergence. This result is consistent with a previous study [34], which suggested that the marked increase in human Alu fixation events with respect to chimpanzee was initiated within the past 4 myrs. The involvement of mobile element amplification activity in the formation of reproductive barriers, and hence speciation, has received a fair amount of attention [55–58], although definitive evidence is lacking. The discovery of high levels of mobile element activity in humans compared to chimpanzees [34,59] has invited speculation as to whether or not the Alu retrotransposition increase might have been involved in the speciation of humans and chimpanzees [59]. While our present results do not support an increase in mobile element activity at the time of the human–chimpanzee divergence, they do not exclude the possibility of such an event during a later hominid speciation event. Furthermore, the possibility remains that an extended simulation model, one that accounts for additional biological and spatial parameters, may generate results that are consistent with a retrotransposon burst at the time of speciation.
Conclusion
We have demonstrated that it is possible to mine information concerning the amplification history of a retrotransposon subfamily from present-day genomic and population data. Overall, there appears to be heterogeneity in both the timing and intensity of human Alu subfamily activity. Our simulations do not presently accommodate the influence of host population subdivision, RR fluctuations (i.e., rate heterogeneity) over time, and selection on patterns of retrotransposon evolution. All of these phenomena will likely have some bearing on the nucleotide divergence and IPL, although the extent of that influence is difficult to anticipate. We plan to extend our simulations to encompass these and other potentially relevant phenomena in further studies. Nevertheless, the present analyses do indicate that the combination of retrotransposon sequence divergence and subfamily polymorphism information has the potential to reveal information about the historical dynamics of mobile element amplification that has thus far remained inaccessible. In particular, by applying our method we are able to rigorously address the issue of the time window during which amplification occurred. A more detailed account of the history of retrotransposon activity will allow for a better understanding of the forces that influence mobile element activity across diverse taxa.
Materials and Methods
Simulating Alu sequence evolution.
We developed a simulator of Alu sequence evolution that takes into account most of the major known properties of Alu elements in terms of subfamily structure and sequence mutation patterns. Specifically, Alu elements begin accumulating nucleotide substitutions stochastically, starting at the time of retrotransposition and until the end of the simulation. Nucleotide substitution was simulated using the Kimura two-parameter reversible mutation model, a neutral mutation rate at non-CpG dinucleotides of 0.0015 mutations/site/myrs [60] and a transition to transversion ratio of four. To accommodate the known mutation bias for Alu CpG dinucleotides as a result of the deamination of methylated cytosines, CpG dinucleotides were allowed to mutate at a 6-fold higher rate than non-CpG dinucleotides [42]. To make the modeling process more computationally tractable, we assumed a scenario of Alu subfamily evolution in which Alu retrotransposition followed a strict master gene model, with a lone, non-mutating source sequence generating offspring that were incapable of additional retrotransposition. We also considered the Alu expansion to have occurred in a single, representative genome, with each successful retrotransposition event equivalent to a “substitution” event at the population level. This allowed for combining Alu retrotransposition events with standard methods for calculating substitution probabilities, greatly reducing simulation complexity and computational time.
Decay of IPLs.
To study the evolution of IPL during the transposition process, we modeled the behavior of IPL under the same model conditions as π (i.e., M0 through M6, as described above). Given the low probability of fixation for each initial insertion event (1/2N
e), several million retrotransposition events must ultimately be followed in order to achieve final subfamily copy numbers comparable to those observed in the human genome. In each model, 7 million insertion events occurring over various windows of time were used to yield approximately 350 fixed elements. To reduce computational time, Kimura's recursion approximation of the diffusion process was used to simulate the neutral drift of retrotransposed elements [61]. The absorption boundaries [0,1] at which alleles were lost or fixed, respectively, were adjusted slightly to compensate for disparities between the continuous results from the recursion equation and the discrete frequencies that real-world alleles can assume. (The continuous values between zero and 1/2N
e are possible return values from the recursion, but not realistic allele frequencies.) A generation time of 25 y and effective population size of 10,000 interbreeding individuals was used. To address uncertainty surrounding ancestral human generation times and effective population sizes, the effects of a range of generation times (20, 25, and 30 y) and effective population sizes (5,000, 10,000, 15,000, and 20,000) were investigated. At the onset of the simulation, the number of retrotranspositions per time increment required to achieve the 7 million insertion target was calculated. Allele frequencies were allowed to drift randomly both during and after transposition windows, and IPL values were calculated and reported at 100,000-y intervals.
Accounting for IPL sampling effects.
To adequately model the element copy number and IPL values observed in the human genome, the manner in which genomic elements are ascertained and characterized was also incorporated into the simulation. The population sample size from which most Alu elements have so far been initially discovered is effectively a single individual (i.e., the human genome draft sequence), and, consequently, a considerable number of polymorphic elements will remain unobserved. When simulating the observed IPL value, the effect of ascertaining elements from a single individual must be accommodated. In order to do so, the number of polymorphic elements that were reported as “observed” at any given time during the simulation was determined by effectively sampling a single individual from the simulated population. In this step, the detection of a given Alu insertion polymorphism within that individual was stochastically determined, with the probability of observing a given insertion being proportional to the frequency of the insertion in the population. The simulations were implemented in a set of C language programs with assisting Perl scripts and are available at http://batzerlab.lsu.edu.
Statistical evaluation of models.
Models were excluded or not excluded based on 95% confidence intervals generated through simulation. For each model scenario (M0 to M6), 1,000 replicates were simulated. IPL and π values were calculated at 100,000-y intervals for the simulated datasets, and the lower and upper 1.265 percentiles were used to determine the 95% confidence interval. Boundary values for 95% confidence interval were adjusted for the effect of two independent tests of the IPL and π parameters resulting from the model. Here, the probability of falling outside the range of some percentage, X, of the simulated data twice (two tests) is given by 1 – (1 − X)2. To determine the boundaries that would be appropriate at the 5% significance level, we solved the equation 1 – (1 − X)2 = 0.05, yielding X = 0.0253. Upper and lower boundaries were then 0.0253/2 = 0.01265. π versus IPL values for real Alu subfamilies were then plotted together with the simulated data. If a given subfamily's π versus IPL data fell outside the 95% confidence interval of a given model, the model was rejected for that subfamily.
Evaluating the impact of subfamily size.
All the analyses above were conducted using subfamily copy numbers of approximately 350 elements for the nucleotide evolution simulation and 7 million insertion events (corresponding to ~350 fixations) for IPL modeling. To assess the impact of subfamily size on the behavior of π, we simulated sequence evolution for T
mut = 2 myrs in an Alu subfamily having generated n = 50, 100, 200, and 400 copies under a retrotransposition model where all elements were produced at t
0 (i.e., T
retro = 0). We performed 100 simulation replicates for each value of n. We found that the major effect of increasing n was to decrease the standard deviation of π among trials, but otherwise copy number had little impact on the behavior of π over time (Figure 4). The reduction of between-trial variance due to increasing family size stabilized at copy numbers greater than 100 elements. We therefore ran the all simulations described above using n = 350 elements, a number that is in the same order of magnitude of size as most of the observed Alu subfamilies used in our study. Similar tests were conducted for IPL simulations using alternate insertion numbers (1 million, 5 million, and 10 million). While some subfamilies in the study, namely Ya5 and Yb8, are considerably larger than 350 in observed copy number, experimentation with copy numbers as high as 5,000 demonstrate that higher subfamily sizes reduces between-replicate variance (data not shown).
Figure 4 Impact of Subfamily Copy Number (n) on the Sequence Variation π Parameter
Increasing subfamily size beyond 100 copies had little effect on between-replicate variation.
Supporting Information
Figure S1 Distribution of Subfamily Sequence Variation π (x-Axis) versus IPL (y-Axis): Generation Time of 25 y and N
e of 5,000 Individuals
Expectations based on 1,000 replicates of expansion models M0–M6. The two lines indicate the boundaries of the 95% confidence interval for each model. Observed (π and IPL) values for ten recent human Alu subfamilies are shown as black diamonds (see legend of Figure 2).
(3325 KB TIF)
Click here for additional data file.
Figure S2 Distribution of Subfamily Sequence Variation π (x-Axis) versus IPL (y-Axis): Generation Time of 25 y and N
e of 20,000 Individuals
Expectations based on 1,000 replicates of expansion models M0–M6. The two lines indicate the boundaries of the 95% confidence interval for each model. Observed (π and IPL) values for ten recent human Alu subfamilies are shown as black diamonds (see legend of Figure 2).
(3.3 MB TIF)
Click here for additional data file.
Figure S3 Distribution of Subfamily Sequence Variation π (x-Axis) versus IPL (y-Axis): Generation Time of 25 y and N
e of 15,000 Individuals
Expectations based on 1,000 replicates of expansion models M0–M6. The two lines indicate the boundaries of the 95% confidence interval for each model. Observed (π and IPL) values for ten recent human Alu subfamilies are shown as black diamonds (see legend of Figure 2).
(3.3 MB TIF)
Click here for additional data file.
Figure S4 Distribution of Subfamily Sequence Variation π (x-Axis) versus IPL (y-Axis): Generation Time of 20 y and N
e of 10,000 Individuals
Expectations based on 1,000 replicates of expansion models M0–M6. The two lines indicate the boundaries of the 95% confidence interval for each model. Observed (π and IPL) values for ten recent human Alu subfamilies are shown as black diamonds (see legend of Figure 2).
(3.4 MB TIF)
Click here for additional data file.
Figure S5 Distribution of Subfamily Sequence Variation π (x-Axis) versus IPL (y-Axis): Generation Time of 30 y and N
e of 10,000 Individuals
Expectations based on 1,000 replicates of expansion models M0–M6. The two lines indicate the boundaries of the 95% confidence interval for each model. Observed (π and IPL) values for ten recent human Alu subfamilies are shown as black diamonds (see legend of Figure 2).
(3.4 MB TIF)
Click here for additional data file.
Table S1
Alu Subfamily Compatibility with Different Retrotransposition Models (M0–M6) for Different Effective Population Sizes (N
e) and a Generation Time of 25 y
(72 KB DOC)
Click here for additional data file.
Table S2
Alu Subfamily Compatibility with Different Retrotransposition Models (M0–M6) for Different Generation Times and an Effective Population Size of 10,000 Individuals
(56 KB DOC)
Click here for additional data file.
We thank Dr. Scott W. Herke for comments on an earlier version of the manuscript. This research was supported by Louisiana Board of Regents Millennium Trust Health Excellence Fund HEF (2000–05)-05 (MAB), (2000–05)-01 (MAB), and (2001–06)-02 (MAB); National Institutes of Health RO1 GM59290 (LBJ and MAB); National Science Foundation BCS-0218338 (MAB), BCS-0218370 (LBJ) and EPS-0346411 (MAB); and the State of Louisiana Board of Regents Support Fund (MAB).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DJH and RC conceived and designed the experiments. DJH performed the experiments. DJH, RC, JX, DJW, ARR, LBJ, and MAB analyzed the data. LBJ and MAB contributed reagents/materials/analysis tools. DJH and RC wrote the paper.
Abbreviations
IPLinsertion polymorphism level
M[number]model [number]
myrsmillion years
RRretrotransposition rate
==== Refs
References
Lander ES Linton LM Birren B Nusbaum C Zody MC 2001 Initial sequencing and analysis of the human genome Nature 409 860 921 11237011
Ostertag EM Kazazian HH Jr 2001 Biology of mammalian L1 retrotransposons Annu Rev Genet 35 501 538 11700292
Batzer MA Deininger PL 2002 Alu repeats and human genomic diversity Nat Rev Genet 3 370 379 11988762
Smit AF 1999 Interspersed repeats and other mementos of transposable elements in mammalian genomes Curr Opin Genet Dev 9 657 663 10607616
Jurka J 2004 Evolutionary impact of human Alu repetitive elements Curr Opin Genet Dev 14 603 608 15531153
Kazazian HH Jr 2004 Mobile elements: Drivers of genome evolution Science 303 1626 1632 15016989
Ostertag EM Goodier JL Zhang Y Kazazian HH Jr 2003 SVA elements are nonautonomous retrotransposons that cause disease in humans Am J Hum Genet 73 1444 1451 14628287
Deininger PL Batzer MA 2002 Mammalian retroelements Genome Res 12 1455 1465 12368238
Dewannieux M Esnault C Heidmann T 2003 LINE-mediated retrotransposition of marked Alu sequences Nat Genet 35 41 48 12897783
Moran JV DeBerardinis RJ Kazazian HH Jr 1999 Exon shuffling by L1 retrotransposition Science 283 1530 1534 10066175
Hayakawa T Satta Y Gagneux P Varki A Takahata N 2001 Alu-mediated inactivation of the human CMP- N-acetylneuraminic acid hydroxylase gene Proc Natl Acad Sci U S A 98 11399 11404 11562455
Bailey JA Liu G Eichler EE 2003 An Alu transposition model for the origin and expansion of human segmental duplications Am J Hum Genet 73 823 834 14505274
Salem AH Kilroy GE Watkins WS Jorde LB Batzer MA 2003 Recently integrated Alu elements and human genomic diversity Mol Biol Evol 20 1349 1361 12777511
Moran JV Holmes SE Naas TP DeBerardinis RJ Boeke JD 1996 High frequency retrotransposition in cultured mammalian cells Cell 87 917 927 8945518
Feng Q Moran JV Kazazian HH Jr Boeke JD 1996 Human L1 retrotransposon encodes a conserved endonuclease required for retrotransposition Cell 87 905 916 8945517
Symer DE Connelly C Szak ST Caputo EM Cost GJ 2002 Human l1 retrotransposition is associated with genetic instability in vivo Cell 110 327 338 12176320
Callinan PA Wang J Herke SW Garber RK Liang P 2005 Alu retrotransposition-mediated deletion J Mol Biol 348 791 800 15843013
Morrish TA Gilbert N Myers JS Vincent BJ Stamato TD 2002 DNA repair mediated by endonuclease-independent LINE-1 retrotransposition Nat Genet 31 159 165 12006980
Lev-Maor G Sorek R Shomron N Ast G 2003 The birth of an alternatively spliced exon: 3′ splice-site selection in Alu exons Science 300 1288 1291 12764196
Perepelitsa-Belancio V Deininger P 2003 RNA truncation by premature polyadenylation attenuates human mobile element activity Nat Genet 35 363 366 14625551
Han JS Szak ST Boeke JD 2004 Transcriptional disruption by the L1 retrotransposon and implications for mammalian transcriptomes Nature 429 268 274 15152245
Kazazian HH Jr 1998 Mobile elements and disease Curr Opin Genet Dev 8 343 350 9690999
Deininger PL Batzer MA 1999 Alu repeats and human disease Mol Genet Metab 67 183 193 10381326
Vieira C Biemont C 2004 Transposable element dynamics in two sibling species: Drosophila melanogaster and Drosophila simulans
Genetica 120 115 123 15088652
Vieira C Lepetit D Dumont S Biemont C 1999 Wake up of transposable elements following Drosophila simulans worldwide colonization Mol Biol Evol 16 1251 1255 10486980
Kidwell MG 1979 Hybrid dysgenesis in Drosophila melanogaster —Relationship between the P-M and I-R interaction systems Genet Res 33 205 217
Brookfield JF 1991 Models of repression of transposition in P-M hybrid dysgenesis by P cytotype and by zygotically encoded repressor proteins Genetics 128 471 486 1649073
Charlesworth B 1988 The maintenance of transposable elements in natural populations Basic Life Sci 47 189 212 2845912
Brookfield JF Badge RM 1997 Population genetics models of transposable elements Genetica 100 281 294 9440281
Biemont C Lemeunier F Garcia Guerreiro MP Brookfield JF Gautier C 1994 Population dynamics of the copia, mdg1, mdg3, gypsy, and P transposable elements in a natural population of Drosophila melanogaster
Genet Res 63 197 212 8082837
Brookfield JF 1986 The population biology of transposable elements Philos Trans R Soc Lond B Biol Sci 312 217 226 2870518
Edwards RJ Brookfield JF 2003 Transiently beneficial insertions could maintain mobile DNA sequences in variable environments Mol Biol Evol 20 30 37 12519903
Montgomery E Charlesworth B Langley CH 1987 A test for the role of natural selection in the stabilization of transposable element copy number in a population of Drosophila melanogaster
Genet Res 49 31 41 3032743
Hedges DJ Callinan PA Cordaux R Xing J Barnes E 2004 Differential alu mobilization and polymorphism among the human and chimpanzee lineages Genome Res 14 1068 1075 15173113
Brookfield JF 2005 The ecology of the genome—Mobile DNA elements and their hosts Nat Rev Genet 6 128 136 15640810
Leeflang EP Liu WM Chesnokov IN Schmid CW 1993 Phylogenetic isolation of a human Alu founder gene: Drift to new subfamily identity [corrected] J Mol Evol 37 559 565 8114108
Quentin Y 1988 The Alu family developed through successive waves of fixation closely connected with primate lineage history J Mol Evol 27 194 202 3138422
Shaikh TH Deininger PL 1996 The role and amplification of the HS Alu subfamily founder gene J Mol Evol 42 15 21 8576958
Carter AB Salem AH Hedges DJ Keegan CN Kimball B 2004 Genome-wide analysis of the human Alu Yb-lineage Hum Genomics 1 167 178 15588477
Otieno AC Carter AB Hedges DJ Walker JA Ray DA 2004 Analysis of the human Alu Ya-lineage J Mol Biol 342 109 118 15313610
Carroll ML Roy-Engel AM Nguyen SV Salem AH Vogel E 2001 Large-scale analysis of the Alu Ya5 and Yb8 subfamilies and their contribution to human genomic diversity J Mol Biol 311 17 40 11469855
Xing JC Hedges DJ Han KD Wang H Cordaux R 2004 Alu element mutation spectra: Molecular clocks and the effect of DNA methylation J Mol Biol 344 675 682 15533437
Batzer MA Kilroy GE Richard PE Shaikh TH Desselle TD 1990 Structure and variability of recently inserted Alu family members Nucleic Acids Res 18 6793 6798 2175877
Jurka J Milosavljevic A 1991 Reconstruction and analysis of human Alu genes J Mol Evol 32 105 121 1706781
Labuda D Striker G 1989 Sequence conservation in Alu evolution Nucleic Acids Res 17 2477 2491 2541408
Roy-Engel AM Carroll ML Vogel E Garber RK Nguyen SV 2001 Alu insertion polymorphisms for the study of human genomic diversity Genetics 159 279 290 11560904
Deininger PL Batzer MA Hutchison CA 3rd Edgell MH 1992 Master genes in mammalian repetitive DNA amplification Trends Genet 8 307 311 1365396
Cordaux R Hedges DJ Batzer MA 2004 Retrotransposition of Alu elements: How many sources? Trends Genet 20 464 467 15363897
Tajima F 1983 Evolutionary relationship of DNA sequences in finite populations Genetics 105 437 460 6628982
Harpending HC Sherry ST Rogers AR Stoneking M 1993 The genetic structure of ancient human populations Curr Anthropol 34 483 496
Graur D Li WH 2000 Fundamentals of molecular evolution Sunderland Sinauer Associates 482 p.
Sherry ST Harpending HC Batzer MA Stoneking M 1997 Alu evolution in human populations: Using the coalescent to estimate effective population size Genetics 147 1977 1982 9409852
Harpending HC Batzer MA Gurven M Jorde LB Rogers AR 1998 Genetic traces of ancient demography Proc Natl Acad Sci U S A 95 1961 1967 9465125
Fenner JN 2005 Cross-cultural estimation of the human generation interval for use in genetics-based population divergence studies Am J Phys Anthropol E-pub ahead of print.
Hurst GD Werren JH 2001 The role of selfish genetic elements in eukaryotic evolution Nat Rev Genet 2 597 606 11483984
Kidwell MG Lisch DR 1998 Hybrid genetics. Transposons unbound Nature 393 22 23 9590685
Ginzburg LR Bingham PM Yoo S 1984 On the theory of speciation induced by transposable elements Genetics 107 331 341 6329903
Evgen'ev MB Zelentsova H Poluectova H Lyozin GT Veleikodvorskaja V 2000 Mobile elements and chromosomal evolution in the virilis group of Drosophila
Proc Natl Acad Sci U S A 97 11337 11342 11016976
Watanabe H Fujiyama A Hattori M Taylor TD Toyoda A 2004 DNA sequence and comparative analysis of chimpanzee chromosome 22 Nature 429 382 388 15164055
Miyamoto MM Slightom JL Goodman M 1987 Phylogenetic relations of humans and African apes from DNA sequences in the psi eta-globin region Science 238 369 373 3116671
Kimura M 1980 Average time until fixation of a mutant allele in a finite population under continued mutation pressure—Studies by analytical, numerical, and pseudo-sampling methods Proc Natl Acad Sci U S A 77 522 526 16592764
|
16201008
|
PMC1239904
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Sep 30; 1(4):e44
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010044
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1620100910.1371/journal.pcbi.001004705-PLCB-RA-0137R2plcb-01-04-09Research ArticleBioinformatics - Computational BiologyEvolutionMolecular Biology - Structural BiologyEubacteriaArchaeaEntropic Stabilization of Proteins and Its Proteomic Consequences Hyperthermostable ProteomesBerezovsky Igor N 1Chen William W 12Choi Paul J 1Shakhnovich Eugene I 1*1 Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts, United States of America
2 Department of Biophysics, Harvard University, Cambridge, Massachusetts, United States of America
Skolnick Jeffery EditorBuffalo Center of Excellence in Bioinformatics, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 9 2005 1 4 e4722 6 2005 1 9 2005 Copyright: © 2005 Berezovsky et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Evolutionary traces of thermophilic adaptation are manifest, on the whole-genome level, in compositional biases toward certain types of amino acids. However, it is sometimes difficult to discern their causes without a clear understanding of underlying physical mechanisms of thermal stabilization of proteins. For example, it is well-known that hyperthermophiles feature a greater proportion of charged residues, but, surprisingly, the excess of positively charged residues is almost entirely due to lysines but not arginines in the majority of hyperthermophilic genomes. All-atom simulations show that lysines have a much greater number of accessible rotamers than arginines of similar degree of burial in folded states of proteins. This finding suggests that lysines would preferentially entropically stabilize the native state. Indeed, we show in computational experiments that arginine-to-lysine amino acid substitutions result in noticeable stabilization of proteins. We then hypothesize that if evolution uses this physical mechanism as a complement to electrostatic stabilization in its strategies of thermophilic adaptation, then hyperthermostable organisms would have much greater content of lysines in their proteomes than comparably sized and similarly charged arginines. Consistent with that, high-throughput comparative analysis of complete proteomes shows extremely strong bias toward arginine-to-lysine replacement in hyperthermophilic organisms and overall much greater content of lysines than arginines in hyperthermophiles. This finding cannot be explained by genomic GC compositional biases or by the universal trend of amino acid gain and loss in protein evolution. We discovered here a novel entropic mechanism of protein thermostability due to residual dynamics of rotamer isomerization in native state and demonstrated its immediate proteomic implications. Our study provides an example of how analysis of a fundamental physical mechanism of thermostability helps to resolve a puzzle in comparative genomics as to why amino acid compositions of hyperthermophilic proteomes are significantly biased toward lysines but not similarly charged arginines.
Synopsis
Comparative genomics sends us profound signals that are not easy to understand. For example, it is well known that proteins from hyperthermophiles are enriched with charged residues, but it has been a mystery why enrichment in positively charged amino acids is almost entirely due to lysines at the expense of very similar arginines. Here, the authors show that lysines (in contrast to arginines) exhibit significant residual dynamics in folded states of proteins, making the entropic cost to fold lysine-rich proteins less unfavorable compared with arginine-rich ones. Therefore, replacements of arginines by lysines provide additional thermal stabilization of proteins via entropic mechanism, making them positively charged residues of choice for evolutionary optimization of hyperthermostable proteins. Apparently, natural selection uses diverse physical mechanisms of thermal stability to achieve adaptation. This study provides an example of how better understanding of protein physics can help in solving genomic mysteries.
Citation:Berezovsky IN, Chen WW, Choi PJ, Shakhnovich EI (2005) Entropic stabilization of proteins and its proteomic consequences. PLoS Comput Biol 1(4): e47.
==== Body
Introduction
Enhancing the stability of globular proteins remains an important task of protein engineering and design [1,2]. The major mechanisms for increasing stability discovered so far vary from introduction of additional chemical bonds (e.g., disulfide bridges) or ion pairs [3–6] to increasing either the enthalpic free energy contributions by the optimizing of hydrophobic core interactions [7–11] or the entropic contributions by varying main-chain degrees of freedom in the unfolded state [12]. This repertoire of mechanisms relies on a variety of underlying physical principles for increasing protein stability [13]. The diversity of extreme environments and the long evolutionary history of organismal proteomes of extremophiles [14,15] suggest, in turn, many possible mechanisms of protein stabilization in response to the demands of the environment. Furthermore, the fact that each proteome contains a variety of structures and functions suggests that nature used all, even seemingly negligible, opportunities and their combinations for structure stabilization when adapting to extreme environmental conditions [15]. Here, we show how side-chain entropy in the native state can provide a mechanism of thermostabilization that is complementary to one of the major mechanisms of thermophilic adaptation, electrostatic interactions [3,4]. The analysis of statistics of rotameric states, together with computational mutation experiments, followed by high-throughput analysis of complete proteomes, reveals a previously unknown mechanism of stabilization via replacement of arginine residues with lysines. This substitution stabilizes the folded state, yet it preserves the charged nature of the substitution position, which may be important for other, perhaps functional, reasons. Thus, possible evolutionary advantages of this mechanism are as follows: (i) avoidance of sterically unfavorable contacts upon substitution, (ii) conservation of the similar-to-the-original (in terms of geometry and size) side-chains, and (iii) preservation of the positive charge and, as a consequence, important electrostatic interactions in the globule [3,4]. These subtle advantages exemplify the elegant work of natural selection and hint at the existence of other, yet undiscovered, mechanisms of protein adaptation.
Results
Monte Carlo Unfolding Simulations of Hydrolases H from Escherichia coli and Thermus thermophilus
The Gō model of protein folding is an idealized model in which the favorable interaction contact terms are exactly those found in the native structure [16]. In this model, the physico-chemical details of protein interactions are replaced by a generic contact energy term that is the same for all contacts between atoms that are found in contact in the native structure, though the complexity of the folded backbone and the side-chain conformations are preserved. It has been argued that such a model is a good representation of such aspects of protein energetics and folding, where non-native contacts do not play a massive role [16,17]. It remains unclear whether such an idealized model can quantitatively predict absolute folding transition temperatures. However, our results suggest that the Gō model predicts the transition temperature accurately enough to discriminate between proteins of thermophilic and mesophilic origin.
Figure 1 shows the unfolding curves for the pair of meso/thermophilic hydrolase H from Escherichia coli and Thermus thermophilus. The Gō model correctly predicts a slightly higher transition temperature for the protein from thermophilic T. thermophilus compared with the one from E. coli (Figure 1). Remarkably, the two unfolding curves coincide up to the transition, at which point they separate and then recombine at higher temperatures (Figure 1). Because the native states are enthalpically identical, and the folds are essentially the same, we surmise the origin of the transition temperature difference to be purely entropic. Specifically, given the nature of the Gō model, the entropic differences must arise from the different number of accessible rotamer states in different proteins. Calculation of average number of rotamers per residue in fully unfolded state [18] gives values of 12.0 and 11.4 for the mesophilic and the thermophilic proteins, respectively. These numbers thus demonstrate that the higher side-chain entropy in the unfolded state of mesophilic hydrolase is partially responsible for the fact that it unfolds at a lower temperature than the thermophilic structure.
Figure 1 The Temperature Dependence of the Energy of Unfolding for Hydrolases, from E. coli (black rhombuses) and T. thermophilus (red squares)
Every simulation of unfolding started from the native structure and lasted for 2 × 106 MC steps; absolute temperature increment is 0.2 and 0.1 in the vicinity of transition temperature. The error bars represent mean square fluctuations of energy at each temperature calculated within productive part of a run when trajectory reached equilibrium after temperature increment.
Lysine and Arginine: Archetypal Signal of Rotamer Entropy in Protein Stability
A careful look at the number of accessible states for each residue type in the folded state of hydrolases (Table 1) leads us to another interesting observation: although arginine and lysine are chemically similar and have the same maximal number of possible rotameric states, 81, they differ greatly in their rotameric accessibilities in the folded state.
Table 1 Average Number of Rotamers per Residue Type in the Folded State (Absolute Temperature T = 1) in Hydrolases H from E. coli and T. thermophilus
There is a total of five groups of amino acid residues with the same maximal number of rotamers in unfolded state (Table 1): (i) Arg, Lys (maximal number of rotamers is 81); (ii) Glu, Met (27); (iii) Ile, Leu, His, Trp (9); (iv) Asp, Phe, Tyr (6); (v) Cys, Ser, Thr, Val (3). The amino acids lysine and arginine are both positively charged; both contain at least five heavy atoms in their side-chains. Both amino acids have four degrees of freedom in their rotatable bonds. The guanidinium group at the end of arginine displays resonance, and, as a consequence, has no internal rotational freedom. The salient difference between arginine and lysine is the fact that lysine is less bulky. Therefore, in the folded state, lysine may have slightly more freedom. Estimates of solvent accessibility of arginine and lysine do not reveal a substantial difference (unpublished data). As a control comparison, we use the pair isoleucine/leucine (each residue has a maximum of nine rotameric states and is similar to the other's physical and chemical properties). Using the Gō model for protein energetics, we sample the number of accessible rotamers as a function of temperature for the hydrolases from E. coli and T. thermophilus (Figure 2). We approximate the entropy of the side-chain with the natural logarithm of the number of observed states in long equilibrium Monte-Carlo simulation. Figure 2 shows the temperature dependence of the natural logarithm of the number of rotamers for pairs Arg/Lys (Figure 2A and 2B), Leu/Ile (Figure 2C and 2D), Thr/Ser (Figure 2E and 2F), and Thr/Val (Figure 2G and 2H) in hydrolases H from E. coli and T. thermophilis. According to Table 1, lysine and arginine residues have different residual side-chain entropy. Lysine residues have many more rotamers in the folded state than arginine residues: on average, 20.1 and 17.2 versus 3.5 and 5.6 rotamers per residue of a particular type (Lys or Arg) in 1INO and 2PRD, respectively. The control group in this analysis is the pair Leu/Ile, which shows a highly similar temperature dependence of the number of rotamers (Figure 2C and 2D) for both proteins. Two last pairs, Thr/Ser and Thr/Val, confirm the role of the side-chain size and, as a consequence, its flexibility in providing number of accessible rotameric states. Each of Thr, Ser, and Val has a maximum of three possible rotamers and, thus, can be compared. Although both Thr and Ser are hydrophilic residues, Ser residues have a slightly greater number of rotameric states in the folded structure (at absolute temperature 1 in our temperature units used for MC simulations) as a result of its smaller side-chain. The hydrophilic/hydrophobic pair Thr/Val (Figure 2G and 2H) exhibit very similar behavior, stemming from the similarity of their side-chain geometries. This result is further substantiated by the temperature-dependence data for the pairs Val/Ser and Phe/Tyr. (The results for averaged temperature dependence, for residue types from both hydrolases, are presented in Figure S1). The bulky side-chains of both Phe and Tyr (Figure S1F) show practically the same temperature dependence, whereas in the pair Val/Ser (Figure S1E) the latter has slightly more rotamers in the folded state.
Figure 2 The Temperature Dependence of the Natural Logarithm of Number of Rotamers
(A) Arg (black rhombuses) versus lysine (red squares) rotamers of hydrolase H from E. coli; (B) Arg (black rhombuses) versus lysine (red squares) rotamers of hydrolase H from T. thermophilus; (C) Leu (dark blue rhombuses) versus Ile (light blue squares) rotamers of hydrolase H from E. coli; (C) Leu (dark blue rhombuses) versus Ile (light blue squares) rotamers of hydrolase H from T. thermophilus; (E) Thr (orange rhombuses) versus Ser (yellow squares) rotamers of hydrolase H from E. coli; (F) Thr (orange rhombuses) versus Ser (yellow squares) rotamers of hydrolase H from T. thermophilus; (G) Thr (orange rhombuses) versus Val (green-blue squares) rotamers of hydrolase H from E. coli; (H) Thr (orange rhombuses) versus Val (green-blue squares) rotamers of hydrolase H from T. thermophilus.
These results suggests that lysine and arginine provide an excellent platform to test a possible entropy-based mechanism of protein stabilization for both genomic and computational studies, for the following reasons: (1) they have similar physico-chemical properties, (2) they maintain the same physical and chemical features, and (3) they have similar rotamer entropies in the unfolded state but different rotamer entropies in the folded states (Table 1). Accordingly, we study the effects of side-chain entropy on protein stability for the chosen pair of types of amino acid residues, arginine and lysine [12,19–23].
Statistics of Rotameric States in a Representative Set of Protein Structures
Let us consider a situation where residues with similar physical and chemical properties have a different number of rotamers in the folded state. The similarity of physical and chemical properties makes it possible to adjust stability due to entropic factor by mutating one residue type into another without changing the structure significantly. The first step to verify this mechanism is a statistical study of the difference in number of accessible rotamers for the folded and unfolded states. We analyzed the ratio of the number of rotamers (in natural logarithm units) at absolute temperature T = 4 (completely unfolded state) to that at T = 1 (folded state) for a representative set of 18 protein structures. Our results do not change from protein to protein and, as such, do not depend on the possible biases in the crystallographic quality of individual structures. Since we perform long runs of MC simulations (total of 107 steps), which equilibrate our system and sample distinct rotamer states, we eliminate the memory of rotamer states in the original experimental structures. The difference between the number of Lys and Arg rotamers is also consistent for the representatives of different protein families, namely hydrolases, rubredoxins, ferredoxins, and chemotaxis protein (Table S1). Figure 3 shows histograms of ratios for the following pairs of amino acid residues: Arg/Lys (Figure 3A), Val/Thr (Figure 3B), and Phe/Tyr (Figure 3C).
Figure 3 Distribution of the Ratios of the Number of Rotamers in Unfolded and Folded States in a Representative Set of Proteins
Completely unfolded state is achieved at absolute temperature T = 4, folded state at T = 1. (A) Lys versus Arg; (B) Ile versus Leu; (C) Phe versus Tyr; (D) Val versus Thr. Upper histogram in each panel corresponds to T = 4, lower histogram corresponds to T = 1.
Arginine and lysine show significantly different rotamer number ratios in the folded state distribution (Figure 3A; mean values of the distribution for lysine and arginine are 2.14 and 1.21 in natural logarithm units, respectively). Ratios for the pairs Leu/Ile, Val/Thr, and Phe/Tyr are very similar (Figure 3B–3D), with mean values of distributions 1.7/1.4, 0.85/0.97, and 1.62/1.58, respectively. These data corroborate that lysine residues contribute entropically to the change in equilibrium between the unfolded and folded states, whereas residues in pairs Leu/Ile, Val/Thr, and Phe/Tyr have similar number of rotamers in the folded state. As a next step, we prove the stabilizing role of lysine versus arginine in a direct computer simulation experiment.
R/K Replacement Computational Experiment: Detecting Changes in Stability by Monte Carlo Unfolding Simulations
As stated above, both a statistical analysis of rotameric states and a comparative high-throughput analysis of complete proteomes demonstrate the particular role of lysine rotamers in protein thermostability. To demonstrate the stabilizing role of lysine residues, we make a replacement of type Arg/Lys and analyze the unfolding simulations in order to detect an anticipated increase in structure stability.
We replaced arginine residues with lysine residues in corresponding positions and locally minimized the resulting structures. We left the rest of the structure intact (see Materials and Methods) in order to influence the native structure as little as possible. The same local minimization was applied to the native structure.
Results of the replacements of arginine residues with lysine residues in both hydrolases H are as follows: in hydrolase H from E. coli position 43 (20 Arg-residue rotamers/31 Lys-residue rotamers in the folded state), 86 (2/5), 88 (4/12), and 171 (3/18); in hydrolase H from T. thermophilus position 24 (1/5), 43 (4/12), 114 (15/47), 156 (14/24), 158 (5/12), 166 (21/25), and 171 (3/10). We also analyzed combinations of Arg-to-Lys replacements in different positions in the structure. We found an increase of transition temperature in the replacement R171K and in combination of all R/K substitutions in positions 43, 86, 88, and 171 in mesophilic hydrolase from E. coli.
Figure 4A and 4B shows a plot of the temperature dependence of the energy in unfolding simulations of structures with replacement R171K in hydrolases H from E. coli and T. thermophilus [24–26]. Though there is a slight increase in the enthalpic term in the modified (R171K) structure of thermophilic hydrolase (3,321 native contacts in the modified structure versus 3,246 in the original, according to the Gō model), and an increase in the number of rotamers in the modified structure, there is no indication of a change in the transition temperature in unfolding simulations (Figure 4A). Similar replacements in the structure of mesophilic hydrolase H from E. coli, on the other hand, cause a change in the transition temperature of approximately 0.1 in absolute units (2.6%). The increase in the number of native contacts in the modified structure (3,226 in modified versus 3,131 in original) accounts for 3% of the difference in transition temperature; entropic factors do not play a stabilizing role in this case (5.38 and 5.34 rotameric states per residue in original unmutated structures, respectively). We detected an increase in the stability of the structure when all arginine residues (positions 43, 86, 88, and 171) were replaced by lysine residues. Taking into account both the decrease in the enthalpic term in the modified structure (3,083 native contacts versus 3,102 in the original, or approximately 0.6% loss) and the simultaneous increase in the transition temperature by 0.05 of absolute units (gain of 1.3%) gives a total increase of 2% in stability, which we conclude to be an effect of entropy stabilization of the structure. The number of rotamers per residue increases from 5.41 in the original to 5.62 in the mutated structure, a 4% difference, which, taking into account the roughness of the estimate, corroborates an increase in stability. The absence of a stabilizing effect of replacements in thermophilic hydrolase H from T. thermophilus can be explained by the high stability of the original protein, as demonstrated earlier [24–26].
Figure 4 The Temperature Dependence of the Energy of Unfolding for Mutated (Red Squares) versus Original Hydrolases H
(A) R171K mutant and wild-type of hydrolase H from T. thermophilus; (B) R171K mutant and wild-type of hydrolase H from E. coli; (C) R43,86,88,171K mutant and wild-type of hydrolase H from E. coli.
We performed a similar experiment with a smaller protein to improve sampling. Cytochrome C from Rhodobacter sphaeroides [27] contains 112 amino acid residues, with positions 24, 26, 53, 58, 74, 80, 87, and 95 occupied by arginine residues. Simulations reveal the following variations in the number of rotamers in each position upon replacement with Lys: position 24 (8/16), 26 (15/17), 53 (37/38), 58 (5/22), 74 (2/6), 80 (43/43), and 95 (8/52). Replacement in individual positions did not reveal an increase in stability. However, simultaneous substitution of all arginine residues by lysines led to a noticeable increase in transition temperature, while the enthalpic term decreased by 0.5% (1,607 native contacts in the modified structure, compared with 1,615 in the original). Figure 5 shows the temperature dependence of the energies, averaged over five runs (each 5 × 107 MC steps). The difference between the transition temperature of the original and the modified structures is ΔT = 0.07 absolute units (3%) increase, which translates into a 3.5% increase in stability when the unfavorable change in enthalpy is taken into account. Indeed, the mutated structure demonstrates an increase in the entropy of the folded state, 5.03 versus 4.56 rotameric states per residue in the original structure.
Figure 5 The Temperature Dependence of the Energy of Unfolding for R24,26,53,58,74,80,87,95K Mutant Compared with the Original Structure of Cytochrome C from R. sphaeroides
These data show that lysine residues contribute greatly to the stabilization of folded states of proteins, compared with their peer positively charged arginine, whereas residues in pairs Leu/Ile, Val/Thr, and Phe/Tyr have similar number of rotamers in the folded state. It is possible that this mechanism of stabilization is employed by nature in its strategies of thermophilic adaptation. If this is the case, it should be manifest in comparative genomics analysis in greater content of lysines in hyperthermophiles compared with mesophiles and, importantly in bias toward Arg-to-Lys substitutions from mesophiles to hyperthermophiles.
Analysis of Complete Proteomes
Amino acid composition biases in hyperthermophilic proteomes.
We performed quantitative analysis on 38 mesophilic and 12 hyperthermophilic proteomes. (For a list of the genomes used, see Tables S2 and S3.)
It has been demonstrated earlier that mesophilic proteins posses rather limited stability [28]. In the case of (hyper)thermophilic proteins, stabilization should be much stronger and, thus, it requires concerted contribution from many possible mechanisms. For this reason, we intentionally considered only hyperthermophilic proteomes in order to capture the most pronounced sequence biases associated with the extreme thermal stability of hyperthermophilic species. Figure 6 and Figure S2 show sets of composition histograms for two types of residues charged and hydrophilic, respectively, presumably associated with variations in thermal stability. While in thermophilic species the percentage of polar residues is high [29], this percentage is the same or even smaller in hyperthermophilic organisms (for instance, Glu, Ser, Thr; see Figure S2). In the case of charged residues, we observe clear under-representation of Asp and His and an increase of Glu (Figure 6) in hyperthermophilic organisms. Increase of the Glu content is usually explained by its longer side-chain, which provides more opportunities for ion interaction [30,31]. It should be noted that increase of Glu at the expense of Asp can be a consequence of higher entropic contribution from Glu compared with Asp. However, Glu also has a longer side-chain, which may naturally increase its enthalpic contribution. Thus, the role of the toward Glu deserves separate consideration with careful analysis of both enthalpic and entropic effects. In addition to the earlier-detected increase of total content in the Arg/Lys pair [30], we found that in ten of the 12 hyperthermophilic genomes lysine content is much higher (not less than 6%), whereas arginine content is distributed evenly mostly between 3% and 6% (Figure 6A and 6B). The dominance of arginine in the pair Arg/Lys in proteomes of Methanopyrus kandleri and Aeropyrum pernix is an exception due to the high GC content in these genomes [32,33]. Mean values for the percentage of Arg, Lys, His, Asp, and Glu in mesophilic and hyperthermophilic organisms (excluding M. kandleri and A. pernix), along with p-values according to binomial distribution calculated for the pair of archetypal representatives of each group, E. coli and Pyrococcus furiosus (Table S4) are: Arg, mesophilic/hyperthermophilic genomes, 5.46/4.94 (p = 8 × 10−11); Lys, 5.35/8.48 (p < 10−14); Asp, 5.28/4.72 (p < 10−14); Glu, 6.1/8.42 (p < 10−14). Thus, ten of the 12 hyperthermophilic organisms show difference in the preference for charged residues in mesophilic and hyperthermophilic genomes. (Note that for Arg and Asp, the difference is inverse: there are more such groups in mesophiles than in thermophiles.) In particular, we detected an increase of lysine content at the expense of arginine content.
Figure 6 Histograms of the Content of Charged Amino Acid Residues in Hyperthermophilic Genomes Compared with Mesophilic Genomes
Top histogram shows percentage of each residue in mesophilic genomes; bottom histogram, in hyperthermophilic genomes. A total of 12 hyperthermophilic and 38 mesophilic genomes were analyzed (for the complete list, see Tables S1 and S2). (A) Arg; (B) Lys; (C) Asp; (D) Glu.
Comparative analysis of hyperthermophilic versus mesophilic proteomes.
A persistently high percentage of Arg+Lys, though biased in most of the proteomes toward increased lysine content, along with the similarity in physical and chemical features of these residues suggests an examination of substitutions of types R/K versus K/R in the alignment of mesophilic sequences (here, E. coli) versus hyperthermophilic ones. We started from the following hypothesis: if, as stated elsewhere [30], only the total content of arginine plus lysine residues matters in determining the stability of hyperthermostable proteins, then there should be no preference for one of the residues (Lys) over the other one (Arg). We used sequences of five hyperthermophilic archaea (Aeropyrum pernix, Methanococcus jannaschii, Nanoarchaeum equitans, P. furiosis, and Sulfolobus tokodaii) and one hyperthermophilic bacteria (Aquifex aeolicus). Outputs of BLAST alignments were used for comparison of sequence substitutions that favor one or the other residue in each pair (Table 2). Our data are presented in Table 2. The number before the slash is the percentage of amino acid residues in the mesophilic sequence, e.g., Leu that was replaced by the other amino acid in the hyperthermophilic sequence, e.g., Ile. The number after the slash reflects the same data for the opposite replacement, e.g., Ile, in the mesophilic sequence by Leu in the hyperthermophilic sequence. The control groups here are the pairs Leu/Ile and Ser/Thr; both residues in each pair are hydrophobic or polar, and both have the same maximal number of possible rotamers, nine and three, respectively. In all alignments of E. coli sequences against those from one of the hyperthermophilic genomes, we obtained equal or very similar numbers of residues substitutions (numbers in parenthesis show ratio of forward to back substitutions). The exceptions are pairs LI/IL and RK/KR in A. pernix, which show bias in the opposite direction explained by the GC content. Unlike the above control groups, the pairs RK/KR demonstrate a remarkable bias toward replacement of arginine in the mesophilic sequence with lysine in the hyperthermophilic sequence (at least 1.6 times in P. furiosis, and up to almost four times in N. equitans). p-Values (calculated according to χ2 criteria) show a statistically significant preference for the arginine-to-lysine substitution as opposed to the reverse one. This challenges the idea that arginine and lysine play the same role in thermostability [30]. Therefore, our comparative genomics analysis strongly supports the conclusion that lysine has a particular or even exceptional role in protein stabilization [31].
Table 2 Percentage of the Forward/Back Replacements in Alignments of Hyperthermophilic Genomes against Mesophilic One (E. coli)
Recently, Jordan et al. [34] attempted to find “universal trend of amino acid gain and loss in protein evolution.” They found two major groups of amino acids: Cys, Met, His, Ser, and Phe; and Pro, Ala, Glu, and Gly, that are accrued and consistently lost. Their major conclusion is that, in agreement with earlier developed amino acid chronology [35], “all amino acids with declining frequencies are thought to be among the first incorporated into genetic code; conversely, all amino acids with increasing frequencies, except Ser, were probably recruited late.” Importantly, amino acid chronology proposed in [34] reflects early stages of protein evolution [35,36], and, technically, it was developed on the basis of codon chronology. The latter started from the GC-rich codons, as first codons are believed to be more thermostable than all repertoire of codons, and corresponding amino acids emerged as result of wobble and transition mutations. Thus, both amino acid chronology and universal trend of amino acid gain and loss in protein evolution demonstrate generic connection between DNA and proteins composition and its evolution. Our findings, however, are of a different nature. Contrary to “universal trend,” which does not discriminate between meso- and (hyper)thermophilic organisms, bias toward Lys residues is a statistically significant trend of hyperthermophilic proteomes (Figure 6). Thus, we discovered here a new mechanism of thermophilic adaptation that happens on the level of amino acid composition and originates from the specific physical chemical features of arginine/lysine residues. Finally, entropic mechanism of stabilization is complementary to generic amino acid chronology, and it demonstrates work of natural selection in order to reach adaptation to extreme environmental condition.
Discussion
Thermodynamical Models of Protein Stability and the Enthalpy/Entropy Relationship as a Manifestation of a Variety of Stabilizing Factors
Most of the data on structure thermostability and its major factors come from experiments aimed at analyzing the role of individual contributors, such as hydrophobic, van der Waals, electrostatic [3,4], and other physical forces [13,14]. This determines a common computational approach to the analysis of protein thermostability: a limited dynamic or static model with a detailed Hamiltonian that partitions the forces into distinct classes [21,37,38].
The approach we presented here straddles the way between a complete description of folding and the limited dynamic models presented in previous studies. We employ a Gō model [16,17], which enables us to account for the enthalpically relevant terms, albeit in a coarse-grained manner. The Gō model also permits us to account accurately for the various entropic contributors to the folding free energy, namely the backbone entropy and side-chain entropies. Finally, and most importantly, the simplicity of the model means that we are able to probe these various free energy effects with multiple folding runs relatively easily. In short, this approach makes it possible to examine the generic aspects of thermodynamics of thermostability.
Our results show the utility of Monte Carlo unfolding simulations with the Gō model as a way to detect the relative contributions of the free energy components, enthalpy and entropy. Furthermore, the description of the unfolding simulation in terms of the enthalpy/entropy relationship highlights the differences in the contribution of different types of amino acid residues to the entropic part of the free energy balance of a protein (Table 1). We found a difference in the number of accessible rotamers in folded state despite the fact that these residues were naively expected to be fully fixed in native states, i.e., all have only one rotamer available in the native state. Logarithm of the ratio of the number of rotamers in the folded and unfolded states gives us the entropy difference upon folding for each residue (Figures 2 and 3). These data demonstrate significantly higher entropy of lysine residues in folded states compared with those of arginine.
We demonstrate here that our top-down approach, from analysis of thermodynamic quantities to discovery of concrete physical processes that give rise to the observed thermodynamic phenomena, can not only detect differences in the free energies of stabilization, but also reveal novel mechanisms of stabilization via the rotamer entropic effect.
Genomic Motivation for the Novel Mechanism of Thermostability
To validate a model of protein stability on the genomic and proteomic level, it is important to find particular expected compositional and sequence biases by means of massive high-throughput analysis. Even if the bulk of the protein in the organism exhibits a particular mechanism of stabilization according to the mechanism of adaptation commonly developed in the proteome [39–41], one or a few proteins may rely on a different/additional mechanism developed under specific environmental conditions. What additional information can we glean from the proteome analysis? First, amino acid compositional analysis reveals a bias toward lysine residues in the pair Arg/Lys, typical for the genomes of hyperthermophiles. Such analysis also a bias toward Lys and Glu in hyperthermophilic proteomes, whereas Asp and His are unfavorable in these organisms. The only exceptions are two hyperthermophilic genomes, A. pernix and M. kandlerii, whose preference for arginine residues is a direct consequence high GC content [31,33,42]. Second, comparative analysis of hyperthermophilic and mesophilic (here, E. coli) proteomes reveals an enrichment of lysine content at the expense of the arginine.
Though bias in amino acid composition toward increasing charged residues is well documented in earlier works [29–31,43–46], the difference in the frequencies of arginine and lysine residues has not been explained unequivocally [30,47].
There is a strong belief that GC content is the major factor in ensuring survival and selective advantages for extremophiles, in particular thermophiles, due to high thermostability of GC pairs [46]. Assuming that this explanation is correct, one would expect (hyper)thermophiles to select arginine over lysine. Arginine is encoded by six codons, four of which (CGU, CGC, CGA, and CGG) are GC-rich, whereas lysine is encoded by two codons (AAA and AAG). Moreover, arginine has a higher charge, which means it forms better salt bridges [47]. Surprisingly, this expectation is confirmed in only a very few cases, for instance in A. pernix and M. kandlerii; whereas in the majority of other hyperthermophilic organisms, we observe significant increase in lysine content, which typically anticorrelates with GC content. Furthermore, as we demonstrated here, lysine content partially increases due to direct replacement of arginine residues (Table 2), which points out the obvious advantage that lysine residues have over arginine. One could argue that (i) composition effect alone can account for the higher substitution rate of Arg/Lys, or that (ii) there was a particular common ancestor enriched by Lys, and the specific compositional bias in contemporary proteins that we observe is due to phylogeny. But the unfolding simulations, the statistical data on rotameric states, and the genomic evidence all point to the advantage of lysine over arginine when thermostability is important. Furthermore, we see excess of lysine only in hyperthermophilic organisms, regardless of their loci on the phylogenetic tree (e.g., archaea and bacteria). Lysine still has some entropic freedom, even in the folded state of a protein, due to its smaller size. In comparison, arginine, with its bulky guanidinium group, does not have the same freedom, and its possible enthalpic advantage is compensated by the drawback of packing of two closely located charges [48].
Adaptation to High Temperatures as a Complex Effect of Different Types of Interactions
We discovered here a novel mechanism of structure thermostabilization that relies on side-chain rotamer entropy [19,23]. To single out the potential effects of rotamer entropy, we compared pairs of amino acid residues with similar physical and chemical properties and the same maximal number of rotameric states. The difference in the rotamer entropy of each pair of residues must, then, be a result only the difference in the rotameric entropy of their side-chains. Statistical data of accessible rotameric states (see Figures 2 and 3 and Table 1) show substantial entropy for lysine residues in both folded and unfolded states, whereas arginine has a significantly decreased side-chain freedom in the folded state. Preference for the lysine is also supported by the genomic data (see Figure 6 and Table 2) and illustrated by the computational mutation experiment (Figures 4 and 5).
In general, just a few mutations can make the difference between a mesophilic protein and its (hyper)thermophilic counterpart. Stability is reached by fine-tuning sequences and structures, rather than by drastic rearrangement. Moreover, in the case of hyperthermophilic proteins, practically all possible means of stabilization appear to be utilized. Any additional element of stabilization must both preserve the already-achieved level of stability and provide additional stabilization by invoking only minor modifications in sequence and structure. Arg/Lys replacement satisfies both of these conditions and, thus, exemplifies using the entropic contribution while simultaneously preserving the charged nature of the residues, which is important for other mechanisms of stabilization, such as the electrostatic [49,50]. Indeed, as it has been thoroughly demonstrated elsewhere [3,4], electrostatics is one of the major factors of thermostability. The entropic mechanism discovered in this work serves as an important complementary factor that provides additional stabilization when the repertoire of other mechanisms has already been possibly exhausted [5,6].
The novel mechanism of thermal stabilization reported here is unique in that it relies not only on the physical and chemical properties of a residue, but also on its dynamics in folded state affecting its entropic contribution. And because the effect is small, it can be revealed only in careful simulations and genomic comparisons. The important pedagogical point we draw from this result is that the study of protein stability on individual proteins using current state-of-the-art energy functions may result in missing subtle thermodynamic evolutionary signals that only become apparent in high-throughput analysis of proteomes and genomes.
Materials and Methods
Statistics of rotameric states.
The number of accessible rotamers in the folded (T = 1) and fully unfolded (T = 4, see Figure 1) states for a representative set of proteins was calculated. The temperature dependence for the number of accessible rotamers in hydrolases H from E. coli and T. thermophilus was calculated at absolute temperatures T = 1, 2, 3, 3.5, and 4. Structure coordinates were recorded at every 105 MC steps for a total of 107 steps. The number of rotamers for every residue were determined as an average over 100 snapshots.
We used the following PDB structures to collect statistics of rotameric states (see also Table S1): (1) hydrolases, (2) rubredoxins, (3) 2Fe-2S ferredoxin, (4) 4Fe-4S ferredoxin, and (5) chemotaxis protein.
Statistics of rotameric states in original and mutated structures of hydrolases and Cytochrome C were collected from recorded structures at every 104 MC steps for a total of 107 steps done for every original/mutated structure (1,000 snapshots). Our results do not depend on crystallographic quality of the structures, and we obtain consistent data for the following reasons: (i) we work with high-resolution structures; and (ii) most important, that we do long runs of MC simulations which equilibrate a system and, thus, eliminate any possible discrepancies in original structures.
High-throughput sequence analysis.
We used the BLAST program [51] to create a set of pair-wise alignments with significant e-value (e = 0.05) using the substitution matrix BLOSUM62. We chose only sequences that had gaps of length 3 or less, and full alignment length of 45 residues or more.
Molecular dynamic minimization.
We used the CHARMM program [52] to minimize the structure upon Arg/Lys replacement. CHARMM minimization was done using the following procedure. Hydrogen coordinates were calculated by bond geometry and inserted into the starting structure; SHAKE was turned on for updating hydrogen positions. A generalized-Born solvation energy function (GBorn) and a dielectric constant with linear distance-dependence were used for dynamics. The residue of interest, and all atoms within a 5-angstrom radius of any atom in that residue, were permitted to move with CHARMM degrees of freedom to ensure that the mutated residue could repack locally. The dynamics simulation was initially constrained to the native state using a harmonic potential. The artificial harmonic constraint was reduced to zero slowly over consecutive cycles of adopted-basis Newton-Raphson minimization. To detect the effects of mutation, we minimized both mutated and original structures with the same protocol in order to use the latter one as a control.
Unfolding Monte Carlo simulations of modified proteins.
Unfolding simulations were performed using an all-atom Gō model developed earlier [53]. In the Gō interaction scheme, atoms that are neighbors in the native structure are assumed to have attractive interactions. Hence, the Gō model of interactions is structure-based. Every unfolding run consists of 2 × 106 steps in the unfolding simulations of hydrolases (Figure 1) and their mutants (Figure 5A–C), and 5 × 107 steps in the case of Cytochrome C (Figure 6). The move set is one backbone move followed by one side-chain move [53].
Supporting Information
Figure S1 The Temperature Dependence of the Natural Logarithm of the Number of Rotamers Averaged over Respective Values in Hydrolases H from E. coli and T. thermophilus
(A) Arginine (black rhombuses) versus Lys (red squares) rotamers; (B) Leu (dark blue rhombuses) versus Ile (light blue squares); (C) Thr (orange rhombuses) versus Ser (yellow squares); (D) Thr (orange rhombuses) versus Val (green-blue squares); (E) Val (green-blue rhombuses) versus Ser (yellow squares); (F) Phe (green-blue rhombuses) versus Tyr (orange squares).
(14 KB PDF)
Click here for additional data file.
Figure S2 Histograms of the Content of Polar Amino Acid Residues in Hyperthermophilic Genomes Compared with Mesophilic Ones
Top histogram shows percentage of respective residue in mesophilic genomes; bottom histogram, in hyperthermophilic ones. Total of 12 hyperthermophilic and 38 mesophilic genomes were analyzed (for the complete list, see Tables S1 and S2). (A) Asn; (B) Gln; (C) His; (D) Ser; (E) Thr; (F) Tyr.
(766 KB TIF)
Click here for additional data file.
Table S1 Set of Proteins Used in Collecting Comparative Rotamer Statistics
(44 KB DOC)
Click here for additional data file.
Table S2 List of Mesophilic Genomes
Total of 38 genomes. Columns are as follows: first, genome accession number in NCBI database of complete genomic sequences; second, name of the organism; third, Life Kingdom (A, archaea; B, bacteria); fourth, size of the proteome in number of protein coding sequences.
(57 KB DOC)
Click here for additional data file.
Table S3 List of Hyperthermophilic Genomes
Total of 12 genomes. Columns are as in Table S2.
(33 KB DOC)
Click here for additional data file.
Table S4 Expected (on the Basis of the Occurrence in E. coli, Column 4) and Observed (Column 5) Frequencies of Charged Amino Acid Residues in P. furiosis
Diff in σ, difference in number of standard deviation between respective expected and observed values. The null model used to calculate p-values represents random uncorrelated distribution of charged amino acids over proteomes resulting in binomial distribution for the content of each type of amino acids, from which p-values were calculated.
(29 KB DOC)
Click here for additional data file.
Accession Numbers
The Protein Data Bank (http://www.rcsb.org/pdb/) accession numbers for products used in this paper are 2Fe-2S ferredoxin (4FXC, 1FRR, 1FRD, 1DOI, and 2CJN); 4Fe-4S ferredoxin (1FCA, 1DUR, 1IQZ, and 1VJW); chemotaxis protein (3CHY, 2CHF, and 1TMY); Cytochrome C (1DW0); hydrolases (1INO [from E. coli] and 2PRD [from T. thermophilus]); rubredoxins (1RDG, 5RXN, 8RXN, and 1CAA).
INB is supported by a Merck Postdoctoral Fellowship for Genome-Related Research. This work is supported by the National Institutes of Health.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. INB, PJC, and EIS performed the experiments. INB, WWC, and EIS analyzed the data. PJC contributed reagents/materials/analysis tools. INB and EIS wrote the paper.
A previous version of this article appeared as an Early Online Release on August 2, 2005 (DOI: 10.1371/journal.pcbi.0010047.eor).
==== Refs
References
Pokala N Handel TM 2001 Review: Protein design—Where we were, where we are, where we're going J Struct Biol 134 269 281 11551185
Malakauskas SM Mayo SL 1998 Design, structure and stability of a hyperthermophilic protein variant Nat Struct Biol 5 470 475 9628485
Dominy BN Perl D Schmid FX Brooks CL 3rd 2002 The effects of ionic strength on protein stability: The cold shock protein family J Mol Biol 319 541 554 12051927
Dominy BN Minoux H Brooks CL 3rd 2004 An electrostatic basis for the stability of thermophilic proteins Proteins 57 128 141 15326599
Sanchez-Ruiz JM Makhatadze GI 2001 To charge or not to charge? Trends Biotechnol 19 132 135 11250029
Loladze VV Ibarra-Molero B Sanchez-Ruiz JM Makhatadze GI 1999 Engineering a thermostable protein via optimization of charge-charge interactions on the protein surface Biochemistry 38 16419 16423 10600102
Chen J Lu Z Sakon J Stites WE 2000 Increasing the thermostability of staphylococcal nuclease: Implications for the origin of protein thermostability J Mol Biol 303 125 130 11023780
Chen J Stites WE 2001 Energetics of side chain packing in staphylococcal nuclease assessed by systematic double mutant cycles Biochemistry 40 14004 14011 11705392
Chen J Stites WE 2004 Replacement of staphylococcal nuclease hydrophobic core residues with those from thermophilic homologues indicates packing is improved in some thermostable proteins J Mol Biol 344 271 280 15504416
Holder JB Bennett AF Chen J Spencer DS Byrne MP 2001 Energetics of side chain packing in staphylococcal nuclease assessed by exchange of valines, isoleucines, and leucines Biochemistry 40 13998 14003 11705391
Korkegian A Black ME Baker D Stoddard BL 2005 Computational thermostabilization of an enzyme Science 308 857 860 15879217
Matthews BW Nicholson H Becktel WJ 1987 Enhanced protein thermostability from site-directed mutations that decrease the entropy of unfolding Proc Natl Acad Sci U S A 84 6663 6667 3477797
Jaenicke R 2000 Stability and stabilization of globular proteins in solution J Biotechnol 79 193 203 10867180
Jaenicke R 1991 Protein stability and molecular adaptation to extreme conditions Eur J Biochem 202 715 728 1765088
Jaenicke R Bohm G 1998 The stability of proteins in extreme environments Curr Opin Struct Biol 8 738 748 9914256
Go N Abe H 1981 Noninteracting local-structure model of folding and unfolding transition in globular proteins. I. Formulation Biopolymers 20 991 1011 7225531
Go N 1983 Theoretical studies of protein folding Annu Rev Biophys Bioeng 12 183 210 6347038
Canutescu AA Shelenkov AA Dunbrack RL Jr. 2003 A graph-theory algorithm for rapid protein side-chain prediction Protein Sci 12 2001 2014 12930999
Doig AJ Sternberg MJ 1995 Side-chain conformational entropy in protein folding Protein Sci 4 2247 2251 8563620
Kussell E Shimada J Shakhnovich EI 2003 Side-chain dynamics and protein folding Proteins 52 303 321 12833553
Privalov PL Makhatadze GI 1993 Contribution of hydration to protein folding thermodynamics. II. The entropy and Gibbs energy of hydration J Mol Biol 232 660 679 8393941
Pickett SD Sternberg MJ 1993 Empirical scale of side-chain conformational entropy in protein folding J Mol Biol 231 825 839 8515453
Sternberg MJ Chickos JS 1994 Protein side-chain conformational entropy derived from fusion data—Comparison with other empirical scales Protein Eng 7 149 155 8170918
Robic S Guzman-Casado M Sanchez-Ruiz JM Marqusee S 2003 Role of residual structure in the unfolded state of a thermophilic protein Proc Natl Acad Sci U S A 100 11345 11349 14504401
Guzman-Casado M Parody-Morreale A Robic S Marqusee S Sanchez-Ruiz JM 2003 Energetic evidence for formation of a pH-dependent hydrophobic cluster in the denatured state of Thermus thermophilus ribonuclease H J Mol Biol 329 731 743 12787674
Hollien J Marqusee S 2002 Comparison of the folding processes of T. thermophilus and E. coli ribonucleases H J Mol Biol 316 327 340 11851342
Leys D Backers K Meyer TE Hagen WR Cusanovich MA 2000 Crystal structures of an oxygen-binding cytochrome c from Rhodobacter sphaeroides
J Biol Chem 275 16050 16056 10821858
Taverna DM Goldstein RA 2002 Why are proteins marginally stable? Proteins 46 105 109 11746707
Kreil DP Ouzounis CA 2001 Identification of thermophilic species by the amino acid compositions deduced from their genomes Nucleic Acids Res 29 1608 1615 11266564
Cambillau C Claverie JM 2000 Structural and genomic correlates of hyperthermostability J Biol Chem 275 32383 32386 10940293
Bharanidharan D Bhargavi GR Uthanumallian K Gautham N 2004 Correlations between nucleotide frequencies and amino acid composition in 115 bacterial species Biochem Biophys Res Commun 315 1097 1103 14985126
Deckert G Warren PV Gaasterland T Young WG Lenox AL 1998 The complete genome of the hyperthermophilic bacterium Aquifex aeolicus
Nature 392 353 358 9537320
Slesarev AI Mezhevaya KV Makarova KS Polushin NN Shcherbinina OV 2002 The complete genome of hyperthermophile Methanopyrus kandleri AV19 and monophyly of archaeal methanogens Proc Natl Acad Sci U S A 99 4644 4649 11930014
Jordan IK Kondrashov FA Adzhubei IA Wolf YI Koonin EV 2005 A universal trend of amino acid gain and loss in protein evolution Nature 433 633 638 15660107
Trifonov EN 2004 The triplet code from first principles J Biomol Struct Dyn 22 1 11 15214800
Trifonov EN Berezovsky IN 2002 Molecular evolution from abiotic scratch FEBS Lett 527 1 4 12220623
Makhatadze GI Privalov PL 1993 Contribution of hydration to protein folding thermodynamics. I. The enthalpy of hydration J Mol Biol 232 639 659 8393940
Makhatadze GI Privalov PL 1995 Energetics of protein structure Adv Protein Chem 47 307 425 8561051
Knight CG Kassen R Hebestreit H Rainey PB 2004 Global analysis of predicted proteomes: Functional adaptation of physical properties Proc Natl Acad Sci U S A 101 8390 8395 15150418
Fukuchi S Yoshimune K Wakayama M Moriguchi M Nishikawa K 2003 Unique amino acid composition of proteins in halophilic bacteria J Mol Biol 327 347 357 12628242
Bastolla U Moya A Viguera E van Ham RC 2004 Genomic determinants of protein folding thermodynamics in prokaryotic organisms J Mol Biol 343 1451 1466 15491623
Kawarabayasi Y Hino Y Horikawa H Yamazaki S Haikawa Y 1999 Complete genome sequence of an aerobic hyper-thermophilic crenarchaeon, Aeropyrum pernix K1 DNA Res 6 83 101 145 152 10382966
Das R Gerstein M 2000 The stability of thermophilic proteins: A study based on comprehensive genome comparison Funct Integr Genomics 1 76 88 11793224
Pe'er I Felder CE Man O Silman I Sussman JL 2004 Proteomic signatures: Amino acid and oligopeptide compositions differentiate among phyla Proteins 54 20 40 14705021
Chakravarty S Varadarajan R 2002 Elucidation of factors responsible for enhanced thermal stability of proteins: A structural genomics based study Biochemistry 41 8152 8161 12069608
Tekaia F Yeramian E Dujon B 2002 Amino acid composition of genomes, lifestyles of organisms, and evolutionary trends: A global picture with correspondence analysis Gene 297 51 60 12384285
Vieille C Zeikus GJ 2001 Hyperthermophilic enzymes: Sources, uses, and molecular mechanisms for thermostability Microbiol Mol Biol Rev 65 1 43 11238984
Hendsch ZS Tidor B 1994 Do salt bridges stabilize proteins? A continuum electrostatic analysis Protein Sci 3 211 226 8003958
Elcock AH 1998 The stability of salt bridges at high temperatures: Implications for hyperthermophilic proteins J Mol Biol 284 489 502 9813132
Karshikoff A Ladenstein R 2001 Ion pairs and the thermotolerance of proteins from hyperthermophiles: a “traffic rule” for hot roads Trends Biochem Sci 26 550 556 11551792
Altschul SF Gish W Miller W Myers EW Lipman DJ 1990 Basic local alignment search tool J Mol Biol 215 403 410 2231712
Brooks BR Bruccoleri RE Olafson BD States DJ Swaminathan S 1983 CHARMM—A program for macromolecular energy, minimization, and dynamics calculations J Comput Chem 4 187 217
Shimada J Kussell EL Shakhnovich EI 2001 The folding thermodynamics and kinetics of crambin using an all-atom Monte Carlo simulation J Mol Biol 308 79 95 11302709
|
16201009
|
PMC1239905
|
CC BY
|
2021-01-05 09:18:23
|
no
|
PLoS Comput Biol. 2005 Sep 30; 1(4):e47
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010047
|
oa_comm
|
==== Front
Aust New Zealand Health PolicyAustralia and New Zealand Health Policy1743-8462BioMed Central London 1743-8462-2-211615015410.1186/1743-8462-2-21ResearchComparison of the uptake of health assessment items for Aboriginal and Torres Strait Islander people and other Australians: Implications for policy Kelaher Margaret [email protected] David [email protected] David [email protected] Ian [email protected] Program Evaluation Unit, School of Population Health, University of Melbourne, Australia2 Onemda VicHealth Koori Health Unit, School of Population Health, University of Melbourne, Australia3 Centre for Health in Society, School of Population Health, University of Melbourne, Australia2005 9 9 2005 2 21 21 24 6 2005 9 9 2005 Copyright © 2005 Kelaher et al; licensee BioMed Central Ltd.2005Kelaher et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Health Assessment (HA) items were introduced in 1999 for Aboriginal and Torres Strait Islander people aged at least 55 years and all Australians aged over 75 years. In 2004 a new item was introduced for HAs among adult Aboriginal and Torres Strait Islander people aged 15–54 years. The new item has been applauded as a major policy innovation however this enthusiasm has been tempered with concern about potential barriers to its uptake. In this study we aim to determine whether there are disparities in uptake of HA items for Aboriginal and Torres Strait Islander people compared to other Australians.
Method
The analysis was based on Health Insurance Commission data. Indigenous status was ascertained based on the item number used. Logistic regression was used to compare uptake of HA items for older people among Aboriginal and Torres Strait Islander people compared to other Australians. Adjustments were made for dual eligibility. Uptake of the HA items for older people was compared to the uptake of the new item for Aboriginal and Torres Strait Islander people aged 15–44 years.
Results
Our analyses suggest a significant and persistent disparity in the uptake of items for older patients among Aboriginal and Torres Strait Islander people compared to other Australians. A similar disparity appears to exist in the uptake of the new adult Aboriginal and Torres Strait Islander HA item.
Conclusion
Further engagement of primary care providers and the community around the uptake of the new HA items may be required to ensure that the anticipated health benefits eventuate.
==== Body
The introduction of Medical Benefits Scheme (MBS) item numbers to reimburse health assessments (HAs) represented a major shift in support for access to health promotion and preventive care in primary care settings. The HA items provide reimbursement for doctors to evaluate patient's physical, psychological and social function in order to optimise health care and education. HA items were first introduced for older patients in 1999. [1] The items included HAs conducted at consulting rooms and not at consulting rooms, hospitals or residential aged care facilities (referred to hereafter as non-consulting room items). [1] Aboriginal and Torres Strait Islander people aged at least 55 years and all Australians aged over 75 years are eligible for these items. The item numbers for Aboriginal and Torres Strait Islander people and all Australians are shown in table 1.
Table 1 MBS Health Assessment item numbers
Health Assessment All Australians 75+ yrs Aboriginal and Torres Strait Islander people 55+yrs
At consulting room 700 704
Not at consulting rooms, hospitals or residential aged care facilities 702 706
The uptake of the HA items and other items introduced as part of the Enhanced Primary Care (EPC) program from 1999–2001 has been rigorously evaluated. HA items had the highest uptake of the Enhanced Primary Care items with around 18% of the eligible population using them. [2,3] No information was available on baseline levels for the provision of HAs but the evaluation did suggest that there was an increase in the use of HAs in case study practices and that reimbursement was an incentive to completing HAs in about one third of practices. Health benefits associated with HA among older patients were relatively small [4,5] and the evaluation suggested that further uptake was required to have significant impact on the health of the target populations. [3] This was particularly true of the items for Aboriginal and Torres Strait Islander people which were used at a significantly lower rate than the items for the general population. [6] It was suggested that this effect may have occurred either because Aboriginal and Torres Strait Islander people might be more likely to have pre-existing care plans or because Aboriginal and Torres Strait Islander people were more likely to use services (e.g. hospitals) where Medicare was not used. [7] In either case it would be expected that the disparity should decrease over time as people required new health assessments and Medicare use among indigenous people increased. [8]
In May 2004, a new item (item 710) was introduced for HAs among adult Aboriginal and Torres Strait Islander people aged 15–54 years. [8,9] Adult HAs could have significant health benefits for indigenous people because of the early age of onset of chronic disease and higher rates of infectious disease in this community compared to other Australians. [10] For example, the rate of sexually transmitted infection was halved at two year follow-up in indigenous rural and remote communities in Queensland where Well Persons Health Checks were conducted. [11] If the new item results in increased HAs, it has the potential to greatly reduce the burden of disease among indigenous Australians; it has rightly been applauded as an example of innovative policy in indigenous health. [9] However this enthusiasm has been tempered with concerns that the potential health benefits of the new item will not be realised because of low uptake. [9]
In this study we aim to establish whether there are likely to be barriers to the uptake of the new HA item by comparing the uptake of the HA items for older people among Aboriginal and Torres Strait Islander people and the rest of the community. We also examine differences in uptake over time and differences between States and Territories. Finally we compare uptake of the HA items for older people to the uptake of the new items for Aboriginal and Torres Strait Islander people aged 15–44 years in the first three quarters after their introduction. It would be expected that structural barriers to the introduction of HAs should have decreased over since 1999 because of the introduction of the HA items. Accordingly it might be expected that the uptake of the new item might be more rapid than the uptake of the items for older Australians.
Data
Data on the use of item numbers (700, 702, 704, 706) by year and by State and Territory were obtained from the Health Insurance Commission statistical reports. [12] Data on the HA items was available from the last quarter in 1999 but this was not used in the general comparison because a full years data was not available. The extract included annual data from 2000–2004.
Data on the use of item numbers (700, 704, 710) in the first three quarters of their introduction was also obtained from the Health Insurance Commission statistical reports. [12] These data are available by State and Territory but figures for the whole of Australia were used because of low numbers. For items 702 and 704 the first three quarters data was for the last quarter of 1999 and the first two quarters of 2000. For item 710 the data was from the last three quarters of 2004. It should be noted that the first quarter data may not include data for the whole quarter.
In addition to the other eligibility requirements, only one claim could be made per person in a 12 month period. Accordingly quarterly and annual data reports should only contain one observation per person. Data are available for smaller geographic areas than State and Territory, such as general practice divisions, however low numbers and a high level of suppressed data made small area analysis problematic.
Population estimates for the Aboriginal and Torres Strait Islander population aged at least 55 years and aged 15–44 years by State and Territory were obtained for the Australian Bureau of Statistics (ABS) projections from the 2001 census for the years 2001 to 2004. [13] Population projections for the years 1999 and 2000 were obtained from series developed from the 1996 census. [14] The projections provide a low and high series of population estimates. In this study the series used had little impact on the results. The low series is reported because it yields the most conservative estimates of the difference between Aboriginal and Torres Strait Islander people and the rest of the community. Population estimates for the general population aged at least 75 years were obtained using ABS time series data. [15]
Analysis
A logistic regression was conducted to analyse differences in the uptake of consulting room (700, 704) and non-consulting room (702, 706) HA items according to Indigenous status and year taking into account variation due to State and Territory. Consulting room and non-consulting room items were analysed separately because there is geographic variation in their use which may be potential source of confounding. The dependent variable was coded dichotomously using service use data to estimate the number of people who used the service and population data to estimate the number of people who did not. Year was coded to enable linear trends in uptake to be tested. Indigenous status was coded dichotomously based on whether the items were only available to Aboriginal and Torres Strait Islander people or available to all Australians.
The 12.2% of Aboriginal and Torres Strait Islander people aged at least 75 years would be eligible for the general population items as well as the Aboriginal and Torres Strait Islander specific items. All analyses were conducted twice to explore whether dual eligibility could have an impact on the results. The first set of analyses was based on observed service use. Service use among Aboriginal and Torres Strait Islander people would be underestimated in these analyses if people with dual eligibility were using general population items. The data were also analysed assuming that Aboriginal and Torres Strait Islander people aged at least 75 years accessed HAs through general population items at the same rate as the rest of community. These instances of service use were then attributed to Aboriginal and Torres Strait Islander people rather than to other Australians. Service use among Aboriginal and Torres Strait Islander people would be overestimated in these analyses because some of the people using the Aboriginal and Torres Strait Islander items are likely to be aged at least 75 years and therefore would be counted twice. Some overestimation would also be expected to occur because the observed rate of service use among Aboriginal and Torres Strait Islander people aged over 75 years is likely to be less than that for the general population.
Differences in rates of consulting room and non-consulting room service use for Aboriginal and Torres Strait Islander people and the rest of the community were calculated for each State and Territory.
A logistic regression was conducted to compare the uptake of older all Australian (700), older Aboriginal and Torres Strait Islander (704) and adult Aboriginal and Torres Strait Islander people (710) HA items. The HA item for adult Aboriginal and Torres Strait Islander people (710) was used as the reference category for comparisons. Quarter was coded to enable linear and quadratic trends in uptake to be tested. The dependent variable was coded dichotomously using service use data to estimate the number of people who used the service and population data to estimate the number of people who did not.
Results
Comparison of the uptake of HA items for older people among Aboriginal and Torres Strait Islander people and other Australians
The result of the logistic regression for use of consulting room HA items (see table 2) suggested that Aboriginal and Torres Strait Islander people (3.0%) were significantly less likely to have HAs than the rest of the community (7.4%). There was a significant linear increase in use of the HA items, with use increasing from 5.1% in 2000 to 8.4% in 2004. There was also a significant interaction between Indigenous status and year with use of the HA items increasing slightly more rapidly for Aboriginal and Torres Strait Islander people than the rest of the community (see figure 1). Disparities remained in all years.
Figure 1 Trends in use of Consulting Room HA items by Indigenous status and Year.
Table 3 shows the per cent use of consulting room HA items by State and Territory and Indigenous status. Percentage uptake generally increased with the size of the eligible population with New South Wales (NSW), Queensland (QLD) and Victoria (VIC) having the highest rates and the Northern Territory (NT) having the lowest. In all States and Territories, with the exception of the NT, use was significantly lower among Aboriginal and Torres Strait Islander people. In the NT the pattern was reversed with Aboriginal and Torres Strait Islander people being more likely than the rest of the community to use the HA items. All differences remained significant if it was assumed that Aboriginal and Torres Strait Islander people aged at least 75 years used services at the same rate as the rest of the community.
Table 3 Per cent use of Consulting room HA items by Indigenous status and State/Territory
State 75+ yrs non-Indigenous 55+ yrs Indigenous Total Difference % (95% CI) Difference % (95% CI)-dual eligibility adjustment
NSW Count 156433 1316 157749
% 7.7% 2.7% 7.5% 5.0 (5.0–5.1) 4.2 (4.1–4.2)
VIC Count 111742 566 112308
% 7.5% 5.8% 7.5% 2.0 (1.9–2.1) 0.8 (0.7–0.9)
QLD Count 92970 1359 94329
% 9.4% 3.3% 9.1% 6.2 (6.1–6.3) 5.0 (4.9–5.0)
SA Count 25788 159 25947
% 4.8% 1.8% 4.7% 3.0 (2.9–3.1) 2.4 (2.3–2.5)
WA Count 27620 758 28378
% 5.7% 3.4% 5.6% 2.2 (2.4–2.5) 1.6 (1.5–1.7)
TAS Count 5529 15 5544
% 3.6% .3% 3.5% 3.4 (3.3–3.5) 3.0 (2.9–3.1)
ACT Count 2496 14 2510
% 4.0% 1.8% 4.0% 2.2 (2.1–2.4) 1.8 (1.6–2.0)
NT Count 197 486 683
% 1.9% 2.5% 2.1% -0.6 (-0.8–-0.3) -0.9 (-1.1–-0.6)
The result of the logistic regression for use of non-consulting room HA items (see table 4) suggested that Aboriginal and Torres Strait Islander people (1.3%) were significantly less likely to have HAs than the rest of the community (6.7%). There was a significant linear increase in use of the HA items, with use increasing from 3.4% in 2000 to 8.2% in 2004. There was also a significant interaction between Indigenous status and year with use of the HA items staying stable among Aboriginal and Torres Strait Islander people while increasing in the rest of the community (see figure 2).
Table 4 Logistic Regression for use of non-Consulting room HA items by Indigenous status and Year controlling for State/Territory
Variable Observed service use Dual eligibility adjustment
OR (95% CI) OR (95% CI)
Indigenous status 0.22 (0.21–0.23) 0.34 (0.33–0.35)
Linear trend for year 1.22 (1.21–1.22) 1.22 (1.21–1.22)
Indigenous status * year 0.85 (0.82–0.88) 0.88 (0.86–0.90)
Figure 2 Trends in use of non-Consulting Room HA items by Indigenous status and Year.
Table 5 shows the per cent use of non-consulting room HA items by State and Territory and Indigenous status. Use of the HA was relatively low in all jurisdictions. Rates of use were much higher in South Australia (SA) and Tasmania (Tas) than in other States and Territories. NT had the lowest take up rate overall. In all States and Territories with the exception of the NT use of non-consulting room HA items was significantly lower among Aboriginal and Torres Strait Islander people. In the NT the trend was reversed with Aboriginal and Torres Strait Islander people being more likely the rest of the community to use the HA items. All differences remained significant when it was assumed that Aboriginal and Torres Strait Islander people aged at least 75 years used services at the same rate as the rest of the community.
Table 5 Per cent use of non-Consulting room HA items by Indigenous status and State/Territory
State 75+ yrs non-Indigenous 55+ yrs Indigenous Total Difference % (95% CI) Difference % (95% CI)-dual eligibility adjustment
NSW Count 132055 658 132713
% 6.5% 1.4% 6.4% 5.2 (5.2–5.3)% 4.5 (4.4–4.5)%
VIC Count 94610 228 94838
% 6.4% 2.3% 6.3% 4.2 (4.1–4.2)% 3.2 (3.2–3.3)%
QLD Count 50911 505 51416
% 5.2% 1.2% 5.0% 4.1 (4.0–4.1)% 3.27 (3.2–3.3)%
SA Count 62203 176 62379
% 11.5% 2.0% 11.3% 9.5 (9.4–9.6)% 8.1 (8.0–8.2)%
WA Count 25725 189 25914
% 5.4% 0.8% 5.1% 4.5 (4.5–4.6)% 3.8 (3.7–3.8)%
TAS Count 15906 29 15935.0
% 10.4% 0.5% 10.1% 9.9 (9.8–10.1)% 8.8 (8.7–9.0)%
ACT Count 2979 4 2983.00
% 4.8% 0.5% 4.8% 4.3 (4.1–4.5)% 3.9 (3.7–4.1)%
NT Count 40 264 304
% 0.3% 1.4% 0.9% -1.0 (-1.2–-0.9)% -1.0 (-1.2–-0.9)%
Uptake of Aboriginal and Torres Strait Islander adult HA item compared to uptake of HA items for older people
The logistic regression for the uptake of consulting room HA items in the first three quarters of their introduction suggested that uptake of the HA items for adult Aboriginal and Torres Strait Islander people (710) was lower than for the uptake of the general population item (700) but was higher than uptake of the item for older Aboriginal and Torres Strait Islander people (704, see table 6). Both linear and quadratic trends were significant because rates of use increased substantially after the first quarter and then stabilised in the second and third (see table 7).
Table 6 Logistic Regression for uptake of Consulting room HA items in the first 3 quarters after their introduction
Variable OR (95% CI)
Linear trend-quarters 2.15 (2.02–2.29)
Quadratic trend-quarters 0.63 (0.60–0.66)
75+ yrs Non-Indigenous HA 2.6 (2.49–2.67)
55+ yrs Indigenous HA 0.70 (0.61–0.80)
15–44 yrs Indigenous HA Reference
Table 7 Per cent use of Consulting room HA items in the first 3 quarters after their introduction
HA Quarter 1 Quarter 2 Quarter 3
%-95%CI %-95%CI %-95%CI
75+ yrs non-Indigenous 0.63 (0.62–0.64) 1.74 (1.73–1.76) 1.73 (1.72–1.75)
55+ yrs Indigenous 0.16 (0.14–0.19) 0.48 (0.44–0.52) 0.51 (0.47–0.55)
15–44 yrs Indigenous 0.23 (0.23–0.24) 0.71 (0.69–0.72) 0.69 (0.67–0.7)
Discussion
Uptake of HA items was relatively low overall and there was significant and persistent disparity in the uptake of HA items for older people among Aboriginal and Torres Strait Islander people compared to the rest of the community. There were significant differences between the jurisdictions in the overall uptake of items. For consulting room items there appeared to be a relationship between overall State or Territory population and uptake although this was unrelated to other factors such as population density. [16] There was no clear pattern for non-consulting room items. NT was the only jurisdiction where Aboriginal and Torres Strait Islander people used HA items more than non-Aboriginal people. This appeared to occur because of low uptake among non-indigenous Australians rather than higher uptake among Aboriginal and Torres Strait Islander people.
The comparison of the uptake of the HA items in the first three quarters of their use suggested that the uptake of the new adult Aboriginal and Torres Strait Islander item was lower than the uptake of the HA item for older members of the general population. This suggests that additional attention to the causes of barriers to the uptake of HAs among Aboriginal and Torres Strait Islander people may be necessary to achieve the potential benefits associated with these items.
In the evaluation of the EPC program it was suggested that disparities in the uptake of HAs for older people could either be a function of Aboriginal and Torres Strait Islander people having pre-existing care plans or the result of Aboriginal and Torres Strait Islander patients being more likely to see doctors who were ineligible to use Medicare. [2] Differences due to both causes would be expected to decrease over time. Any variation in uptake due to difference in levels of pre-existing HAs would be reduced over time as HAs were renewed. Since the original evaluation an exemption under section 19(2) of the National Health Act has enabled salaried doctors in approved services to bill through Medicare. This has resulted in increased rates of Medicare use at Aboriginal and Torres Strait Islander services. There was some evidence of a slightly faster rate of increase in use among Aboriginal and Torres Strait Islander people for consulting room items though rates appeared stable for non-consulting room items. The persistence of the disparity suggests neither explanation accounts for a large part of the difference in HA uptake between older Aboriginal and Torres Strait Islander people and other Australians.
The EPC evaluation found that awareness of HA items was high among doctors but that lack of awareness of the items among consumers and allied health workers was a barrier to their uptake. [3] Consumer awareness may be particularly important in the use of Aboriginal and Torres Strait Islander items where client identification is an issue. Uptake of HA items was facilitated in practices where practice nurses rather than the doctor undertook the information gathering components. [2] The provision of additional assistance to conduct HAs may be particularly important in Aboriginal and Torres Strait Islander health services where the ratio of walk-in to appointment-based consultations is far higher than in main stream services, making it difficult for doctors to block out the time required for HAs. Even greater barriers may exist in communities were there is no full-time doctor. Others barriers include racism and problems with cross-cultural communication. [10] Barriers associated with cultural appropriateness may be addressed by initiatives such as the development of a guide to conducting health assessments in Aboriginal and Torres Strait islander people. [9] However a multifaceted approach is likely to be required. [2,9]
In any analysis of health services data where clinical data is absent it is difficult to determine appropriate levels of HA use. However it does not seem clinically plausible that Aboriginal and Torres Strait Islander people should be less in need of HAs than comparable other Australians. Ameliorating this situation may require not only further promotion of the items with doctors but further engagement of local primary health infrastructure and the community. [2,5] The evaluation of the HA items and previous initiatives to promote health checks[11] in Aboriginal and Torres Strait islander communities are valuable resources in developing approaches to ensure that the potential health benefit deriving from the new and existing items are delivered.
Statement of Competing cnterests
The author(s) declare that they have no competing interests.
Authors' contributions
Margaret Kelaher conceptualised the paper and conducted the analysis. David Dunt, Ian Anderson and David Thomas collaborated in drafting the paper.
Table 2 Logistic Regression for use of Consulting room HA items by Indigenous status and Year controlling for State/Territory
Variable Observed service use Dual eligibility adjustment
OR (95% CI) OR (95% CI)
Indigenous status 0.37 (0.36–0.38) 0.51 (0.50–0.53)
Linear trend for year 1.12 (1.12–1.12) 1.12 (1.12–1.12)
Indigenous status * year 1.11(1.10–1.14) 1.03(1.01–1.06)
Acknowledgements
Margaret Kelaher is funded by an NHMRC career development award and VicHealth. David Thomas is supported by NHMRC Population Health Capacity Building Grant No. 236235. Core funding for Onemda VicHealth Koori Health Unit is provided by the Victorian Health Promotion Foundation and the Commonwealth Department of Health and Ageing. Our thanks to Sophie Couzos for her comments on an earlier draft of this paper.
==== Refs
Australian Department of Health and Ageing Medicare Benefits Schedule Book Operating from 1 November 2004 2004 Canberra: Commonwealth of Australia
Wilkinson D Mott K Price K Morey S Beilby J Best J McElroy H Pluck S Eley V Evaluation of the Enhanced Primary Care Medical Benefits Schedule Items and the General Practice Education, Support and Community Linkages Program. Final Report 2004 Canberra: Australian Department of Health and Ageing
Wilkinson D McElroy H Beilby J Mott K Price K Morey S Best J Variation in levels of uptake of enhanced primary care item numbers between rural and urban settings, November 1999 to October 2001 Aust Health Rev 2002 25 123 30 12536871
Byles JE A thorough going over: Evidence for health assessments for older persons Aust N Z J Public Health 2000 24 117 123 10790930
Byles JE Tavener M O'Connell RL Nair BR Higginbotham NH Jackson CL Mckernon ME Francis L Heller RF Newbury JW Marley JE Goodger BG Randomised control trial of health assessments for older Australian veterans and war widows Med J Aust 2004 181 186 190 15310251
Wilkinson D McElroy H Beilby J Mott K Price K Morey S Best J Uptake of health assessments, care plans and case conferences by general practitioners through the Enhanced Primary Care program between November 1999 and October 2001 Aust Health Rev 2002 25 1 11 12404961
Australian Institute of Health and Welfare (AIHW) Expenditures on health services for Aboriginal and Torres Strait Islander people 1998–99 2001 Canberra: AIHW and Commonwealth Department of Health and Aged Care AIHW cat no. IHW
Department of Health and Ageing New Medicare Checks for Indigenous Australians
Mayers NR Couzos S Towards equity through an adult health check for Aboriginal and Torres Strait Islander people Med J Aust 2004 181 531 532 15540961
Trewin D Madden R The Health and Welfare of Australia's Aboriginal and Torres Strait Islander Peoples 2003 2004 Canberra: Australian Institute of Heath and Welfare (AIHW) and Australian Bureau of Statistics ABS Catalogue no. 4704.0 AIHW Catalogue no. IHW11 ISSN 1441–2004
Miller G McDermott R McCulloch B Leonard D Arabena K Muller R The Well Person's Health Check: A population screening program in indigenous communities in north Queensland Aust Health Rev 2004 25 1 11
Health Insurance Commission Medicare accessed March 2005
Australian Bureau of Statistics (ABS) Experimental Estimates and Projections, Aboriginal and Torres Strait Islander Australians 2004 Canberra: ABS ABS catalogue no. 3238.0
Australian Bureau of Statistics (ABS) Experimental Projections of the Aboriginal and Torres Strait Islander Population Canberra: ABS ABS catalogue no. 3231.0
Australian Bureau of Statistics (ABS) Population by Age and Sex, Australian States and Territories 2004 Canberra: ABS Time Series Spreadsheet 3201.0
Australian Bureau of Statistics (ABS) National Regional Profile 2004 Canberra: ABS ABS catalogue no. 1379.0.55.001.
|
16150154
|
PMC1239906
|
CC BY
|
2021-01-04 16:38:28
|
no
|
Aust New Zealand Health Policy. 2005 Sep 9; 2:21
|
utf-8
|
Aust New Zealand Health Policy
| 2,005 |
10.1186/1743-8462-2-21
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2171613525510.1186/1471-2105-6-217Methodology ArticleAutomation of gene assignments to metabolic pathways using high-throughput expression data Popescu Liviu [email protected] Golan [email protected] Department of Computer Science, Cornell University, Ithaca, NY2005 31 8 2005 6 217 217 18 7 2005 31 8 2005 Copyright © 2005 Popescu and Yona; licensee BioMed Central Ltd.2005Popescu and Yona; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Accurate assignment of genes to pathways is essential in order to understand the functional role of genes and to map the existing pathways in a given genome. Existing algorithms predict pathways by extrapolating experimental data in one organism to other organisms for which this data is not available. However, current systems classify all genes that belong to a specific EC family to all the pathways that contain the corresponding enzymatic reaction, and thus introduce ambiguity.
Results
Here we describe an algorithm for assignment of genes to cellular pathways that addresses this problem by selectively assigning specific genes to pathways. Our algorithm uses the set of experimentally elucidated metabolic pathways from MetaCyc, together with statistical models of enzyme families and expression data to assign genes to enzyme families and pathways by optimizing correlated co-expression, while minimizing conflicts due to shared assignments among pathways. Our algorithm also identifies alternative ("backup") genes and addresses the multi-domain nature of proteins.
We apply our model to assign genes to pathways in the Yeast genome and compare the results for genes that were assigned experimentally. Our assignments are consistent with the experimentally verified assignments and reflect characteristic properties of cellular pathways.
Conclusion
We present an algorithm for automatic assignment of genes to metabolic pathways. The algorithm utilizes expression data and reduces the ambiguity that characterizes assignments that are based only on EC numbers.
==== Body
Background
Pathways are cellular procedures that are associated with a specific functionality in the cell, such as amino acid synthesis and degradation, energy metabolism, signal transduction, molecular oxidation, and more. The complexity of a cell is a function of its underlying procedures. Therefore, there is a strong interest in identifying the active pathways in an organism. This knowledge can not only shed light on the mechanisms the cell uses to acquire its functional role; by assigning genes to pathways one can also better understand the exact role of these genes, and identify key genes whose existence is crucial to sustain normal cell functionality.
A wealth of experimental data about molecular complexes and cellular processes that has been accumulated in the literature initiated several projects that attempted to compile the existing knowledge into publicly available databases. Among these are EMP [1], MPW [2], WIT [3], UM-BBD [4], KEGG [5], MetaCyc [6], ERGO [7] and SEED [8]. These databases store valuable information about hundreds of pathways and cellular processes.
Much of the research on pathways so far focused on extrapolating pathways from one organism to other. The goal of this research goes beyond just storing, analyzing and extrapolating the biochemical information and strives to improve the known data by discovering variations to pathways in different organisms as well as to discover novel pathways.
Attempting to complement the experimental data and extend its utility to other systems and newly sequenced genomes, several methods were developed for pathway prediction. One approach to pathway reconstruction is to utilize the existing knowledge on enzymatic reactions to create a complete graph of a possible metabolic network [9-12]. However, this approach is sought with complexity problems and it is hard to verify the validity of these predictions. Several studies manually constructed and curated the metabolic networks for organisms like Escherichia coli [13-15], Haemophilus influenzae [16] and Saccharomyces cerevisiae [17,18], from a variety of data sources and literature. These studies have an advantage over automated pathway reconstruction, as the reconstructed networks are more likely to be biologically plausible. However, this approach requires close human intervention.
Perhaps the most popular approach for pathway prediction is based on extrapolation. Procedures developed by WIT, KEGG, MetaCyc and ERGO use blueprints of pathways collected either from biochemical charts or from actually observed pathways in different organisms, and assign genes to pathways based on homology between genes across organisms, database annotations, and manual curration. Specifically, many known pathways are metabolic pathways that consist mostly of sets of reactions catalyzed by specific enzymes that are designated by their Enzyme Classification (EC) number [19]. Most of the existing methods for metabolic pathway prediction that are based on pathway blueprints assign the vast majority of genes to pathways based on their EC designation (some address also the problem of finding missing enzymes [20-24]). However, since certain reactions appear in multiple pathways, this method will assign all enzymes that can catalyze a certain reaction (termed isozymes) to all pathways that contain this reaction. For example, genes that are designated as malate dehydrogenase (EC 1.1.1.37) are classified to several different pathways (including mixed acid fermentation, gluconeogenesis, superpathway of fatty acid oxidation and glyoxylate cycle, respiration, and more), all of which use the same oxidization reaction that is catalyzed by these genes.
Clearly, this nondiscriminatory assignment of genes to pathways is suboptimal, as it is unlikely that all genes with the same EC designation are used in all pathways that contains the corresponding reaction. Rather, it is more likely that different genes are used in different pathways, and it has been suggested [25] that "the primary role of isozymes is to allow for differential regulation of the reactions that are shared by different processes". However, this information is sparse and without additional experiments it is very hard to make this type of functional differentiation. The extent of this problem is not negligible. For example, of the 469 pathways in MetaCyc, 336 have at least one reaction in common with another pathway. Since in most genomes there are multiple instances of some enzyme families, the common method for pathway prediction (that is based only on EC numbers) results in many-to-many ambiguous mapping between genes and pathways.
Pathway assignment can be aided by the existence of microarray technology [26-29]. This technology enables genome-wide measurements of cell activity, providing us with snapshots of the molecular machinery at different times along the cell cycle and under different experimental conditions. This data can help to identify groups of genes that are co-expressed, i.e. that are likely to exist in the cell at the same time or under the same set of conditions. Although the sequence of reactions in a pathway does not take place simultaneously, given the time-resolution of the mRNA expression measurements these reactions can be considered to occur instantly and simultaneously for all practical purposes. Therefore, it is expected that genes that participate in the same pathway will have similar expression profiles, i.e. they will co-exist and will be concurrently available at the cell's disposal to complete the pathway. Indeed, correlation in expression profiles has been observed for linear pathways that consist of sequences of reactions [25]. It has been also shown that prediction based on search in the pathway space improves when pathways are scored using expression data [30]. Other studies used expression data to score gene classes and pathways, in search of interesting classes or modules of genes or to verify the existence of certain pathways in a genome [31-38]. Expression data can also suggest the existence of control mechanisms and pathway switches. For example, when a pathway has a fork, isozymes might be used to switch between the alternate routes, resulting in anti-correlation [25]. A detailed discussion of these studies and others in the field of pathway prediction and analysis appears in Appendix A.
Here, based on this premise, we propose a method for improving the gene-to-pathway assignment problem and refining the large-scale predictions of pathways provided by systems like WIT, Pathway Tools and KEGG (that use EC designation only). Our method utilizes pathway blueprints, statistical models of protein families and expression data. As opposed to previous methods, our algorithm focuses on elucidating the correct assignment of genes to pathways and expression data is used to score assignments rather than pathways. Our algorithm predicts all assignments simultaneously, while resolving possible conflicts and optimizing the correlated expression.
The paper is organized as follows. We first describe our model and the prediction algorithm. Next we evaluate our methods by running a full-scale prediction on the yeast genome. Finally we compare our predictions to the few assignments that were verified experimentally.
Results
Our model organism is Yeast. This choice was motivated by the myriad of experimental data available for the Yeast genome, and specifically, time-series expression data which is not readily available for other genomes. Our study integrates pathway data with expression data and sequence data. Information on the datasets used in this study is available in the 'Methods' section.
There are many definitions of pathways in the literature and on-line, depending on the context in which they are used. In our work we adopt the same definition that is used in many other studies and underlies the pathways in databases such as MetaCyc and KEGG. As was characterized concisely in [39]: "A metabolic pathway is a sequence of consecutive enzymatic reactions that brings about the synthesis, breakdown, or transformation of a metabolite from a key intermediate to some terminal compound. A metabolic pathway may be linear, cyclic, branched, tiered, directly reversible, or indirectly reversible."
We formalize the concept of a metabolic pathway according to this definition and it is assumed that each pathway P consists of a set of enzymatic reactions which together perform a certain function. Each reaction can be catalyzed by enzymes that are typically associated with one Enzyme family F.
Pathway assignments – algorithm overview
Our algorithm for assigning genes to pathways takes as input
• A genome G = {g1, g2, ..., gN}
• Expression data E = {Ei} where Ei is the expression profile of gene gi
• An assignment of genes to enzyme families F = {F1, .., FJ}
• A set of metabolic pathways P = {P1, .., PK}.
Our method consists of the following steps:
1. Search for probable pathways. For each pathway Pk ∈ P:
(a) match enzymes with the reactions that make up the pathway;
(b) eliminate the pathway if more than θ of the reactions cannot be associated with genes (here we set θ = 0.5).
The resulting set of pathways is denoted P'
2. Compute initial pathway assignments and sort assignments according to the score from high to low.
3. Refine assignments. Given the assignments from the previous step:
(a) compute the conflict graph;
(b) compute the connected components in the conflict graph;
(c) solve the conflicts within each connected component.
We now proceed to describe each step in detail.
Search for probable pathways
To assign genes to pathways in a given sequenced and annotated genome, we use the descriptions of the pathways from MetaCyc and the classification of genes to enzyme families (based on annotations or statistical models, as described in 'Methods') to initially match each enzyme with a reaction and therefore with a pathway.
We denote by F(Pk) = {F1, F2, ..., Fm} the set of pathway families – the protein families that catalyze the reactions that make up pathway Pk, where m is the number of different reactions (the number of reactions need not be equal to the number of families, however, each reaction in a pathway is usually associated with one family). A pathway is kept if at least m/2 of its reactions can be assigned with enzymes. Formally, denote by F(G) the set of families that can be associated with at least one gene in the genome G. A pathway Pk is considered probable in the genome G if |F(Pk) ∩ F(G)| ≥ m/2. We denote the set of probable pathways by P', and our algorithm proceeds only with pathways in P'. Note that at this stage there might be multiple genes assigned to the same reaction.
Initial pathway assignments
After eliminating the improbable pathways we generate initial assignments by computing the best individual assignment for each candidate pathway, independently. We are given a pathway Pk with m reactions. In search for the optimal assignment we consider all genes in each one of the families F1, F2, ..., Fm ∈ F(Pk), resulting in |F1| × |F2| × ...|Fm| possible assignments. Each possible combination is considered and we evaluate its significance by computing the total correlation score between genes. I.e. the score of assignment A = (g1, g2, ..., gm) s.t. gi ∈ Fi is defined as the average co-expression score
where sim(Ei, Ej) is the expression similarity of genes gi and gj as described in 'Methods' and wi is the weight that represents the likelihood that gene gi belongs to family Fi and is defined as where evalue(i) is the significance of the match between gene i and the statistical model of family Fi (see 'Methods'). For example, assume the best match with family Fi is observed for an annotated gene with evalue of 10-20. Then a gene that is classified to that family with evalue of 10-10 is assigned a weight of 0.5.
After computing all the assignment scores we sort them in the order from best to worst. The best assignment is selected as the one that maximizes the average score.
Multi-domain proteins
Of the 71,638 proteins in our database with an EC designation (see section 'Data sets' in 'Methods'), about 1241 have multiple enzymatic domains. Of which, the majority (1076 proteins) are two-domain proteins that form 173 unique combinations. A simple statistical analysis reveals that these proteins are more likely to contribute all their domains to the same pathway. Specifically, we computed the fraction of two-domain enzymes that can be completely mapped to a single pathway (i.e. there exist at least one pathway such that all the enzymatic domains take part in). Of the 173 two-domain combinations, 67 are combinations of domains that are in our pathway data set. Of which 48 (72%) can be mapped completely to a single pathway. The expected fraction is estimated assuming that the two domains are chosen at random from the domain library, and computing how many random pairs appear in the same pathway. Of the 6786 possible combinations of domain pairs (using the domain library derived from the set of 67 combinations used above) only 199 (3%) are mapped to a single pathway. The significant difference (72% vs. 3%) indicates a clear bias for multi-domain proteins.
This is not surprising, as multi-functional proteins would be thermodynamically favorable in pathways. If two reactions in a pathway can be catalyzed by the same protein, the efficiency of the reaction can significantly increase, since it saves the need to localize and control the expression of multiple proteins. If the two reactions are consecutive, it is quite likely that the output of one reaction is immediately transferred as an input to the second reaction catalyzed by the second domain. To account for this scenario in our model and create a natural bias toward multi-domain proteins, we use the self-similarity score when assigning these genes to two (or more) different reactions within the same pathway. With that bias, multi-domain proteins will be preferred whenever some or all their domains can be utilized in the same pathway.
Computational issues
To find the best assignment of genes to a given pathway we exhaustively enumerate all possible assignments. This is possible for most pathways, genomes and families. For example, most of the pathways in Yeast have less than a hundred possible assignments in our current setting. However, some of the protein families are fairly large (hundreds and even thousands of members), resulting in a large number of possible assignments. The maximum number of pathway assignments in Yeast is observed for the tRNA charging pathway which has 49,152 possible assignments. Considering all possible combinations in the cross-product is computationally intensive, and also unnecessary. To reduce the number of assignments that are considered one can first compute the similarity scores of all possible pairs and use only pairs that have significant similarity score (see the 'Metrics' section in 'Methods') or limit the analysis to the top N scoring pairs. In practice, given the size of a typical pathway as well as the number of possible genes catalyzing a reaction, the complete enumeration of assignments is possible in a reasonable time (a matter of minutes).
Refining the assignments
The initial set of assignments is likely to produce a good unique mapping between genes and pathways (see 'Discussion'). However, since each pathway is analyzed independently it might happen that the same gene is assigned to the same reaction in multiple pathways. Each such assignment is considered a conflict. Although in some cases the same gene might play the same functional role in different pathways, our hypothesis is that if there are multiple enzymes in the same genome that can catalyze the same reaction, and that reaction takes place in multiple pathways, then it is more probable that each enzyme is "specialized" to catalyze this reaction in a different pathway. To eliminate the conflicts we revisit the assignments and resolve them whenever it is possible, as described next.
The pathway conflict graph
We start by constructing the pathway relation graph. In this graph each pathway is a node, and two nodes are connected by an edge if the two pathways represented by the nodes share a reaction (see Figure 1a). We introduce one edge for each such reaction (i.e. there might be multiple edges connecting the same two nodes). The pathway conflict graph is derived from this graph: we mark an edge as a conflict if the corresponding reaction is associated with the same gene in both pathways, based on the initial assignments (see Figure 1b).
Figure 1 Pathway graphs. Left: the pathway relation graph. Each pathway is represented as a node, and an edge is drawn between two pathways for each reaction that they share in common. Middle: the pathway conflict graph. Thick edges represent conflicts (i.e. the same gene was assigned to catalyze the same reaction in both pathways connected by the edge). Right: the final conflict graph. The edge between pathways P9 and P10 is a flat edge (no alternative assignments exist for that reaction) and therefore it is unmarked. At the end we are left with only two connected components with possibly solvable conflicts.
Connected components
The pathway conflict-graph can be split into connected components, each of which is composed of several pathways connected by edges (reactions), some of which are marked and indicate possible conflicts. If several genes are associated with such a reaction, it might be possible to resolve this conflict. Clearly, the assignments in one connected component have no effect on the other connected components and therefore we can revisit these assignments independently for each connected component.
It should be noted that not all conflicts can be resolved. If in the given genome there is only a single gene that can be associated with a specific reaction, then clearly it is not possible to refine conflicts associated with that reaction. An edge that is linked with such a reaction is referred to as a flat edge. Since for flat edges no alternative assignments exist (given the gene data), we unmark these edges in the conflict graph. Our algorithm operates only on connected components with marked edges (see Figure 1c).
Assignment of genes to pathways in a connected component
To find the best non-conflicting assignment of genes to pathways in a connected component we generalize our scoring function such that the score of an assignment is the sum of the scores of the assignments to pathways contained in the component, with the restriction that no enzyme can be used twice to catalyze the same reaction in different pathways. We ignore inter-pathway expression data correlations, assuming different pathways are associated with different cellular processes and therefore are not expected to be correlated.
Formally, given a set of pathways P = P1, P2, .., Pk and an assignment A, the assignment score is simply the total weighted co-expression score
where A(Pi) is the subset of genes assigned to pathway Pi and Score(A(Pi)) is as defined previously. As before we enumerate all possible assignments of genes to pathway families and each assignment is also marked with the number of conflicts it introduces.
Ideally we would like to find high scoring assignments that are conflict-free. However, this ideal situation is not always attainable as some shared reactions are central and are best catalyzed by the same enzyme. For example, reaction 2.6.1.1 that is shared by the asparagine and aspartate biosynthesis pathways is catalyzed in both pathways by gene AAT2, although there exist another gene (AAT1) that can catalyze this reaction (see 'Discussion'). Moreover, some pathways are superpathways of other pathways and are naturally composed of the same genes. Therefore, not all conflicts can and should be resolved. To accommodate these possible scenarios we consider all assignments that are within Δ from the maximal score that is obtained when conflicts are allowed, and pick the one that has the minimal number of conflicts within that range (without a better methodology at this point, the exact value of Δ is currently set ad-hoc to 1). A significant drop in the score of a conflict-free assignment (compared to the highest scoring assignment) suggests that some reactions are indeed catalyzed by the same gene, despite the fact that alternative genes do exist to perform similar functions.
Discussion
Evaluating our pathway prediction algorithm requires the availability of well studied and annotated genome for which high-quality expression data and empirical knowledge of pathways exist. Since most pathway databases assign genes to pathways collectively based on the EC designation it was hard to find an extensive set of literature-curated pathways. We used one of the curated PGDB Yeast Biochemical Pathways [44] at the Saccharomyces Genome Database (SGD) [45]. This database was computationally derived from the Yeast sequenced and annotated genome using the Pathway Tools software [46] and the pathway blueprints from the Metacyc database [6], and was then manually curated by mining the literature. Not all the assignments were done based on direct phenotype experiments and the confidence in the assignments varies depends on the type of the evidence used. The database contains 58 pathways, many of which did not exist in the Metacyc database or did not match perfectly with the pathway blueprints in Metacyc. A few other pathways contained genes that we were not able to map to our Yeast protein database, and were eliminated as well. This left us with 25 pathways that were used for testing. Each curated pathway in the SGD database describes a sequence of reactions as well as the genes that catalyze the reactions. Some reactions are not associated with a specific gene and were not considered when evaluating the correctness of an assignment. Also, some of the reactions are unclassified reactions that either have an incomplete EC number or do not have an EC number at all. These reactions are currently ignored in our experiments.
It should be noted that some of the curated pathways associate multiple genes with the same reaction. In general, it seems that there are two possible explanations. It might be the case that a complex of proteins catalyzes the reaction and the genes associated with the reaction are part of this complex. In this case we want to assign all proteins to the reaction. This is not taken into account in our algorithm currently. The other more common case is when each protein can catalyze the reaction by itself, for example under different specific cellular conditions. This can be verified in knockout experiments and has been observed in several systems (e.g. [47]). While it is possible that all these genes are used concurrently, our assumption is that only a few of them actually do. In these cases, our algorithm can assess the "affinity" of each gene with the pathway. In the next sections we discuss our results and compare them with the curated assignments of the pathways in the test set.
Pathway assignment for curated pathways
To test our predictions, we run the algorithm on the Yeast genome, using the time-series expression data and the blueprints of the 25 pathways in our test set. The EC annotations were updated to be consistent with those used by SGD. It should be noted that for many pathways all possible assignments are curated as valid assignments by SGD. A summary of the results is given in Table 1.
Table 1 Summary of pathway assignments. For each pathway in the test set we report the number of reactions, the number of assignments considered, the number of curated (SGD verified) assignments, and the maximal and minimal assignment scores. The score reported is the weighted average score per pair of compared enzymes. The score reflects the average significance of a pairwise relation within a pathway. The larger the score the more significant is the relation. Negative scores suggest anti-correlation and near-zero scores provide no evidence that the two genes are related. Pathways are sorted based on assignment score.
Pathway Number of reactions Number of assignments Number of curated assignments Max(Min) Score Normalized
methionine and S-adenosylmethionine synthesis 2 2 2 10.45 (7.34)
isoleucine biosynthesis I 5 12 4 10.32 (3.00)
leucine biosynthesis 4 4 4 10.08 (4.01)
valine biosynthesis 4 4 4 10.14 (4.99)
asparagine biosynthesis I 2 4 4 8.80 (-4.88)
proline biosynthesis I 3 1 1 8.43 (8.43)
homoserine methionine biosynthesis 2 1 1 7.33 (7.33)
tryptophan biosynthesis 5 2 2 5.29 (4.13)
aspartate biosynthesis II 2 4 4 4.85 (0.75)
non-oxidative branch of the pentose phosphate pathway 5 8 8 4.82 (0.84)
folic acid biosynthesis 11 48 32 4.63 (1.26)
glutamate biosynthesis I 2 2 2 4.08 (-4.88)
glutathione biosynthesis 2 1 1 4.03 (4.03)
glutamate degradation VIII 5 1 1 3.92 (3.92)
serine biosynthesis 3 2 2 3.58 (-0.58)
purine biosynthesis 2 14 16 8 2.50 (2.06)
homocysteine and cysteine interconversion 3 2 1 2.35 (2.01)
biotin biosynthesis I 3 1 1 2.27 (2.27)
homocysteine degradation I 2 1 1 2.01 (2.01)
threonine biosynthesis from homoserine 2 1 1 0.87 (0.87)
glutamine – glutamate pathway II 1 1 1 0.00 (0.00)
tyrosine biosynthesis I 3 2 2 -0.53 (-0.58)
glycine biosynthesis I 2 2 2 -0.91 (-3.60)
phenylalanine biosynthesis I 3 2 2 -2.09 (-2.80)
Almost all curated assignments are assigned high positive scores (results not shown). There are some exceptions and a few curated assignments have a negative score. In these cases most or all other assignments have negative scores as well. For 13 out of the 25 pathways the maximum normalized score is greater than 4. The score is an indication of how significant is the similarity of two expression profiles [43]. An average score greater than 4 means that the enzymes assigned to the pathways are similarly expressed with high confidence and are likely to be functionally linked. Moreover, for 10 out of the 25 pathways, all pairwise scores (for all pairwise relations) in the top-scoring assignment are positive. These results support our assumption that proteins that participate in the same cellular process are similarly expressed. It is also observed that curated assignments are assigned better scores than the non curated assignments and the best assignment is usually a curated assignment. In the next subsections we take a closer look at some interesting pathways.
The isoleucine biosynthesis pathway
As Table 1 indicates, the expression data strongly supports the existing knowledge about pathways and can be used for prediction. The isoleucine biosynthesis pathway is one such example (Figure 2). This pathway consists of 5 reactions. A total of 12 assignments are considered, of which 4 are curated and are considered true assignments, and 8 are considered false assignments. Table 2 lists detailed information about each candidate assignment. Note that curated assignments are assigned a high positive score, and the normalized score of the best assignment is well over 4. Moreover, the true and false assignments are well separated in the sorted list. The baseline score is determined by the two enzymes (EC 1.1.1.86 and EC 4.2.1.9) that have no alternative genes and are shared by all assignments. Looking at the break-up of pairwise similarities within the pathway we note that almost all of them have positive scores for curated assignments, while false assignments contain more pairs with negative pairwise scores.
Figure 2 The Isoleucine Biosynthesis pathway diagram. The pathway layout is retrieved from the MetaCyc database. For each reaction we list the genes that can catalyze the reaction. A plus or minus sign indicates if the gene was assigned to the pathway in SGD. The expression profiles and their similarity score are shown for selected pairs of genes. Mapping between gene names and Biozon identifiers is given in Table 6.
Table 2 Assignments for the pathway isoleucine biosynthesis I. Only reactions with alternative assignments are listed (last column), and the selection number refers to Figure 2. For example, the top assignment selects the second gene (ILV1) to catalyze reaction 4.3.1.19. Assignments are sorted based on the normalized score. Second column marks which assignments are true assignments, and which are considered false assignments. For each assignment we list the total number of pairwise similarities, the number of positive and negative scoring pairs and the number of zero scoring pairs (when no expression data is available).
Number Match Normalized Score Number of Pairs Positive Pairs Negative Pairs Zero Pairs Assignments
1 + 10.32 10 10 0 0 4.3.1.19 : 2
4.1.3.18 : 2
2.6.1.42 : 1
2 + 9.54 10 10 0 0 4.3.1.19 : 2
4.1.3.18 : 1
2.6.1.42 : 1
3 + 7.37 10 10 0 0 4.3.1.19 : 2
4.1.3.18 : 2
2.6.1.42 : 2
4 - 6.49 10 8 2 0 4.3.1.19 : 3
4.1.3.18 : 2
2.6.1.42 : 1
5 + 6.30 10 9 1 0 4.3.1.19 : 2
4.1.3.18 : 1
2.6.1.42 : 2
6 - 6.08 10 6 0 4 4.3.1.19 : 1
4.1.3.18 : 2
2.6.1.42 : 1
7 - 5.52 10 6 0 4 4.3.1.19 : 1
4.1.3.18 : 1
2.6.1.42 : 1
8 - 5.00 10 7 3 0 4.3.1.19 : 3
4.1.3.18 : 1
2.6.1.42 : 1
9 - 4.78 10 8 2 0 4.3.1.19 : 3
4.1.3.18 : 2
2.6.1.42 : 2
10 - 3.84 10 6 0 4 4.3.1.19 : 1
4.1.3.18 : 2
2.6.1.42 : 2
11 - 3.00 10 6 4 0 4.3.1.19 : 3
4.1.3.18 : 1
2.6.1.42 : 2
12 - 3.00 10 5 1 4 4.3.1.19 : 1
4.1.3.18 : 1
2.6.1.42 : 2
Note that the first two assignments are both assigned high positive scores. These two assignments differ in the gene used to catalyze the EC reaction 4.1.3.18. The first is using ILV2 while the second is using ILV6 Interestingly, these proteins form a complex which catalyzes the reaction 4.1.3.18 [45]. This and similar cases will be handled in future versions of our algorithm (see the Conclusion section). Of the curated assignments, the fourth one leads to one negative pairwise score of -3.04 for proteins ILV6 and BAT2. Protein BAT2 can catalyze the reaction 2.6.1.42. This reaction can also be catalyzed by protein BAT1, and its selection results in better assignment scores. The two proteins are very similar in sequence (77% identity), however, the former is highly expressed during stationary phase of the cell-cycle and down-regulated during the logarithmic phase of growth (as is documented in the SwissProt record of that gene), while the later exhibits the opposite behavior. In view of the expression data it is unlikely that the two genes participate in this pathway at the same time, and gene BAT2 is probably assigned only during the stationary phase where the pathway activity is reduced.
We compared our results with the reconstructed metabolic network of Saccharomyces cerevisiae, as described in [18] (see 'Related Studies'). This network is not compartmentalized into separate metabolic pathways, however, the reactions are grouped according to the cellular process they are involved with. The comparison revealed discrepancies between the pathway data from Metacyc and SGD and the metabolic network model, which complicated the comparison of the results.
For example, the "isoleucine biosynthesis I" MetaCyc pathway overlaps with the reaction group "Valine, leucine, and isoleucine metabolism". The group has 24 reactions while the MetaCyc pathway has five, of which four are part of the group and the fifth (reaction 4.3.1.19 which appears first) is part of the "Threonine and Lysine Metabolism" group. Moreover, while the EC numbers and the sequence of reactions with respect to the EC numbers are the same in MetaCyc and the network model, the reactions are different because they do not use the same substrates as intermediary metabolites.
Interestingly, the first four reactions in the isoleucine biosynthesis MetaCyc pathway take place inside the mitochondrion, while the last step of the pathway, reaction 2.6.1.42 (catalyzed by BAT1 and BAT2), takes place both in the mitochondrion and in the cytoplasm. Indeed, it has been verified experimentally that BAT1 resides in the mitochondrion while BAT2 resides in the cytoplasm (see Figure 3). In order to obtain cytoplasmic isoleucine, a transport reaction is necessary to transfer the final intermediary metabolite. That might explain why BAT2 is slightly uncorrelated with the other genes in the pathway. Such situations lead to "forks", where two branches are uncoupled even if they have the same EC number. Our assignments are consistent with these observations.
Figure 3 The Isoleucine Biosynthesis pathway from the reconstructed metabolic network of Saccharomyces Cerevisiae [18]. Reproduced with permission from Cold Spring Harbor Laboratory ©2004 (Duarte et al. 2004 [18]). The EC numbers and the genes associated with the reactions were added to diagram. The part that overlaps with the MetaCyc isoleucine biosynthesis pathway is circled.
The folic acid biosynthesis
The metabolic network along with other cellular processes form the computational elements of the cell and as with any computation of this magnitude it needs to be regulated and synchronized. This regulation is reflected in the expression levels of genes. A possible synchronization device might require for example that reaction A is not started until reaction B is completed. Therefore, the enzymes that can catalyze A and B are not expected to be active at the same time, a state that can be achieved by controlling the expression levels of the corresponding genes. This type of mechanism will create a functional anti-correlation which will be awarded with a negative score by our scoring system. Beyond controlling timing of reactions, anti-correlation might also reflect a control mechanism that is used to govern pathway activity and metabolic rate.
An illustration of this mechanism is the pathway folic acid biosynthesis whose trajectory traverses both the mitochondrion and the cytoplasm. This is a quite complex pathway (see Figure 4) and the mechanism might occur between genes FOL1, FOL2 and FOL3. FOL1 is present in the mitochondrion while FOL2 and FOL3 are in the cytoplasm. Gene FOL1 is strongly anti-correlated with genes FOL2 and FOL3 while these two genes are strongly correlated between them. Note that the input to the last reaction catalyzed by FOL1 is the output from two different parallel branches of the pathway (top part). The anti-correlation might serve as a synchronization mechanism to control the reactants flow in the pathway in the presence of forks (that split into or merge different branches).
Figure 4 The folic acid biosynthesis pathway diagram. See Figure 2 for description. Note that FOL1, ADE3 and MIS1 are multi-functional enzymes.
Interestingly, gene FOL1 is an enzyme with multiple enzymatic functions positioned between the reactions catalyzed by FOL2 and FOL3. FOL1 has three different enzymatic domains, classified as 4.1.2.25, 2.7.6.3 and 2.5.1.15. There are no other genes that are classified (based on database annotations or sequence similarity) to either of these three enzyme classes. This further supports our assumption that multi-domain enzymes are more likely to catalyze several reactions in the same pathway, are are preferred over different enzymes, each assigned to a different reaction.
Multi-domain enzymes are also used in the lower part of the pathway. Both MIS1 and ADE3 catalyze three different consecutive reactions. Both are assigned to this pathway by SGD, however, surprisingly, their mutual expression similarity is negative (-2.09), indicating anti-correlation. Interestingly, the three reactions are shared with other pathways (glycine degradation, formylTHF biosynthesis and carbon monoxide dehydrogenase pathway), and it is hypothesized that the two isozymes, MIS1 and ADE3, serve as switches, to control the pathway activity and its coupling with other pathways. Indeed, such a mechanism has been suggested in [25] to control pathway flow.
To better understand these mechanisms we compared our results with the metabolic network model of [18]. The pathway from MetaCyc has 15 reactions, of which 12 overlap with the reaction group "Folate Metabolism", which has 29 reactions. Four out of these 12 reactions are duplicated in the metabolic network model with one instance in the cytoplasm and one in the mitochondrion. The discrepancy between MetaCyc and the network model involves the sequence of reactions 4.1.2.25, 2.7.6.3 and 2.5.1.15, all catalyzed by FOL1 gene, which are differently connected in the network model (see Figure 5). The network model also shows a fourth catalytic function for FOL1. The location of the enzymes and reactions in the network model indicate the intricate trajectory between cytoplasm and mitochondrion.
Figure 5 The folic acid biosynthesis pathway from the reconstructed metabolic network of Saccharomyces Cerevisiae [18]. Reproduced with permission from Cold Spring Harbor Laboratory ©2004 (Duarte et al. 2004 [18]). The EC numbers and the genes associated with the reactions were added to diagram. The parts that overlap with the MetaCyc folic acid biosynthesis pathway are circled. The green circles indicate consistency while the red one indicates inconsistency.
In view of these discrepancies, choosing the "right" model for this pathway is difficult. However, our results confirm the interplay between cytoplasm and mitochondrion and can help distinguish between mitochondrial and cytoplasmic genes, as each subgroup is mutually co-expressed, suggesting that the pathway expression is controlled by two distinct regulatory programs.
The asparagine biosynthesis pathway
Not always it is possible to explain negative pathway scores (anti-correlation or no correlation). Sometimes, a gene that can catalyze a specific reaction in a pathway is not coordinated with the other genes in the pathway. This might be due to the fact that the gene functions in the pathway only under certain conditions while inactive under others [27]. Or the gene might serve as a backup gene that is activated only when the main one is missing or is malfunctioning [47]. This problem is especially pronounced if the main gene has not been identified yet. Indeed, despite extensive annotation efforts, many genes have not been characterized yet.
By analyzing pairwise scores within a pathway, our method can suggest which genes fit together better in the context of the pathway and which genes are unlikely to work together. Moreover, if the overall assignment score is negative then it might be the case that the pathway is not active in the expression data collected or the pathway might not exist in the organism at all. Negative scores might also expose errors in pathway assignments. One interesting example is the asparagine biosynthesis pathway (Figure 6). This pathway is intriguing, having four curated assignments, two of them with negative scores. This is a small pathway with only two reactions. It is gene AAT1, which catalyzes the first reaction of the pathway (2.6.1.1), that is responsible for the negative scores of two assignments. This gene is strongly anti-correlated with genes ASN1 (-2.24) and ASN2 (-4.87), which catalyze the second reaction. On the contrary, gene AAT2 is strongly correlated with both genes ASN1 (8.80) and ASN2 (8.56). Interestingly, the reaction 2.6.1.1 is shared with other three pathways (asparagine degradation, aspartate biosynthesis and glutamate degradation VII). Our results suggest that the two isozymes, which can catalyze the same reaction, are used selectively in different pathways; AAT2 is involved in asparagine and aspartate biosynthesis, while AAT1 is involved in asparagine and glutamate degradation (where it is assigned a high positive score). But why was AAT1 assigned to the asparagine biosynthesis pathway? A closer look at the entry for AAT1 in the SGD database reveals that the curator assigned this enzyme to the pathway based on its enzymatic activity only, which was determined experimentally. In the literature AAT1 is associated with aspartate degradation. Obviously synthesis and degradation cannot appear at the same time and hence the anti-correlation between AAT1 and genes ASN1 and ASN2. This is a clear example of the assignment problem, suggesting that even curated assignments can be further improved using our method. The metabolic network model [18] confirms the previous conclusions. This pathway has two reactions that are entirely contained in the "Alanine and aspartate metabolism" group, which has 15 reactions (see Figure 7). There are 3 instances of the reaction 2.6.1.1 in the network model, one in peroxisome (catalyzed by AAT2), the second in cytoplasm (also catalyzed by AAT2) and the third in mitochondrion (catalyzed by AAT1) (see Figure 7). On the other hand, the reaction 6.3.5.4 takes place in cytoplasm. The expression profiles are in agreement with these subcellular locations and indeed the cytoplasm genes ASN1/ASN2 are highly correlated with AAT2, while anti-correlated with the mitochondrion AAT1.
Figure 6 The asparagine biosynthesis pathway. See Figure 2 for description. Both ASN1 and ASN2 are correlated with AAT2 but are anti-correlated with AAT1 (selected pairwise similarities are shown). The later is localized to a different cellular compartment than the others, and is likely to be involved in other pathways (see text for details).
Figure 7 The asparagine biosynthesis pathway from the reconstructed metabolic network of Saccharomyces Cerevisiae 18. Reproduced with permission from Cold Spring Harbor Laboratory ©2004 (Duarte et al. 2004 [18]). The part that overlaps with the MetaCyc asparagine biosynthesis pathway is circled.
Genome wide results
We repeated our analysis, this time with a larger set of pathways from the MetaCyc database, to generate genome wide assignment of genes to pathways. Most pathways were not represented in the Yeast genome, and we restricted our analysis to pathways for which we could assign genes to all reactions (64 pathways). We eliminated pathways that had only one fully characterized reaction since our algorithm is based most dominantly on expression similarity, and therefore assumes at least two reactions in a pathway. Reactions with incomplete EC number or which could not be assigned to a gene in the Yeast genome were ignored. In total, 52 pathways were considered.
We ran our procedure using the two different expression data sets (see the 'Data sets' section). The results are summarized in Table 3, where the pathways are divided into four categories based on their assignment score. Note that the majority of the pathways is assigned a high positive score > 4, indicating strong correlation between the expression profiles of members in these pathways, and supporting our very initial assumption. Only a few pathways are assigned negative scores. These are usually short pathways where one gene is highly anti-correlated with the others. It should be noted that both expression datasets generate very similar results. However, since the Rosetta data set is based on many more experiments than the time-series data set, the pairwise expression similarity scores are much more significant, resulting in higher pathway assignment scores (results not shown). In other words, our confidence in the assignments is stronger with the Rosetta dataset (detailed information about assignments is available at [48] and will be later made available at the Biozon website at [41].
Table 3 Distribution of pathway assignment scores. For each data set we ran our algorithm for pathway assignment. The algorithm considers all pathways simultaneously attempting to maximize expression similarity while minimizing the number of conflicts. The final normalized pathway assignment scores Score(A(P)) are divided into four categories based on the average expression similarity of their genes: strongly correlated genes (4 ≤ Score), mildly correlated genes (1 <Score < 4), weakly or uncorrelated genes (-1 ≤ Score ≤ 1) and anti-correlated genes (Score < -1).
Data Set Assignment Score < -1 score -1 ≤ Score ≤ 1 1 <Score < 4 4 ≤ Score
Time-series 4 5 11 32
Rosetta 2 2 7 41
Lastly, it is interesting to compare the assignments before and after resolving conflicts. The pathway relation graph for the 52 pathways contains 10 connected components and 30 singletons. When using the time-series data to assign genes to pathways, we observe conflicts for 9 connected components. The final conflict graph contains 9 connected components and possibly 12 resolvable conflicts (shared edges). 12 of these conflicts are resolved with a small decrease in the assignment score, as is reported in Table 4. Information on the final assignments is given in Table 5. Overall, only a few additional negative pairs are reported after conflicts are resolved, with 32 of the 52 pathways consisting solely of positively scoring pairs (compared to 33 pathways, before conflicts are resolved).
Table 4 Genome wide analysis. Connected components' scores before and after resolving conflicts. For each component we list the names of the constituent pathways, the number of conflicts (shared assignments) and the component score. Note that not all conflicts are solvable. For example, the first connected component contains three pathways, and the best initial assignment results in 6 conflicts. Of these only two are solvable (i.e. there are multiple enzymes that can be assigned to these reactions). The final assignment resolves these conflicts while reducing the score of the connected component only slightly (9.31 compared to 10.22).
Component Number Pathways Number of Conflicts (solvable conflicts) Component score
Before After Before After
1 isoleucine biosynthesis I
valine biosynthesis
leucine biosynthesis 6(2) 4 10.22 9.31
2 aerobic glycerol degradation II
glycolysis 5(3) 2 7.58 7.22
3 asparagine biosynthesis I
glutamate – aspartate pathway
glutamate degradation VI
aspartate biosynthesis II
aspartate biosynthesis and degradation 4(0) 4 6.82 6.82
4 trehalose anabolism
galactose metabolism
UDP-glucose conversion
trehalose biosynthesis 4(1) 3 6.64 6.62
5 pentose phosphate pathway, Mycoplasma pneumoniae
ribose degradation
non-oxidative branch of the pentose phosphate pathway 5(2) 3 5.60 4.78
6 serine biosynthesis
cysteine biosynthesis II 3(0) 3 3.98 3.98
7 glycine biosynthesis I
glycine cleavage
folic acid biosynthesis 3(2) 1 3.41 3.41
8 arginine biosynthesis, Bacillus subtilis
de novo biosynthesis of pyrimidine ribonucleotides 0(0) 0 2.08 2.08
9 alanine degradation 3
alanine biosynthesis II 1(1) 0 0 0
10 phenylalanine biosynthesis I
tyrosine biosynthesis I 2(1) 1 -1.31 -1.33
Table 5 Genome wide analysis. Statistics of the final assignments. For each pathway we list the number of possible assignments, the maximum and minimum scores observed over these assignments, and the final score (note that the final score might not be the maximum score, due to conflicts that were resolved at the refinement stage). The last column gives the number of pairwise relations considered in each assignment, and the number of negative-scoring pairs in the final assignment (in parentheses). Negative scores indicate anti or no correlation. Pathways are sorted by the final assignment score. Note that most pathways are assigned a high positive score, and almost all pairs in the final assignments are positive pairs.
Pathway Number of Assignments Max Score Min Score Final Score Number of pairs (negative Pairs)
pentose phosphate pathway, Mycoplasma pneumoniae 2 11.03 -0.73 11.03 1(0)
sulfate assimilation 2 1 11.02 11.02 11.02 1(0)
methionine and S-adenosylmethionine synthesis 2 10.45 7.34 10.45 1(0)
isoleucine biosynthesis I 12 10.32 3.00 10.32 10(0)
valine biosynthesis 4 10.14 4.99 10.14 6(0)
trehalose biosynthesis 2 10.02 9.65 9.65 1(0)
glutamate degradation I 1 9.64 9.64 9.64 3(0)
arginine biosynthesis I 1 9.58 9.58 9.58 3(0)
chorismate biosynthesis 2 9.24 8.19 9.24 21(0)
glycolysis 180 8.98 3.16 8.87 28(0)
asparagine biosynthesis I 4 8.80 -4.88 8.80 1(0)
trehalose anabolism 8 8.80 0.27 8.80 6(0)
proline biosynthesis I 1 8.43 8.43 8.43 3(0)
galactose metabolism 4 7.91 4.62 7.91 6(0)
glycine degradation III 2 7.73 7.73 7.73 1(0)
methylglyoxal degradation 2 7.59 0.98 7.59 1(0)
tRNA charging pathway 49152 7.41 1.69 7.41 171 (2)
glyoxylate cycle 72 7.37 0.27 7.37 10(0)
homoserine methionine biosynthesis 1 7.33 7.33 7.33 1(0)
pyruvate dehydrogenase 2 6.43 4.33 6.43 1(0)
removal of superoxide radicals 4 5.93 -0.45 5.93 1(0)
aerobic glycerol degradation II 180 6.19 1.64 5.58 28(1)
aspartate biosynthesis II 4 4.85 0.75 4.85 1(0)
non-oxidative branch of the pentose phosphate pathway 8 4.82 0.84 4.82 10(1)
oxidative branch of the pentose phosphate pathway 6 4.78 1.07 4.78 3(0)
arginine biosynthesis, Bacillus subtilis 3 4.55 2.69 4.55 34(4)
leucine biosynthesis 4 10.08 4.01 4.31 3(0)
UDP-N-acetylglucosamine biosynthesis 1 4.22 4.22 4.22 1(0)
cysteine biosynthesis II 2 4.19 0.94 4.19 6(0)
tryptophan biosynthesis 2 4.13 3.99 4.13 10(1)
glutamate biosynthesis I 2 4.08 -4.88 4.08 1(0)
glutathione biosynthesis 1 4.03 4.03 4.03 1(0)
arginine degradation I 1 4.00 4.00 4.00 3(0)
arginine proline degradation 1 3.84 3.84 3.84 3(0)
serine biosynthesis 2 3.58 -0.58 3.58 3(0)
folic acid biosynthesis 48 3.49 0.23 3.49 55 (12)
histidine biosynthesis I 1 2.48 2.48 2.48 12(2)
purine biosynthesis 2 16 2.43 2.01 2.43 90 (27)
homocysteine and cysteine interconversion 2 2.35 2.01 2.35 3(1)
biotin biosynthesis I 1 2.27 2.27 2.27 3(2)
homocysteine degradation I 1 2.01 2.01 2.01 1(0)
glutamate degradation VIII 1 1.86 1.86 1.86 8(2)
homoserine biosynthesis 1 1.14 1.14 1.14 3(1)
threonine biosynthesis from homoserine 1 0.87 0.87 0.87 1(0)
de novo biosynthesis of pyrimidine ribonucleotides 12 0.14 -0.69 0.14 43 (25)
ornithine spermine biosynthesis 2 -0.24 -2.48 -0.24 3(2)
tyrosine biosynthesis I 2 -0.53 -0.58 -0.58 3(1)
glycine biosynthesis I 2 -0.91 -3.60 -0.91 1(1)
UDP-glucose conversion 4 -1.32 -2.04 -1.32 3(2)
ribose degradation 2 8.06 -1.74 -1.74 1(1)
phenylalanine biosynthesis I 2 -2.09 -2.80 -2.09 3(2)
tryptophan kynurenine degradation 1 -2.46 -2.46 -2.46 1(1)
Conclusion
Ongoing sequencing and annotation efforts produce a wealth of data consisting of genes and their products. On the other hand, new types of biological data such as expression and interaction data provide new insights into the mechanisms governing cellular activity. In this light, data integration is necessary in order to accurately analyze the function of genes and other biological entities. The study of biochemical pathways is especially central to these efforts.
Information on cellular pathways is available for several genomes that were studied extensively. However, for most genomes pathway information is not available, what triggered the development of pathway prediction algorithms. Pathway prediction is a difficult problem; Since pathways are not a physical entity, there is no consensus on the definition of a pathway. The pathways that are defined by databases like BioCyc are small subgraphs of a large network of reactions. However, in reality these pathways do not function independently but are rather linked and coordinated with other subnetworks. In an attempt to understand the processes involving metabolism the network has been traditionally divided into smaller subnetworks that can be associated with specific functions. These subnetworks can be considered as the building blocks of the metabolic network and the whole network can be partially reconstructed by integrating the metabolic knowledge contained in these pathways.
In attempt to extrapolate metabolic pathways from one organism to another, several studies developed procedures for assigning genes to pathways. However these procedures ambiguously assign genes to pathways as they usually rely solely on the enzyme class of genes and therefore assign each gene to all the pathways that contain the reactions it can catalyze.
In this paper we present an algorithm for accurate assignment of genes to pathways that attempts to eliminate this ambiguity. For this task, our algorithm utilizes expression data. It has been argued that the metabolic network is co-expressed locally, and an enzyme is co-expressed with the genes catalyzing reactions upstream and downstream of the reaction it catalyzes. We further assume that for the most part pathways are local neighborhoods in the metabolic network and therefore genes assigned to each pathway tend to be co-expressed. Based on this premise, our algorithm assigns genes by maximizing the co-expression of genes that participate in the same pathway. Our algorithm addresses the assignment problem on a genome level, by simultaneously optimizing the co-expression scores for multiple pathways while minimizing the number of conflicts (genes that are shared between different pathways). Our assumption is that if there are multiple genes that can catalyze the same reaction, and that reaction is used in multiple pathways, then each gene is optimized for a different pathway. Conflicts that are detected after initial assignment are reconsidered and our algorithm proceeds by refining the assignment of genes to pathways within connected components in the pathway conflict graph.
Our tests show that for most pathways it is possible to identify a group of genes that can catalyze the pathway reactions and are similarly expressed. Our algorithm can find the most probable assignment of specific genes for each pathway, detect erroneous assignments and suggest control mechanisms of pathways, given a specific expression dataset. The algorithm tackles also the special case of multi-functional enzymes. Since it is difficult to analyze the global network, an alternative approach to detecting pathways of prescribed functions is to search for subnetworks or local neighborhoods in the metabolic network that consist of co-expressed genes, regardless of pathway blueprints. Finding the co-expressed subnetworks of a metabolic network is the methodology of [49] and other studies (as discussed on 'Related Studies' in the paper). However, while this assumption is valid in general it does not always hold and unfortunately these co-expressed subnetworks do not necessarily correspond or overlap with known pathways (as is also indicated by some of our examples). This discrepancy makes it difficult to assess and compare pathway prediction algorithms.
The manually curated pathways that are stored in databases such as MetaCyc and SGD provide an excellent benchmark and perhaps the most accurate reflection of the existing biochemical knowledge, as of today. Our goal is extrapolate that knowledge when predicting pathways in organisms that haven't been studied so far and refine procedures that rely on pathway blueprints and use just EC numbers. Since our algorithm does not rely on manual analysis, it can be most successfully applied to the genomes of organisms that have not fully characterized, once expression data for these genomes becomes available. With the pace in which new genomes are revealed it would be impossible to peruse manual analysis for all and the need for automated procedures becomes evident. The examples we provided prove the effectiveness of our method.
While our algorithm makes advances in the field of pathway prediction it is also faced with several problems. For example, when isozymes are similarly expressed our method picks the best assignment (given the expression data) and only one isozyme is associated with every reaction. However, in some cases multiple isozymes might participate in the same pathway in response to slightly different conditions and substrates. Future versions of our algorithm will handle these cases and estimate the affinity of each isozyme to each pathway that contains the corresponding reaction.
A host of other problems add to the ambiguity of gene-to-pathway assignments, not all them can be addressed with expression data. For example, some enzymes have low specificity and can accept diverse substrates and therefore participate in several different reactions. On the other hand an EC number might specify not a single reaction but rather a class of reactions having common characteristics. One such example is the alcohol dehydrogenases which oxidize a variety of alcohols. The corresponding EC number 1.1.1.1 represents the class of reactions in which either a primary or a secondary alcohol is oxidized, and all alcohol dehydrogenases are annotated with the EC number 1.1.1.1. In yeast there are 6 enzymes annotated with 1.1.1.1. These are the genes ADH1, ADH2, ADH3, ADH4, ADH5 and SFA1. All ADH genes can catalyze the reactions reducing the aldehydes indole acetaldehyde, phenylacetaldehyde and acetaldehyde into the respective alcohols (indole-3-ethanol, phenylethanol and ethanol). However, SFA1 takes as substrate only indole-3-ethanol and phenylethanol. Therefore, as this example demonstrates, EC numbers might not be specific enough, and even database annotations may not be sufficient to differentiate between the different functions of these enzymes.
Our method uses a collection of data sets, including pathways, expression data and statistical models of protein families. We intend to augment these data sets with other relevant biological data sets. For example, integration of interaction data and regulator-regulatee data is necessary in order to predict the global structure of pathways correctly in situations as the one described in 'Discussion' for the isoleucine biosynthesis pathway. Future versions of our algorithm will also account for the topology of the network within pathways and the subcellular location of genes. Other future enhancements include better methods for prediction of enzyme domain families from sequence, to detect new candidates for assignments (thus improving the accuracy of our method) and better mapping procedures from protein annotations to reactions. It should be noted though that our method can be easily extended to other pathways with non-enzymatic reactions. Finally, we are working on probabilistic algorithms which are based on the Expectation-Maximization algorithm, to predict simultaneously gene functions, the existence of pathways, and gene assignments.
Methods
Data sets
Pathways
As the pathway blueprints we used the set of 468 pathways in the MetaCyc database [6] as of May 2003. This database contains a complete biochemical description of pathways that are observed in different organisms. These descriptions are used as templates when predicting similar pathways in other organisms. We extracted from these descriptions the composition of a pathway as a collection of EC classes. It should be noted that most of the pathways in the MetaCyc database were observed experimentally and are linear as opposed to the pathways in KEGG where a reference pathway might integrate the metabolic information from multiple alternative pathways.
Expression data
We used two different expression data sets. The first is the publicly available cell-cycle data set from the Saccharomyces cerevisiae website [26,40]. From this data set we extracted four time series of synchronized S. cerevisiae cells going through the cell cycle. In our analysis each ORF is represented by an extended expression profile derived by concatenating these time series together. The dimensions of these expression vectors range from 1 to 73. This data set has been normalized by [26] to correct for experimental variation between the different microarrays. The second set is the Rosetta Inpharmatics Yeast compendium data [27]. This data set consists of 300 different conditions, mostly deletion mutants. We refer to this set as the Rosetta data set.
Sequence data
Our sequence data is the set of protein sequences in the Yeast sequence database with a total of 6298 proteins. Almost all (5894 out of 6298) of the ORFs in the expression data sets can be mapped to genes in the Yeast sequence database through the ORF label.
Enzyme families
Each pathway is associated with a set of families, usually a list of enzyme families with their enzyme classification (EC) numbers. To assign proteins to EC families we use a composite non-redundant (NR) database that contains more than 1 million unique sequence entries compiled from more than 20 different databases (the database is available at Biozon [41]). Based on the annotations in these databases, 71,638 proteins can be assigned to one (or more) of 2051 EC families. A total of 70,397 are assigned to a single enzyme family, 1241 are possibly multi-domain proteins with at least two different EC designations, and 498 are ambiguous (or suspicious) in the sense that different databases assign them to different EC families (but within the same level of the EC hierarchy, i.e. the first two levels are identical).
To assign Yeast genes to EC families we match the Yeast sequence database against this composite database. Of the 6298 Yeast genes, 832 can be assigned an EC number, either based on their annotation or the annotation of entries with identical sequences from the other databases. Of these genes, 27 are proteins with multiple enzymatic domains.
Predicted EC membership
We extend the set of enzymes by creating a model (sequence profile) for each EC family. The profiles are generated by first grouping proteins with the same, known EC designation from the Biozon NR database. For each EC family we then use an iterative PSI-BLAST procedure [42] to generate a profile. It should be noted that in most cases several profiles are needed to cover all members of the protein family. This is because of the large sequence diversity observed in enzyme families, some of which are composed of several subfamilies that do not exhibit any apparent sequence similarity [22]. Of the 2051 EC families, 597 are composed of more than one subfamily. These models are searched against the Yeast genome, and all genes that are detected as similar with evalue < 0.001 are assigned to the corresponding family, with a confidence value that depends on the evalue.
Metrics
In a previous study [43] we analyzed and assessed the sensitivity and accuracy of different measures of similarity between expression profiles. The measures were assessed in terms of their ability to detect functional links between genes, such as protein-protein interactions, pathway membership, promoter co-regulation, and sequence homology. Our analysis showed that the z-score based measure that combines the Pearson correlation and the Euclidean metric has the maximal information content. Formally, given two expression vectors V and U of dimension d, denote by Dist(V, U) the normalized Euclidean metric
and denote by Corr(V, U) the Pearson correlation of the two vectors
The two distance measures are converted to zscores based on the permutation method described in [43]. This method provides reliable measure of significance as it adjusts to the "compositions" of the vectors compared. The zscores are then summed to determine the final similarity score. Since higher correlation scores are assigned positive zscores, and smaller Euclidean distances are assigned negative zscores, the final score is defined as
sim(V, U) = Z [Corr(V, U)] - Z [Dist(V, U)]
with higher scores indicating stronger similarity. For analysis and performance evaluation see [43]
Appendix – Related work
Metabolic processes make up a substantial part of the cell's activity, and therefore much of the research on pathways so far focused on creating new databases for metabolic pathways as well as extrapolating the known biochemical information from one organism to other. The goal of this research goes beyond just storing, analyzing and extrapolating the metabolic information and strives to improve the known data by discovering variations to pathways in different organisms as well as to discover novel pathways. In this section we review the literature on the main pathway databases and metabolic pathway reconstruction methods and especially methods that use microarray expression data to analyze pathways.
Pathway databases
Most pathway databases were created by compiling metabolic information from different literature sources. Among the first such databases was the Enzymes and Metabolic Pathways (EMP) database [1,50]. It contained information about enzymes and metabolic pathways from over 10000 journal articles, and as of 1996 it stored 2180 pathways from about 1400 organisms. This database was later replaced by the Metabolic Pathways Database (EMP/MPW) [2]. The latter was used as the reference database for metabolic pathway prediction in the WIT ("What Is There?") system [3]. This collection contains 2800 pathway diagrams and their logical structure is encoded in terms used for electronic circuits. Another metabolic database is KEGG [5,51,52]. This database is represented as a graph structure based on binary relations between data items [53]. The pathway database consists of more than 200 reference diagrams taken from the biochemical charts that represent all known realizations of a pathway. The database has three parts: the pathway part, the genes part and the reaction and compound part [54]. BioCyc [55] is composed of a family of databases called Pathway Genome databases (PGDB) where each one is centered around a specific genome. The exception is MetaCyc [6,56] that contains over 491 pathways from multiple organisms.
It is also worth mentioning The University of Minnesota Biocatalysis/Biodegradation Database (UM-BBD) that specializes microbial catabolic metabolism of xenobiotic organic compounds [4,57]. Other pathway databases include aMAZE [58], NCGR PathDB [59], ExPASy – Biochemical Pathways [60] and Biocarta [61].
Pathway prediction based on pathway blueprints
One approach for pathway prediction/recovery in a new organism is based on associating genes that encode enzymes with blueprints of pathways collected either from biochemical charts or from actually observed pathways in different organisms. For example, in order to predict pathways in new genomes, WIT matches the identified enzymes in that genomes with the pathway diagrams from the MPW database [3]. Recently WIT was extended in systems like PUMA2, SEED and ERGO. For example, PUMA2 [62] uses comparative evolutionary analysis of genomes and MPW pathways to extrapolate pathways to new genomes. SEED [8] is an open source system for manual genome annotation where experts can annotate a specific subsystem in multiple genomes at once. It is built around the notion of a "molecular subsystem" which is a collection of functional roles that together fulfill a function. ERGO [7] is a private domain tool that is based on similar principles, and integrates different types of data such as genomic data, regulatory data and essentiality data. No details are available as for the procedures that are used for functional annotation or pathway reconstruction. KEGG matches enzymes to the reference pathways and depending on the degree of completion it assumes that the pathway exists or not [63]. PathFinder [64] is a system that predicts and visualize pathways using the KEGG pathway blueprints using a similar methodology. In UM-BBD, biodegradation pathways of chemical compounds are predicted using a knowledge-based system that matches the compound to a set of biotransformation rules [65]. A biotransformation rule is composed of a sequence of biotransformation functions that transform a compound into its products. The prediction is completed when the resulting compound can no longer be transformed using the rules in the knowledge base or it is one of the termination compounds. In BioCyc, the MetaCyc database is used as the blueprint for the pathway prediction software Pathologic [46,66] which matches enzyme coding genes in a specific genome to reactions in known pathways (but unlike KEGG, they do not re-annotate genes but rather use only existing annotations). Applying Pathologic on a genome results in the creation of a computationally derived PGDB. After creation, a PGDB is curated by mining the literature and new pathways are studied and added to the database. The curated pathways are integrated into MetaCyc to improve the diversity of the database.
All these programs try to address also the problem of finding missing enzymes either by considering alternative reactions or by looking for similar proteins based on sequence similarity or using machine learning models [20-22,24].
Reconstructing pathways from metabolic networks
Reconstructing pathways from metabolic networks is an emerging direction in pathway prediction that does not use the previously known pathway blueprints. This approach uses existing knowledge on reactions and enzymes and chemical rules to create a complete graph of a possible metabolic network, where pathways are defined as sequences of reactions that transform a metabolite into another. For example, in [9] each metabolite is considered a state and a reaction is considered as a transformation from one state to another. The reactions are compiled from the KEGG Ligand database [54]. This state space is searched heuristically for pathways that link metabolites using the A* algorithm with a cost function that is based on the chemical efficiency of the pathway. Similarly, in [10] the metabolic information is structured as a directed graph with two types of nodes: participants (substrates, enzymes) and events (reactions), and edges link reactions to their constituents. Reactions are weighted with the probability that an enzyme catalyzing this reaction exists in the input genome, using sequence similarity. This graph is then searched for maximally weighted pathways using a depth-first strategy. A similar graph is built in [11], who assert pathways from clusters of co-regulated genes that correspond to connected subgraphs. In [12] the authors represent the metabolic information as Petri nets which are bipartite graphs where nodes are of two types: places and transitions. Reactions correspond to transitions and metabolites to places. Pathways are then generated as sequences of transitions in these Petri nets. Related to the prediction of pathways is the analysis of the topological properties of metabolic networks [67]. They shows that metabolic networks from different organisms have the same scaling properties. Furthermore these networks comply with the design principles of scale-free networks.
Similar principles were used in several studies that constructed genome-wide metabolic networks for organisms such as Escherichia coli [13-15], Haemophilus infiuenzae [16], Helicobacter pylori [68], and Saccharomyces cerevisiae [17,18]. In contrast to the automatic methods described above, these studies were based on manual analysis of multiple data sources and mostly the literature. Though time-consuming and expensive, manual analysis is also more accurate and the constructed networks enabled realistic simulations of metabolic networks. For example, in [17] the Saccharomyces cerevisiae metabolic network is reconstructed and its basic features are analyzed. The information was compiled from databases such as KEGG, YPD, SGD and the literature and was augmented with manual functional annotations. The pathways in the model are compartmentalized between cytosol, mitochondria and extra-cellular, and transport steps are added to the model. Extended information about the reactions such as stoichiometry, reversibility and cofactors is also added to the model in order to facilitate the analysis later on. This model is extensively analyzed in [69] and the phenotype of yeast is simulated using a procedure that considered stoichiometric, thermodynamic and reaction capacity constraints. They tested the effect of gene loss and different growth media on the network viability. Most of the simulations were in agreement with the experimental data. In [18] the model is extended by fully compartmentalizing the metabolic reactions by adding five more cellular locations to the model and revising functional assignments for gene products. They also refine the definitions of reactions to include factors such as mass conservation and charge balance. Their results were quite consistent with the experimental data.
Expression data and pathway prediction
Another approach to pathway prediction is based on the analysis of expression data. The main idea behind this approach is that genes participating in the same cellular process are functionally interconnected and this interconnection can be induced from expression data by clustering (e.g. [70]). For example, in [31] the authors cluster genes using expression data, and if multiple genes from a cluster belong to a certain pathway they infer that the other members of the cluster might also belong there. Clustering is also used in [32]. The authors define a distance function between enzyme coding genes that is a combination of the distance between the two reactions they catalyze in the pathway reaction graph, and the correlation-distance between their expression profiles. Similarly, in [33] the expression data and the metabolic information is encoded into two kernel functions and canonical correlation analysis is used to search for correlations between pathways and expression data and therefore identify active pathways. The work is extended in [71], by including a kernel function based on protein-protein interactions. An approach for filing holes in pathways based on expression data is presented in [23]. In this paper a scoring function based on a distance function between expression profiles and the topology of the metabolic network is used to score candidate genes.
Module discovery from expression data is another approach to pathway prediction related to clustering. The assumption is that each cellular process is a module involving multiple genes that are co-regulated and hence are co-expressed. Moreover, the same gene may participate in more than one process (module) and therefore each process accounts for a fraction of the genes expression at a particular measurement. In [72], a probabilistic relational model for each processes is defined and an algorithm to train it is introduced. A similar model based on combined expression data and protein-protein interaction data is developed in [34]. A model for the discovery of transcriptional modules and their common binding site motifs as well as the learning algorithm is developed in [73]. The work is extended to co-regulated gene modules and their regulation program (a small common set of regulators) in [74]. In that work the modules were considered disjoint but in [75] a new model is developed which considers overlapping processes and tries to find the regulation program for each process. All the above models are probabilistic graphical models that employ EM like learning methods.
A different approach is taken in [49], who focus on metabolites as the driving force behind the evolution of metabolic regulation. They search for metabolites around which the most significant transcriptional changes occur (as measured by the expression data of the genes that catalyze reactions in which this metabolite is involved) and identify significantly correlated subnetworks of enzymes. In [25], the authors study regulation in metabolic networks and construct a hierarchy of pathways based on their mutual correlation as measured by expression data. The authors also suggest that correlation in expression profiles is an indication of linear pathways that consist of sequences of reactions. Different isozymes might be independently co-regulated with different groups of genes and therefore might be used to switch between the alternate routes or in the differential regulation of reactions that are shared between different pathways. Expression data was used not only for pathway prediction but also in pathway analysis. Efforts for integrating expression data with metabolic information started by trying to visualize the expression data on top of the pathways diagrams. In Pathway Processor [36] the system tries to assess the probability that the expression of a large number of genes in any given pathway is significantly changed in a given experiment and each pathway is scored using this probability. Similarly, MAPPFinder [76] annotates the GO hierarchy with expression data. The method first associate the GO terms with genes and then calculates the percentage of the genes that meet a user specified criterion. A zscore is computed in order to quantify the significance of the obtained percentage. PathMAPA [77] is a system that visualizes metabolic pathways and expression data in Arabidopsis Thaliana, where pathways are represented in terms of enzymes annotated with EC numbers. The tool estimates the significance of a pathway being up regulated or down regulated in a given experiment. In [35] the authors suggest three functions to score pathways: based on the activity of the genes in the pathway, co-regulations of the genes and the topology of the pathway. The method is then applied to putative pathways in the KEGG database in order to asses the biological significance of these pathways. In [37] the authors present a scoring method for classes of genes. These classes are based on Gene Ontology classification and the scoring is based on expression data. Three types of scores are proposed: co-expression of genes in the same class, statistical significance of gene expression changes, and the learnability of the classification. The scores are converted to p-values to assess their statistical significance, in search of classes with significant scores. Similarly, the biological significance of the pathways asserted in [12] (see previous subsection) is computed by using a scoring function based on expression data in [30]. They score both pathways and genes using two different types of scores (conspicuousness of the expression profile and the synchrony), and the scores are used to asses which are the most probable pathways. Another pathway scoring approach was developed in [38], in search of active pathways. This approach scores a gene set (the set of genes which catalyze reactions in a pathway) by summing all pairwise similarity of the genes in the set. The score obtained is then transformed to a pvalue. All these approaches are related to our approach. However, our method does not score pathways but rather it scores gene assignments to determine the best assignment and identify alternative assignments. Furthermore, our algorithm is geared toward simultaneous prediction of multiple pathways while minimizing shared assignments.
Authors' contributions
LP implemented the model, ran experiments, compared to other models and analyzed the result sets. GY conceived of the study, designed the model and analyzed the results.
Table 6 The correspondence of genes to Biozon NR identifiers. We refer to genes using their unique and stable Biozon NR identifiers, at [41]. To view an entry with identifier x follow the URL: x.
Gene NR Identifiers
ILV1 005760000068
CHA1 003600000165
YKL218C 003260000219
ILV2 003090000098
ILV6 006870000019
ILV5 003950000069
ILV3 005850000040
BAT1 003930000034
BAT2 003760000122
FOL2 002430000075
FOL1 008640000008
FOL3 004270000071
DFR1 002110001504
MIS1 009750000001
ADE3 009460000003
SHM1 005650000392
SHM2 004690000046
YKL132C 004300000053
MET7 005480000035
AAT1 004510000006 004510000730
AAT2 004320000601 004170000010
ASN2 005720000349 005710000020
ASN1 005720000348 005710000019
Supplementary Material
Additional File 1
Assignments of genes to pathways with the time series dataset. For each pathway we list the 10 highest scoring and the 10 lowest scoring assignments (or all assignments, if the number of assignments is 100 or less).
Click here for file
Additional File 2
Assignments of genes to pathways with the rosetta dataset. For each pathway we list the 10 highest scoring and the 10 lowest scoring assignments (or all assignments, if the number of assignments is 100 or less).
Click here for file
Acknowledgements
The authors thank Eurie Hong from SGD for providing us with the SGD pathway data, and William Dirks for help with expression data analysis. We also thank the reviewers for their invaluable comments. This work is supported by the National Science Foundation under Grant No. 0218521 to Golan Yona.
==== Refs
Selkov E Galimova M Goryanin I Gretchkin Y Ivanova N Komarov Y Maltsev N Mikhailova N Nenashev V Overbeek R Panyushkina E Pronevitch L Selkov JE The metabolic pathway collection: an update Nucleic Acids Res 1997 25 37 38 9016500
Selkov JE Grechkin Y Mikhailova N Selkov E MPW: the Metabolic Pathways Database Nucleic Acids Res 1998 26 43 45 9407141
Overbeek R Larsen N Pusch GD D'Souza M Jr ES Kyrpides N Fonstein M Maltsev N Selkov E WIT: integrated system for high-throughput genome sequence analysis and metabolic reconstruction Nucleic Acids Res 2000 28 123 125 10592199
Ellis LBM Hou BK Kang W Wackett LP The University of Minnesota Biocatalysis/Biodegradation Database: post-genomic data mining Nucleic Acids Res 2003 31 262 265 12519997
Kanehisa M Goto S Kawashima S Okuno Y Hattori M The KEGG resource for deciphering the genome Nucleic Acids Res 2004 D277 280 14681412
Krieger CJ Zhang P Mueller LA Wang A Paley S Arnaud M Pick J Rhee SY Karp PD MetaCyc: a multiorganism database of metabolic pathways and enzymes Nucleic Acids Res 2004 D438 442 14681452
Overbeek R Larsen N Walunas T D'Souza M Pusch G Selkov EugeneJ Liolios K Joukov V Kaznadzey D Anderson I Bhattacharyya A Burd H Gardner W Hanke P Kapatral V Mikhailova N Vasieva O Osterman A Vonstein V Fonstein M Ivanova N Kyrpides N The ERGO genome analysis and discovery system Nucleic Acids Res 2003 31 164 171 12519973
The SEED: an Annotation/Analysis Tool Provided by FIG
McShan D Rao S Shah I PathMiner: predicting metabolic pathways by heuristic search Bioinformatics 2003 19 1692 1698 12967966
Sirava M Schafer T Eiglsperger M Kaufmann M Kohlbacher O Bornberg-Bauer E Lenhof HP BioMiner–modeling, analyzing, and visualizing biochemical pathways and networks Bioinformatics 2002 18 S219 230 12386006
van Helden J Gilbert D Wernisch L Schroeder S Mand Wodak Applications of regulatory sequence analysis and metabolic network analysis to the interpretation of gene expression data Lecture Notes in Computer Sciences 2001 2066 155 172
Kuffner R Zimmer R Lengauer T Pathway analysis in metabolic databases via differential metabolic display (DMD) Bioinformatics 2000 16 825 836 11108705
Pramanik J Keasling J Stoichiometric model of Escherichia coli metabolism: Incorporation of growth-rate dependent biomass composition and mechanistic energy requirements Biotechnol Bioengg 1997 56 398 421
Edwards JS Palsson BO The Escherichia coli MG1655 in silico metabolic genotype: Its definition, characteristics, and capabilities Proc Natl Acad Sci USA 2000 97 5528 5533 10805808
Reed J Vo T Schilling C Palsson B An expanded genome-scale model of Escherichia coli K-12 (iJR904 GSM/GPR) Genome Biol 2003 4 R54 12952533
Edwards J Palsson B Systems Properties of the Haemophilus influenzae Rd Metabolic Genotype J Biol Chem 1999 274 17410 17416 10364169
Forster J Famili I Fu P Palsson BO Nielsen J Genome-Scale Reconstruction of the Saccharomyces cerevisiae Metabolic Network Genome Res 2003 13 244 253 12566402
Duarte NC Herrgard MJ Palsson BO Reconstruction and Validation of Saccharomyces cerevisiae iND750, a Fully Compartmentalized Genome-Scale Metabolic Model Genome Res 2004 14 1298 1309 15197165
Enzyme Nomenclature
Bono H Goto S Fujibuchi W Ogata H Kanehisa M Systematic Prediction of Orthologous Units of Genes in the Complete Genomes Genome Inform Ser Workshop Genome Inform 1998 9 32 40 11072319
Shah I Predicting enzyme function from sequence PhD thesis 1999 George Mason University
Syed U Yona G Using a mixture of probabilistic decision trees for direct prediction of protein function Proceedings of the seventh annual international conference on Computational molecular biology 2003 ACM Press 289 300
Kharchenko P Vitkup D Church GM Filling gaps in a metabolic network using expression information Bioinformatics 2004 20 il78 185 15262797
Green M Karp P A Bayesian method for identifying missing enzymes in predicted metabolic pathway databases BMC Bioinformatics 2004 5 76 15189570
Ihmels J Levy R Barkai N Principles of transcriptional control in the metabolic network of Saccharomyces cerevisiae Nat Biotechnol 2004 22 86 92 14647306
Spellman PT Sherlock G Zhang MQ Iyer VR Anders K Eisen MB Brown PO Botstein D Futcher B Comprehensive Identification of Cell Cycle-regulated Genes of the Yeast Saccharomyces cerevisiae by Microarray Hybridization Mol Biol Cell 1998 9 3273 3297 9843569
Hughes TR Marton MJ Jones AR Roberts CJ Stoughton R Armour CD Bennett HA Coffey E Dai H He YD Kidd MJ King AM Meyer MR Slade D Lum PY Stepaniants SB Shoemaker DD Gachotte D Chakraburtty K Simon J Bard M Friend SH Functional Discovery via a Compendium of Expression Profiles Cell 2000 102 109 126 10929718
McCormick SM Frye SR Eskin SG Teng CL Lu CM Russell CG Chittur KK McIntire LV Microarray analysis of shear stressed endothelial cells Biorheology 2003 40 5 11 12454381
Yoo MS Chun HS Son JJ DeGiorgio LA Kim DJ Peng C Son JH Oxidative stress regulated genes in nigral dopaminergic neuronal cells: Correlation with the known pathology in Parkinsons disease Brain Res Mol Brain Res 2003 110 76 84 12573535
Zien A Küffner R Zimmer R Lengauer T Altman R Analysis of Gene Expression Data with Pathway Scores Proc Int Conf Intell Syst Mol Biol 2000 La Jolla, CA: AAAI 407 417 10977101
Nakao M Bono H Kawashima S Kamiya T Sato K Goto S Kanehisa M Genome-scale Gene Expression Analysis and Pathway Reconstruction in KEGG Genome Inform Ser Workshop Genome Inform 1999 10 94 103 11072346
Hanisch D Zien A Zimmer R Lengauer T Co-clustering of biological networks and gene expression data Bioinformatics 2002 18 S145 154 12169542
Vert JP Kanehisa M Extracting active pathways from gene expression data Bioinformatics 2003 19 II238 II244 14534196
Segal E Wang H Koller D Discovering molecular pathways from protein interaction and gene expression data Bioinformatics 2003 19 i264 272 12855469
Kurhekar M Adak S Jhunjhunwala S Raghupathy K Genome-Wide Pathway Analysis And Visualization Using Gene Expression Data Pac Symp Biocomput 2002 462 473 11928499
Grosu P Townsend JP Hartl DL Cavalieri D Pathway Processor: A Tool for Integrating Whole-Genome Expression Results into Metabolic Networks Genome Res 2002 12 1121 1126 12097350
Pavlidis P Lewis D Noble W Exploring gene expression data with class scores Pac Symp Biocomput 2002 474 485 11928500
Rahnenfuhrer J Domingues FS Maydt J Lengauer T Calculating the Statistical Significance of Changes in Pathway Activity From Gene Expression Data Statistical Applications in Genetics and Molecular Biology 2004 3
Stenesh J Dictionary of Biochemistry and Molecular Biology 1989 2 John Wiley & Sons
Cho RJ Campbell MJ Winzeler EA Steinmetz L Conway A Wodicka L Tyra G Wolfsberg AEG Landsman D Lockhart DJ Davis RW A Genome-Wide Transcriptional Analysis of the Mitotic Cell Cycle Mol Cell 1998 2 65 73 9702192
Birkland A Yona G BIOZON: a system for unification, management and analysis of heterogeneous biological data
Altschul S Madden T Schaffer A Zhang J Zhang Z Miller W Lipman D Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694
Dirks W Yona G A comprehensive study of the notion of functional link between genes based on microarray data, promoter signals, protein-protein interactions and pathway analysis 2003
Gene Function, Metabolic Pathways and Comparatiev Genomics in Yeast 2003 IEEE Computer Society
Christie KR Weng S Balakrishnan R Costanzo MC Dolinski K Dwight SS Engel SR Feierbach B Fisk DG Hirschman JE Hong EL Issel-Tarver L Nash R Sethuraman A Starr B Theesfeld CL Andrada R Binkley G Dong Q Lane C Schroeder M Botstein D Cherry JM Saccharomyces Genome Database (SGD) provides tools to identify and analyze sequences from Saccharomyces cerevisiae and related sequences from other organisms Nucleic Acids Res 2004 D311 314 14681421
Karp PD Paley S Romero P The Pathway Tools software Bioinformatics 2002 18 S225 S232 12169551
Miklos G Rubin G The role of the Genome Project in determining gene function: insights from model organisms Cell 1996 86 521 529 8752207
Supplemental information
Patil KR Nielsen J Uncovering transcriptional regulation of metabolism by using metabolic network topology Proc Natl Acad Sci USA 2005 102 2685 2689 15710883
Selkov E Basmanova S Gaasterland T Goryanin I Gretchkin Y Maltsev N Nenashev V Overbeek R Panyushkina E Pronevitch L Selkov JE Yunus I The metabolic pathway collection from EMP: the enzymes and metabolic pathways database Nucleic Acids Res 1996 24 26 28 8594593
Kanehisa M Goto S KEGG: Kyoto Encyclopedia of Genes and Genomes Nucleic Acids Res 2000 28 27 30 10592173
Kanehisa M Goto S Kawashima S Nakaya A The KEGG databases at GenomeNet Nucleic Acids Res 2002 30 42 46 11752249
Goto S Bono H Ogata H Fujibuchi W Nishioka T Sato K Kanehisa M Organizing and computing metabolic pathway data in terms of binary relations Pac Symp Biocomput 1997 9390290
Goto S Okuno Y Hattori M Nishioka T Kanehisa M LIGAND: database of chemical compounds and reactions in biological pathways Nucleic Acids Res 2002 30 402 404 11752349
BioCyc Database
Karp PD Riley M Paley SM Pellegrini-Toole A The MetaCyc Database Nucleic Acids Res 2002 30 59 61 11752254
Ellis LBM Hershberger CD Wackett LP The University of Minnesota Biocatalysis/Biodegradation Database: microorganisms, genomics and prediction Nucleic Acids Res 2000 28 377 379 10592280
Lemer C Antezana E Couche F Fays F Santolaria X Janky R Deville Y Richelle J Wodak SJ The aMAZE LightBench: a web interface to a relational database of cellular processes Nucleic Acids Res 2004 D443 448 14681453
PATHDB :: Biochemical Pathways
ExPASy Biochemical Pathways
Biocarta
PUMA2 – Evolutionary Analysis of Metabolism
Bono H Ogata H Goto S Kanehisa M Reconstruction of Amino Acid Biosynthesis Pathways from the Complete Genome Sequence Genome Res 1998 8 203 210 9521924
Goesmann A Haubrock M Meyer F Kalinowski J Giegerich R PathFinder: reconstruction and dynamic visualization of metabolic pathways Bioinformatics 2002 18 124 129 11836220
Hou BK Wackett LP Ellis LBM Microbial Pathway Prediction: A Functional Group Approach J Chem Inf Comput Sci 2003 43 1051 1057 12767164
Paley SM Karp PD Evaluation of computational metabolic-pathway predictions for Helicobacter pylori Bioinformatics 2002 18 715 724 12050068
Jeong H Tombor B Albert R Oltvai Z Barabasi A The Large-Scale Organization of Metabolic Networks Nature 2000 407 651 654 11034217
Schilling CH Covert MW Famili I Church GM Edwards JS Palsson BO Genome-Scale Metabolic Model of Helicobacter pylori 26695 J Bacteriol 2002 184 4582 4593 12142428
Famili I Forster J Nielsen J Palsson BO Saccharomyces cerevisiae phenotypes can be predicted by using constraint-based analysis of a genome-scale reconstructed metabolic network Proc Natl Acad Sci USA 2003 100 13134 13139 14578455
Eisen MB Spellman PT Brown PO Botstein D Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci USA 1998 95 14863 14868 9843981
Yamanishi Y Vert JP Nakaya A Kanehisa M Extraction of correlated gene clusters from multiple genomic data by generalized kernel canonical correlation analysis Bioinformatics 2003 19 i323 330 12855477
Segal E Battle A Koller D Decomposing gene expression into cellular processes Pac Symp Biocomput, Kaua'i 2003 89 100
Segal E Yelensky R Koller D Genome-wide discovery of transcriptional modules from DNA sequence and gene expression Bioinformatics 2003 19 i273 282 12855470
Segal E Shapira M Regev A Pe'er D Botstein D Kollerl D Friedman N Module networks: identifying regulatory modules and their condition-specific regulators from gene expression data Nat Genet 2003 34 166 176 12740579
Battle A Segal E Koller D Probabilistic discovery of overlapping cellular processes and their regulation Proc 8th Inter Conf on Research in Computational Molecular Biology (RECOMB) 2004 New York, NY, USA: ACM Press 167 176
Doniger S Salomonis N Dahlquist K Vranizan K Lawlor S Conklin B MAPPFinder: using Gene Ontology and GenMAPP to create a global gene-expression profile from microarray data Genome Biol 2003 4 R7 12540299
Pan D Sun N Cheung KH Guan Z Ma L Holford M Deng X Zhao H PathMAPA: a tool for displaying gene expression and performing statistical tests on metabolic pathways at multiple levels for Arabidopsis BMC Bioinformatics 2003 4 56 14604444
in silico Organisms – Saccharomyces cerevisiae (baker's yeast)
|
16135255
|
PMC1239907
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Aug 31; 6:217
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-217
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2201614657510.1186/1471-2105-6-220Methodology ArticleA method for finding single-nucleotide polymorphisms with allele frequencies in sequences of deep coverage Wang Jianmin [email protected] Xiaoqiu [email protected] Department of Computer Science, Iowa State University, Ames, Iowa 50011, USA2005 7 9 2005 6 220 220 1 2 2005 7 9 2005 Copyright © 2005 Wang and Huang; licensee BioMed Central Ltd.2005Wang and Huang; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The allele frequencies of single-nucleotide polymorphisms (SNPs) are needed to select an optimal subset of common SNPs for use in association studies. Sequence-based methods for finding SNPs with allele frequencies may need to handle thousands of sequences from the same genome location (sequences of deep coverage).
Results
We describe a computational method for finding common SNPs with allele frequencies in single-pass sequences of deep coverage. The method enhances a widely used program named PolyBayes in several aspects. We present results from our method and PolyBayes on eighteen data sets of human expressed sequence tags (ESTs) with deep coverage. The results indicate that our method used almost all single-pass sequences in computation of the allele frequencies of SNPs.
Conclusion
The new method is able to handle single-pass sequences of deep coverage efficiently. Our work shows that it is possible to analyze sequences of deep coverage by using pairwise alignments of the sequences with the finished genome sequence, instead of multiple sequence alignments.
==== Body
Background
Information concerning the allele frequencies of single-nucleotide polymorphisms (SNPs) is needed to select an optimal subset of common SNPs for use in association studies [1]. One approach to finding common SNPs with allele frequencies is to generate DNA sequences from a sufficient number of samples in a population. This approach requires that computational methods have an ability to handle thousands of sequences from the same genome location (sequences of deep coverage). In this paper, we describe a computational method for finding common SNPs with allele frequencies in sequences of deep coverage. We present results from the method on human expressed sequence tags (ESTs) of deep coverage, which are currently a major source of DNA sequences of deep coverage. The method is also expected to be useful for finding common mutations in sequences of deep coverage produced in a cancer genome project [2].
The PolyBayes program is widely used to find SNPs in redundant DNA sequences [3,4]. It first constructs a multiple sequence alignment based on pairwise alignments of each sequence with a high-quality genomic sequence called an anchor. Then it identifies and removes paralogous sequences that have a high number of observed differences with the anchor sequence. Next it computes an SNP probability score for each column of the multiple sequence alignment based on a rigorous Bayesian formula. The formula uses the prior probabilities of all the nucleotide permutations for the column, which are estimated from the quality scores of the bases on the column.
We enhance the PolyBayes program in several aspects to handle single-pass sequences (query sequences) of deep coverage. First, all the paralogous regions of the finished human genome sequence are included as anchor sequences. Each query sequence is assigned to the corresponding anchor sequence that is different from each of the remaining anchor sequences at some positions but is identical to the query sequence at most of the positions. This approach separates paralogous sequences by making use of the positions where paralogous sequences differ but sequences from the same genome location agree.
Second, pairwise alignments of corresponding query and anchor sequences are used to construct profiles, one per anchor sequence. At each position of an anchor sequence, its profile contains the numbers and types of high-quality query bases that are aligned to the position of the anchor sequence. Candidate SNPs are produced based on the profiles, instead of multiple sequence alignments for the following reason. As the number of single-pass sequences in a multiple sequence alignment increases, the number of gap columns in the alignment increases but the number of identity columns in the alignment does not increase. Thus, it is difficult to construct an accurate multiple sequence alignment for single-pass sequences of deep coverage.
Third, because the pairwise alignment of corresponding query and anchor sequences may contain regions of low similarity due to sequencing errors or contaminants, the highly similar regions of the alignment are found by a dynamic programming algorithm. Only the highly similar regions are used in generation of the profile.
Our computer program named PolyFreq was compared with PolyBayes on eighteen data sets of human EST sequences of deep coverage. Results from PolyFreq and PolyBayes indicate that PolyFreq ran to completion and used almost all input sequences in computation of the allele frequencies of SNPs for every data set.
Results
The method for finding SNPs with allele frequencies was implemented as a computer program. The source code of the program is freely available for academic use [5, see Additional file 1]. The program takes as input a set of high-quality anchor sequences and a set of query sequences with quality scores. The set of anchor sequences includes all the paralogous regions of the genome for the set of query sequences. The anchor and query sequences are from the same species.
The program reports candidate SNPs in the anchor sequences. For each candidate SNP, the program reports its position in the anchor sequence, its local context in the anchor sequence, and base types with a frequency greater than a cutoff. The frequency of a base type is also given in a rational form with the number of query bases of the type as the enumerator and the total number of query bases as the denominator.
To evaluate PolyFreq, eighteen data sets of human EST sequences of deep coverage were constructed as follows. Eighteen clusters of human EST sequences, each containing at least 1,000 EST sequences with trace data, were randomly selected from the April, 2005 release of the UniGene database [6]. The eighteen UniGene clusters also contain EST sequences without trace data. For each of the eighteen UniGene clusters, an EST data set was obtained by selecting all EST sequences with trace data from the cluster. The set of quality score sequences for each of the eighteen data sets was produced from the trace data with Phred [7]. The quality score q of a base is obtained by the formula q = -10 log p, where p is the estimated error probability of the base [8]. For example, a quality score of 20 corresponds to an error probability of 0.01. The EST sequences in each of the eighteen sets were produced from 71 to 118 cDNA libraries derived from diverse human tissues and cell lines [9]. Each of the eighteen data sets of full-length EST sequences without any masking was used as a query set.
For each query set, its set of anchor sequences was obtained by comparing the query sequences with the finished genome sequence at the BLAT web server [10]. By using stringent settings for BLAT, a set of two human anchor sequences was produced for each of three query data sets, and a set of one human anchor sequence was produced for each of the remaining query data sets. Each set of anchor sequences was screened for repeats with RepeatMasker [11].
The PolyFreq program was run on each pair of query and anchor sets. The PolyFreq program ran successfully to completion for each of the eighteen data sets. The following values were used for the parameters of the program: 50, minimum depth of coverage for each candidate SNP; 0.1%, minimum minor allele frequency; 5 bp, minimum perfect block length; 20, minimum base quality score in the perfect block; 90%, minimum percent identity for query-anchor alignments; 97%, minimum percent identity for the highly similar regions of query-anchor alignments.
Although PolyBayes was not designed to deal with data sets of deep coverage, we tested PolyBayes on the eighteen data sets of deep coverage to see how PolyBayes would behave on the data sets. Because PolyBayes takes only one anchor sequence, the corresponding anchor sequence was selected and given to PolyBayes for each data set. On eight of the eighteen data sets, PolyBayes ran successfully to completion. On the remaining data sets, PolyBayes terminated abnormally without producing any output file after running for a few hours. The default values for all the parameters but the SNP probability output cutoff of PolyBayes were used; PolyBayes terminated abnormally more frequently under other parameter values. A value of 0.75 for the SNP probability output cutoff was used to produce a lower number of false positives than the default value of 0.5.
The abnormal termination of PolyBayes might be related to the deep coverage of the data set and full-length EST sequences with low-quality ends or contaminants. Thus, for each set of full-length EST sequences, a set of trimmed EST sequences was produced by removing the ends of every sequence that are not highly similar to the corresponding anchor sequence. For each data set, the number of trimmed sequences was close to the number of full-length sequences. The PolyBayes program was also run on each set of trimmed sequences. It ran to completion for thirteen out of the eighteen data sets.
All the tests were performed on a Dell workstation with two 3.0-Ghz processors and 4 Gb of main memory. PolyFreq took less than one hour on every data set, whereas PolyBayes was two to ten times slower than PolyFreq on every data set. The memory requirements of PolyFreq and PolyBayes on the data sets were similar and were about 30 to 40 times the input size.
The PolyFreq and PolyBayes programs were compared on every data set for which PolyBayes ran successfully to completion on either the set of full-length sequences or the set of trimmed sequences. For each data set, results produced by PolyFreq on the set of full-length sequences were included in the comparison, whereas results produced by PolyBayes on both sets of full-length and trimmed sequences were included. The SNPs from the dbSNP database [12] that were mapped by the following method to the anchor sequences were used as true SNPs for the comparison. Each SNP from dbSNP is specified by a local sequence context. For each data set of EST sequences with a RefSeq sequence [13], each SNP from dbSNP that occurs in the RefSeq sequence was determined by finding the exact occurrence of its sequence context in the RefSeq sequence. Each SNP from dbSNP in the RefSeq sequence was mapped to a corresponding anchor sequence by using a spliced alignment of the RefSeq and anchor sequences. Because four data sets had no RefSeq sequence, no SNPs from dbSNP were mapped to the anchor sequences for the data sets.
For each program on every data set with a RefSeq sequence, the number of true positives, the number of false positives, and the number of false negatives were computed. The number of true negatives was not collected because of its large value. Also reported were the number of sequences in the data set and the number of sequences that were used by the program to compute candidate SNPs. The comparison results are shown in Table 1.
Table 1 Results by PolyFreq and PolyBayes on eighteen data sets of EST sequences
Data set Size PolyBayes (trimmed) PolyFreq (full length) PolyBayes (full length)
TP FP FN NSU TP FP FN NSU TP FP FN NSU
Hs.119589 4403 12 170 50 1491 5 24 57 4391 12 152 50 1531
Hs.129673 1665 7 48 11 1457 5 11 13 1662 T/A
Hs.148340 1603 3 36 9 1563 4 9 8 1583 3 67 9 1560
Hs.170622 1514 1 37 12 429 3 18 10 1507 2 73 11 365
Hs.178551 1685 5 62 12 1632 6 14 11 1676 T/A
Hs.180909 1017 3 50 6 983 3 20 6 1012 3 164 6 996
Hs.187199 2041 T/A N/A 1997 T/A
Hs.198281 3156 9 110 22 3077 15 54 16 3149 T/A
Hs.350927 1017 5 42 11 976 9 21 7 1015 6 139 10 995
Hs.356331 1441 2 55 9 318 1 14 10 1436 2 85 9 239
Hs.356572 2822 0 46 2 2534 0 17 2 2821 T/A
Hs.439552 7163 T/A N/A 6873 T/A
Hs.444467 4033 N/A 805 N/A 4028 N/A 679
Hs.446628 1490 4 32 12 1338 5 12 11 1486 T/A
Hs.520640 4120 T/A 9 52 27 4099 T/A
Hs.522463 8294 T/A N/A 8280 T/A
Hs.524390 4462 T/A 9 48 24 4454 T/A
Hs.544577 7537 14 60 43 1716 10 17 47 7517 18 175 39 1556
The mark N/A means that no SNP from dbSNP was mapped to the anchor sequence because of lack of a RefSeq sequence. The mark T/A means that PolyBayes terminated abnormally without producing any output file. A candidate SNP from the program is considered as true positive (TP) if it is in dbSNP or false positive (FP) otherwise. A SNP from dbSNP that occurs in the data set is considered as false negative (FN) if it is not reported as a candidate SNP from the program. The number of sequences used (NSU) by the program in generation of candidate SNPs is reported.
The results in Table 1 indicate that PolyFreq could handle the data sets of full-length reads with problem regions and with very deep coverage. PolyFreq used 1,997 to 8,280 sequences on the five data sets for which PolyBayes terminated abnormally. On the data sets for which PolyBayes ran to completion, PolyFreq was similar to PolyBayes in the number of true positives and the number of false negatives, and is better than PolyBayes in the number of false positives. PolyBayes used significantly fewer sequences than PolyFreq on some of the data sets. Note that the ability to use as many sequences as possible is necessary for accurate computation of the allele frequencies of SNPs.
Discussion
We originally developed a method for assembling sequences of deep coverage. The method constructs multiple sequence alignments of large width for contigs. The method has to deal with a large number of gap columns in the multiple sequence alignment. We later agreed with one of the reviewers that it is not necessary to construct multiple sequence alignments for analysis of sequences of deep coverage. The reviewer also suggested that we focus on SNP analysis in sequences of deep coverage. Those suggestions motivated us to develop the method reported in this paper.
The PolyFreq program keeps PolyBayes' feature of performing comparisons between query and anchor sequences, instead of performing comparisons among query sequences. In addition, PolyFreq constructs profiles by using the highly similar regions of pairwise alignments of corresponding query and anchor sequences, instead of multiple alignments of query and anchor sequences. Thus, the efficiency and accuracy of PolyFreq are not significantly affected by query sequences of deep coverage. On the contrary, PolyFreq can compute the allele frequencies of SNPs more accurately in query sequences of deep coverage.
As sequencing costs are significantly reduced in the future, single-pass sequences from hundreds to thousands of individuals will be produced. Those sequences will be of deep coverage. Our current work suggests that it is possible to analyze sequences of deep coverage by using pairwise alignments of the sequences with the finished genome sequence, instead of multiple sequence alignments.
Methods
We first present the major steps of our method for finding common SNPs with allele frequencies in a set of query sequences and a set of anchor sequences. Then we describe each step in detail. The method consists of the following major steps:
1. Compute an alignment of anchor sequences for each pair of anchor sequences.
2. Compute an alignment of query and anchor sequences for each pair of similar query and anchor sequences.
3. For each query sequence, find the corresponding anchor sequence that is different from each of the remaining anchor sequences at some positions but is identical to the query sequence at most of the positions.
4. Find the highly similar regions of their alignment for each pair of corresponding query and anchor sequences.
5. For each anchor sequence, use the highly similar regions of every alignment involving the anchor sequence to construct a profile for the anchor sequence. At each position of the anchor sequence, its profile contains the numbers and types of high-quality query bases that are aligned to the position of the anchor sequence.
6. Report each candidate SNP with major and minor allele frequencies if its minor allele frequency is greater than a cutoff.
In step 1, for each pair of anchor sequences, an alignment of the sequences in given orientation as well as an alignment of the sequences in opposite orientation is constructed with GAP3, a global alignment program specially designed for genomic sequences with long different regions between similar regions [14]. One of the two alignments with a larger score is saved for the pair of sequences. The alignments saved in this step are to be used in step 3 for finding the corresponding anchor sequence for each query sequence.
In step 2, pairs of similar query and anchor sequences are found with DDS2, which produces a high-scoring chain of segment pairs (ungapped alignment fragments) between the two sequences in the pair [15]. For each pair of similar query and anchor sequences, an alignment of the sequences in the pair is constructed with GAP22, an improved version of the GAP2 program [16] for quickly computing an alignment in a small area of the dynamic programming matrix, which is determined based on the chain of segment pairs. If the percent identity of the alignment is greater than a cutoff, then the alignment is saved for the pair of sequences.
In step 3, for each query sequence that is highly similar to two or more anchor sequences, the corresponding anchor sequence for the query sequence is selected among the anchor sequences through pairwise comparisons. Initially, one anchor sequence is taken as the current leader. Then the rest are compared with the current leader one at a time. Consider the comparison between the current leader and the current challenger. The winner between the two anchor sequences is produced by using the alignment of the two anchor sequences and their alignments with the query sequence. A common match occurs at a position of the query sequence, a position of the current leader, and a position of the current challenger if the three positions are pairwise aligned on each of the three alignments and contain the same base. The winner between the two anchor sequences is the one with a larger number of uncommon matches in its alignment with the query sequence. The winner becomes the current leader. After all the pairwise comparisons, the final leader is the corresponding anchor sequence for the query sequence.
In step 4, for each pair of corresponding query and anchor sequences, the highly similar regions of the alignment of the two sequences are identified in linear time with LCP, a program for finding regions of a sequence that meet a content requirement [17]. Each of the highly similar regions found by LCP has a percent identity greater than or equal to a cutoff p and is strictly optimal. The score of a region of the alignment is the sum of scores of every base match and every base difference in the region, where the score of every base match is 1 - p and the score of every base difference is - p. A region is optimal if its score is not less than the score of any other region that overlaps with it. An optimal region is strictly optimal if it is not completely contained in any optimal region other than itself.
In step 5, only substitutions in the highly similar regions of every alignment of corresponding query and anchor sequences are used to construct a profile for the anchor sequence because the remaining regions of the alignment have a high rate of difference, which is likely due to sequencing errors or contaminants in the query sequence. Additional requirements are introduced below because a long highly similar region may still contain a packet of sequencing errors in the middle. A sufficiently long section in a highly similar region of an alignment is a perfect block if the section consists only of exact base matches and the quality score of each query base in the section is greater than or equal to a cutoff [18]. A substitution in a highly similar region of an alignment is acceptable if it is immediately flanked on each side by a perfect block. Acceptable and unacceptable substitutions are illustrated in Figure 1A.
Figure 1 Acceptable and unacceptable substitutions in a pairwise alignment and a candidate SNP from a real data set. (A) The line shows an alignment of query and anchor sequences with thick parts indicating highly similar regions. The large rectangular box gives a detailed view of the small box. On the left is an unacceptable substitution that is flanked by a block of low-quality bases, and on the right is an acceptable substitution that is flanked by a perfect block on each side. The quality value of each query base in the large box is shown next to the base. (B) Shown is a candidate SNP with allele frequencies from PolyFreq on a real data set (Hs. 119589) in Table 1.
For each anchor sequence, its profile contains four counts at each position: one count for each query base type. For example, the base A count at the anchor position is the number of acceptable substitutions at the anchor position and at a query sequence position with base A, in a highly similar region of an alignment of the anchor sequence with the query sequence. A count for a query base type at the anchor position is 0 if there is no acceptable substitution at the anchor position and at any query sequence position with the query base type. The frequency of each of the four counts is the count divided by the sum of the four counts if the sum is positive.
In step 6, each profile is scanned for candidate SNPs. A candidate SNP occurs at an anchor sequence position if the sum of the four counts for the position is greater than or equal to a cutoff and at least two of the four counts have a frequency greater than a cutoff. All candidate SNPs with allele frequencies are reported along with a local anchor sequence region for each candidate SNP. A candidate SNP with allele frequencies from one of the examples in Table 1 is shown in Figure 1B.
Author's contributions
XH designed the strategy for solving the problem and provided guidance to JW. JW worked out the details of the strategy, developed the program, and produced results on data sets with the program. XH wrote the paper and JW formatted it in Word. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
A file named PolyFreq.tar is included. The file contains the source code of all programs in the package. The file is unpacked by using the Unix command "tar xvf PolyFreq.tar" on a Unix or Linux computer.
Click here for file
Acknowledgements
We thank Geo Pertea and John Quackenbush for discussions on assembly of sequences of deep coverage, and Brian Haas for suggestions on and evaluation of GAP22. We are grateful to the reviewers for suggestions that motivated us to develop the new method.
JW and XH were supported in part by NIH Grants R01 HG01502-05 and R01 HG01676-05 from NHGRI.
==== Refs
Hinds DA Stuve LL Nilsen GB Halperin E Eskin E Ballinger DG Frazer KA Cox DR Whole-genome patterns of common DNA variation in three human populations Science 2005 307 1072 1079 15718463 10.1126/science.1105436
Kaiser J NCI gears up for cancer genome project Science 2005 307 1182 15731412 10.1126/science.307.5713.1182a
Marth GT Korf I Yandell MD Yeh RT Gu Z Zakeri H Stitziel NO Hillier L Kwok P Gish WR A general approach to single-nucleotide polymorphism discovery Nat Genet 1999 23 452 456 10581034 10.1038/70570
The International SNP Map Working Group A map of human genome sequence variation containing 1.42 million single nucleotide polymorphisms Nature 2001 409 928 933 11237013 10.1038/35057149
PolyFreq Program
Wheeler DL Church DM Federhen S Lash AE Madden TL Pontius JU Schuler GD Schriml LM Sequeira E Tatusova TA Wagner L Database resources of the National Center for Biotechnology Nucleic Acids Res 2003 31 28 33 12519941 10.1093/nar/gkg033
Ewing B Hillier L Wendl M Green P Basecalling of automated sequencer traces using phred. I. Accuracy assessment Genome Res 1998 8 175 185 9521921
Ewing B Green P Basecalling of automated sequencer traces using phred. II. Error probabilities Genome Res 1998 8 186 194 9521922
Mammalian Gene Collection (MGC) Program Team Generation and initial analysis of more than 15,000 full-length human and mouse cDNA sequences Proc Natl Acad Sci USA 2002 26 16899 16903 10.1073/pnas.242603899
Kent WJ BLAT – The BLAST-like alignment tool Genome Res 2002 12 656 664 11932250 10.1101/gr.229202. Article published online before March 2002
Smit AFA Hubley R Green P RepeatMasker Open-3.0 1996
dbSNP
Pruitt KD Tatusova T Maglott DR NCBI Reference Sequence (RefSeq): A curated non-redundant sequence database of genomes, transcripts and proteins Nucleic Acids Res 2005 33 D501 D504 15608248 10.1093/nar/gki025
Huang X Chao K-M A generalized global alignment algorithm Bioinformatics 2003 19 228 233 12538243 10.1093/bioinformatics/19.2.228
Huang X Ye L Chou H-H Yang I-H Chao K-M Effcient combination of multiple word models for improved sequence comparison Bioinformatics 2004 20 2529 2533 15117751 10.1093/bioinformatics/bth279
Huang X Adams MD Zhou H Kerlavage AR A tool for analyzing and annotating genomic sequences Genomics 1997 46 37 45 9403056 10.1006/geno.1997.4984
Huang X An algorithm for identifying regions of a DNA sequence that satisfy a content requirement Comput Appl Biosci 1994 10 219 225 7922676
Altshuler D Pollara VJ Cowles CR Van Etten WJ Baldwin J Linton L Lander ES An SNP map of the human genome generated by reduced representation shotgun sequencing Nature 2000 407 513 516 11029002 10.1038/35035083
|
16146575
|
PMC1239908
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Sep 7; 6:220
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-220
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2211614657910.1186/1471-2105-6-221SoftwareGASH: An improved algorithm for maximizing the number of equivalent residues between two protein structures Standley Daron M [email protected] Hiroyuki [email protected] Haruki [email protected] Institute for Protein Research, Osaka University, 3-2 Yamadaoka, Suita, Osaka 565-0871, Japan2 Japan Science and Technology Agency, Institute for Bioinformatics Research and Development (BIRD), Japan3 Division of Bioinformatics, Medical Institute of Bioregulation, Kyushu University, 3-1-1, Maidashi, Higashi-ku, Fukuoka, Fukuoka 812-8582, Japan2005 8 9 2005 6 221 221 27 4 2005 8 9 2005 Copyright © 2005 Standley et al; licensee BioMed Central Ltd.2005Standley et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We introduce GASH, a new, publicly accessible program for structural alignment and superposition. Alignments are scored by the Number of Equivalent Residues (NER), a quantitative measure of structural similarity that can be applied to any structural alignment method. Multiple alignments are optimized by conjugate gradient maximization of the NER score within the genetic algorithm framework. Initial alignments are generated by the program Local ASH, and can be supplemented by alignments from any other program.
Results
We compare GASH to DaliLite, CE, and to our earlier program Global ASH on a difficult test set consisting of 3,102 structure pairs, as well as a smaller set derived from the Fischer-Eisenberg set. The extent of alignment crossover, as well as the completeness of the initial set of alignments are examined. The quality of the superpositions is evaluated both by NER and by the number of aligned residues under three different RMSD cutoffs (2,4, and 6Å). In addition to the numerical assessment, the alignments for several biologically related structural pairs are discussed in detail.
Conclusion
Regardless of which criteria is used to judge the superposition accuracy, GASH achieves the best overall performance, followed by DaliLite, Global ASH, and CE. In terms of CPU usage, DaliLite CE and GASH perform similarly for query proteins under 500 residues, but for larger proteins DaliLite is faster than GASH or CE. Both an http interface and a simple object application protocol (SOAP) interface to the GASH program are available at .
==== Body
Background
The coordinates of over 30,761 protein structures are currently available at the Protein Data Bank (PDB [1]), and each year thousands of new structures are deposited. A quantitative analysis of this data requires accurate tools for superimposing protein structures, measuring their similarity, and identifying structurally equivalent residues. However, unlike sequence analysis, there is no universally accepted measure of structural similarity. Moreover, even if such a measure existed, structure alignment is so much more complex than sequence alignment, that none the most popular programs available on the Web (e.g., Dali [2,3], CE [4], or VAST [5]) can guarantee an optimal structural alignment in every case. For this reason, it is very useful to have several publicly-available structure alignment tools, as well as a single measure of structural similarity that can be applied to all of them in order to select the best result.
Recently, we introduced an intuitive and convenient measure of structural similarity, the Number of Equivalent Residues (NER), and evaluated several popular structural alignment servers based on this score [6]. By using a single metric (NER) we were able to show that the servers generally converged on the same solution, a result that was not apparent when two metrics (e.g. RMSD and number of aligned residues), or raw scores were used to compare server results.
Another result was that there were occasionally significant differences between servers. This was particularly true for proteins with repeating motifs (e.g. TIM barrels), multiple domains, or in cases where the structurally equivalent residues represent only a small subset of the total. More recently, Levitt and co-workers concluded that there was "wide variation" in alignment quality among different programs and that the performance of any single method was much lower than using the best result from several methods [7]. Our own abservations along these linese motivated us to design a structural alignment algorithm that is robust in locating the global maximum of the NER score even for very difficult cases.
Since the NER score requires an initial alignment or superposition, a straightforward way to add robustness to the optimization algorithm is to increase the number of initial alignments. For this purpose a new alignment program, Local ASH, based on the double dynamic programming algorithm, was developed. In contrast to our earlier program, Global ASH [8,9], that computed only the globally optimal alignment, the new program computes multiple, locally-optimal alignments.
In addition to accepting multiple initial alignments, the GASH program allows crossover between alignments, as is done in genetic algorithms (hence the "G" in GASH). Since both the number of initial alignments and the number of crossovers is an adjustable parameter, the GASH program can make a very good estimate of the true maximum of the NER score for an arbitrary pair of protein structures. One can even import alignments from other programs, and we give an example of combining Local ASH, DaliLite, and CE alignments in this study.
The test-set presented here has been significantly expanded compared to earlier work. In addition to the Fischer-Eisenberg set of structural pairs, our new set consists 3,101 pairs representing many different folds, as defined by SCOP [10]. In addition, GASH is compared directly with the DaliLite and CE executables, allowing CPU time as well as accuracy to be evaluated. As in previous work, accuracy is defined by both the NER score and the number of residues aligned within a given RMSD threshold. In addition, several structure pairs that were not aligned properly by our previous program, Global ASH, are eximined in detail in terms of the alignment of functionally conserved residues.
Implementation
The overall approach is to globally optimize the NER score in three steps:
1. Produce a set of locally optimal alignments.
2. Parse each alignment into geometrically-consistent sub-alignments using distance matrix comparison.
3. Cross the alignments a fixed number of times and select the best unique set by NER maximization.
This procedure as well as our earlier protocol are illustrated in figure 1, and each of the steps is described in more detail below.
Figure 1 GASH flowchart. A flow chart of the Global ASH/NER (OLD) and GASH (New) methods is shown. The key differences between the old and new methods are: the generation of multiple initial alignments, a modified parsing algorithm for generation of sub-alignments, and the further generation of hybrid alignments by crossover.
Target function
The NER score has been described in detail previously. In brief, NER is a sum over all aligned residue pairs of a similarity function S:
where k corresponds to an aligned residue pair, dk is the intermolecular distance between Cα atoms in the aligned pair, and the similarity score is a Gaussian curve with unit amplitude at zero distance:
The parameter dcut defines the tolerance in the similarity score. Since the best value of dcut depends on the problem at hand, we make it an adjustable parameter on our public server. In all calculations a value of 4Å was used.
Local ASH program
Local ASH is a local structural alignment program that utilizes the double dynamic programming algorithm (DDP) [11]. In DDP, a local frame is used to describe the environment of each residue. A set of vectors from the beta carbon of the residue in question to those of all the other residues in the structure is calculated using the local frame of the residue. The vectors are ordered in the set, according to the position of the destination residue in the primary structure. This set is called the structural environment of the residue. The next step is to form an optimal vector-to-vector correspondence between each pair of structural environments using standard dynamic programming (DP). The similarity between a pair of residues is given by the score resulting from an alignment of the corresponding structural environments. The similarity matrix thus obtained is used in a second DP calculation to yield the residue-to-residue correspondence or alignment. The DP used to evaluate the similarity between a pair of structural environments is called the 'lower level DP', whereas the DP used to solve the residue-to-residue correspondence is called the 'upper level DP'. The method is called DDP, since DP is used at two different levels. Taylor and Orengo further extended the method to local structural alignment [12] by using the Smith-Waterman algorithm for the upper level DP [13].
Both Global and Local ASH use DDP, but with some important modifications. The first modification is a distance cutoff used to define the structural environment. A sphere with a given radius is located at the beta carbon of each residue. The structural environment of a residue is expressed as an ordered set of the vectors from the beta carbon of the residue to those of the residues whose beta carbons are within the sphere. The modified environment is called the local environment. The similarity between a pair of residues is calculated as the alignment score between the corresponding local environments. The similarity obtained from the comparison of the local environments is used for the local structural alignment. In order to avoid confusion with the distance cut-off used in the NER score, discussed below, we will refer to the ASH distance cutoff as the alignment radius.
Local ASH uses the Smith-Waterman algorithm for local structural alignment. When a pair of structures share multiple structural similarities that can not be expressed in a single alignment, the trace-back procedure is repeated in order to enumerate high-scoring solutions. After each trace-back operation, the scores corresponding to the alignment path and the neighboring region in the similarity matrix for the upper level DP are cleared. This ensures that each alignment will trace out a unique path. Orengo and Taylor introduced a window surrounding the local alignment path as the region to be cleared [12]. In contrast, Local ASH adopts the declump algorithm, which has been developed for local sequence alignment [14]. Local ASH also has the option to detect local similarity derived from circular permutation [15], although the option was not used in this study.
For results described here, the alignment radius we set to 14Å, and the maximum number of alignments output was 25. The internal gap opening and extension penalties for the lower level DP calculation were set to 10 and 0.5, respectively. For terminal regions the gap penalties were all zero. For the upper DP calculation, the gap opening and extension penalties were also set to 10 and 0.5, respectively. The Local ASH source code can be freely downloaded [16].
Parsing alignments using distance matrix comparison
Local ASH can not "see" beyond the alignment radius. If two or more regions of structural similarity exist and are separated by a distance greater than the alignment radius, an alignment may be constructed that runs through both regions. If these multiple regions do not correspond to a superposition with the same rotation and translation values, there will be more than one maximum in the landscape of the NER scoring function for the alignment. The distance matrix comparison step is used to decouple such geometrically-distinct sub-alignments.
A necessary condition for an alignment to be geometrically consistent is for all of the intramolecular distances between aligned residues in one structure to be approximately the same length as the corresponding distances in the other structure. Such a comparison of distance matrices forms the basis of the Dali target function [2]. Here we use it as a constraint. The sub-alignments thus constrained correspond to sets of residue pairs where the difference in corresponding intramolecular distances agree within a specified tolerance.
The parsing is iterative: the first aligned pair of residues (iq,it) initiates a sub-alignment (Here, the subscript q refers to the query and t to the template). Then we attempt to add a second aligned pair (jq,jt) to this sub-alignment. If we find that any of the intramolecular distances in the query differ by more than a cutoff value from the corresponding distance in the template, , the residue pair (jq,jt) is rejected from the sub-alignment; otherwise, it is accepted. If the pair is rejected from all existing sub-alignments, a new sub-alignment is initialized. Each subsequent pair of residues is compared to each existing sub-alignment in this manner until all residue pairs in the original alignment have been accounted for.
There are two differences between the algorithm used in GASH and that used by us previously: first, a residue pair can be in more than one sub-alignment, as long as the difference cutoff is satisfied. Second, the cutoff used in the current work is 10Å rather than 20Å.
In figure 2 we show some of the sub-alignments from a single local alignment between 1sftB and 1ezwA. Although about two-thirds of the aligned pairs have not been plotted, in order to make it easier to see individual pairs, it can be seen that the aligned pairs from a particular sub-alignment are not clustered together in sequence number but are distributed over a wide range of values.
Figure 2 Alignment parsed by distance matrix. The parsing of a single local alignment into geometrically consistent sub alignments is illustrated. Only five sub-alignments are shown, and consecutive aligned residue pairs belonging to the same sub-alignment are represented by a single point in order to make the plot easier to see. The secondary structure (helices in blue and strands in red) is plotted along the axis.
NER maximization
NER maximization involves first optimizing the superposition, given an initial alignment, and then re-optimizing the alignment based on the new superposition. For optimizing the superposition, we first minimize the Cα RMSD of the aligned residues, then directly maximize the NER score by conjugate gradient optimization. The method of Mclachlan is used for RMSD minimization [17]. For the conjugate gradient optimization step we use the Fletcher-Reeves-Polak-Rebiere method as implemented in the Numerical Recipes program frprmn [18].
We then re-calculate the alignment based on the residue-based similarity score (eqn. 2) using an ordinary dynamic programming calculation [19]. This re-alignment step is similar in approach to that used by May and Johnson [20,21] as well as Gerstein and Levitt [22]; however, in our case we do not iterate between superposition and alignment. From the new alignment, we calculate the optimal NER4 score. Note that the final NER score will be sensitive to the relative gap penalty used in the dynamic programming step, so one must be sure to use the same parameters when comparing alignments. The best choice for the gap penalty depends on the problem at hand. For all results presented here, we use .25 for internal gap opening, and .125 for internal extension.
Crossing alignments
Genetic algorithms have been used to generate alignments [23] or superpositions [20,21]. Here we use Local ASH to generate a very reasonable set of initial alignments, and use the crossover operation to exchange information between this initial set of alignments in order to obtain a globally optimal solution. We do not require the mutate, or other local operations, as we are not attempting to generate new information at this point, and because the NER maximization procedure can locally optimize an imperfect alignment generated by the crossover step. Crossover is the only stage where a random element is explicitly introduced into the procedure, and thus, in combination with the choice of initial alignments, is a point where the extent of sampling can be adjusted. The crossover algorithm used in GASH is as follows:
1. A stack to hold hybrid alignments is initialized with a fixed maximum length.
2. Two alignments as well as a splice-point in one of them are chosen at random.
3. The splice point in the second alignment is chosen so that the C-terminal portion is as long as possible without containing any of the residues in the N-terminal portion of the first alignment.
4. The alignments are crossed with a single progeny: the N terminal portion of alignment 1 and the C-terminal portion of alignment 2.
5. The CA RMSD of the new alignment is minimized, and the NER score calculated from the RMSD-minimized superposition. If the stack-size has not reached the maximum value, the new alignment is saved; otherwise, if this new NER score is greater than the lowest saved NER score in the stack, the lowest saved alignment is replaced by the new alignment. (This means we don't have to sort the stack but just keep track of the worst saved solution.)
6. Return to step 1 a fixed number of times (Ntry).
The entire cross-over cycle is repeated Nstart times, starting from the same initial set of alignments. After each cycle, the top Ntop alignments are subjected to full NER maximization. The values Nstart, Ntry, and Ntop are all parameters that can be used to adjust the extent of sampling, as described in the GASH Variants section, below.
In figure 3 we show the final GASH alignment between 1sftB and 1ezwA along with some of the initial alignments produced by Local ASH. We can see in this example that the final GASH alignment samples at least three different local alignments, and that this solution is completely different from the one obtained by Global ASH.
Figure 3 Local and global alignments. The crossover operation is illustrated here by showing the final GASH alignment between 1sftB and 1ezwA. Four of the initial Local ASH alignments are shown as scatter plots, which are partially sampled by the final GASH alignment, as well as the Global ASH alignment.
Saving unique results
It should be emphasized that, while in the present study we focus on global optimization, the GASH program produces multiple solutions. The selection of final solutions involves sorting the saved results and removing lower-scoring alignments from the list that are similar (as defined by an RMSD threshold) to higher-scoring ones. On the GASH server, the RMSD threshold, as well as an NER threshold for accepted solutions, can be adjusted in order to modify the number of retained alternate solutions.
DaliLite program
The DaliLite program was obtained from the FSSP [24] server [25]. The DaliLite NER scores were calculated by dynamic programming as described above using the DaliLite superposition without modification. The calculation was identical to that used to evaluate the superposition of the GASH structures.
CE program
The CE program was obtained from the CE server [26]. The NER scores were calculated in a manner identical to that used for Dali, except that the superimposed coordinates were generated by our own script based on the rotation matrix and translation vector produced by the CE program.
GASH variants
In order to assess the necessity and sufficiency of the crossover step, we report results using four versions of GASH: Default, No Cross-over, High Cross-over, and Meta. These variants are summarized below.
Default GASH
The default settings for Nstart, Ntry, and Ntop are 2,100, and 20, respectively. Initial alignments are generated by Local ASH.
No cross-over GASH
In order to assess whether the extent of crossover specified by the default parameters is necessary, GASH was run with no crossing over (Nstart = 0).
High cross-over GASH
In order to assess whether the extent of crossover specified by the default parameters is sufficient, GASH was run with extensive crossing-over (Nstart = 200).
Meta GASH
In order to assess whether the initial set of alignments generated by Local ASH is sufficient, alignments extracted from DaliLite and CE were added to the initial set. In the case of DaliLite, we used all alignments generated by the program as well as those extracted from the DaliLite superpositions, using the re-alignment procedure, described above.
Test sets
Two test sets are used to assess the performance of GASH. The first was derived using SCOP[10] and FSSP[24]. The second was derived by Fischer and Eisenberg has been used by others[4] to benchmark structure comparison algorithms. We will refer the these as the SCOP-FSSP and the Fischer-Eisenberg sets, respectively.
The SCOP-FSSP set was generated by a two-step procedure. First, SCOP [10] entries from different families were chosen by hand then checked to see if they were among the pre-computed lists of structure pairs at the FSSP [24] server. One entry, trehalose-6-phosphate synthase, was not found on FSSP, so trehalose-6-phosphate phosphatase related protein (1u02A) was used instead. The resulting 17 structures are referred to as "queries". Next, for each query, all structural neighbors were taken from the FSSP server. These structures are referred to as "templates". The template list was obtained by browsing the "FOLD Index," based on the PDB90 representative set of proteins [27], starting from the PDB ID of the SCOP entry. Note that while no two templates share more than 90% sequence identity, one of the templates is likely to have a high sequence similarity to the query as this closest matching representative is used to select the best pre-computed list. Examples are 1sftB (query) 1bd0A (representative template) and 1ab8A (query) 1cs4B (representative template).
A number of multi-domain queries were selected in this way under the assumption that these structures would yield a diverse and challenging set of templates. The result was a set of 3,593 structure pairs containing many different types of proteins. Since FSSP, which uses Dali to generate structure pairs, was used to create our test-set, there should be no bias toward GASH in the SCOP-FSSP set. Subsequently 491 of these structure pairs were eliminated because one or more of the programs failed to produce a meaningful alignment (see below), resulting in a set of 3,102 structure pairs that were used for the present analysis. The set of queries, along with their fold classifications is given in Table 1.
Table 1 Query List. The set of queries used to generate the SCOP-FSSP set is shown. The chain ID, if non-blank, is appended to the PDB ID (column 1). The number of residues refers to the entire protein chain. The Class and Fold are taken from SCOP, except in the case of 1u02A, which was not classified by SCOP.
PDB ID Protein Name Nres Class Fold
1ab8A Type II Adenylyl Cyclase C2 Domain 208 α+β Ferredoxin-like
1bxrA Carbamoyl Phosphate Synthetase(CPS) 1104 1. α
2. α/β
3. α/β
4. α/β
5. α/β
6. α+β 1. CPS connection domain
2. Swivelling β/β/α domain
3. Flavodoxin-like
4. Methylglyoxal-like
5. PreATP Grasp domain
6. ATP Grasp
1frvA Nickel-Iron Hydrogenase (NIH) 293 1. α+β
2. α+β 1. NIH Large subunit
2. NIH Small Subunit
1mniA Myoglobin 184 α Globin-like
1qgtB Hepatitis B Viral Capsid 174 α Hepatitis B Viral Capsid
1u02A Trehalose-6-Phosphate Phosphatase Related Protein 253 α/β
1bgw DNA topoisomerase II, C-terminal fragment (residues 410–1202) 709 Multi-dom α+β Type II DNA Topoisomerase
1dwuA Ribosomal Protein L1 244 Multi-dom α+β Ribosomal Protein L1
1gqeA Polypeptide release factor 2 385 Multi-dom α+β Polypeptide release factor 2
1nvbB Dehydroquinate Synthase (DHQS) 422 Multi-dom α+β DHQS-like
1r6fA Low Calcium Response Protein V (LcrV) 303 Multi-dom α+β Virulence-associated V antigen
1udyA Medium Chain acyl-CoA Dehydrogenase 416 1. α
2. Multi-dom α+β 1. Bromodomain-like
2. Acyl-CoA dehydrogenase NM domain-like
1bwwA Immunoglobulin Light Chain Kappa Variable Domain 140 β Immunoglobulin-like β-sandwich
1e03L Antithrombin 454 Multi-dom α+β Serpins
1kyqB Bifunctional dehydrogenase/ferrochelatase Met8p 298 1. α/β
2. Multi-dom α+β 1. NAD(P)-binding Rossmann-fold domains
2. Siroheme Synthase Middle domain-like
1obaA Endolysin 369 1. β
2. α/β 1. β-hairpin stack
2. TIM β/α-barrel
1sftB Alanine Racimase 411 1. β
2. α/β 1. Domain of α+β subunits of F1 ATP synthase-like
2. TIM β/α-barrel
The Fischer-Eisenberg set was taken from table VI in Shindyalov and Bourne [4].
Results
Overview
The results were generated by running each program on the command line on one of 14 identical personal computers (Intel Pentium4 processor, Linux RedHat 8.0 operating system). In order to make a fair comparison, the same re-alignment procedure was used to evaluate the NER scores and the numbers of aligned residues for all alignment methods. The numbers of aligned residues were computed for three RMSD cut-offs: 2,4, and 6Å, which we will refer to as N2, N4, and N6, respectively. Occasionally one or more of the programs failed to produce a result, or only aligned a few residues. Since we could not distinguish between software errors (e.g. parsing the initial PDB file or some incompatibility between our system and one of the programs) and true algorithmic deficiencies, we eliminated any template from the list if any one of the methods failed to produce an NER score greater than 10. Based on the higher number of failures for DaliLite, CE and Global ASH (171,180, and 100, respectively) compared to the default, no crossover, and high crossover GASH (58,64, and 56, respectively) it is unlikely that eliminating templates in this way biased the results in favor of GASH.
Averages of all similarity measures as well as the number of internal gaps were computed for each query set as well as for the entire set of results (see Additional file 1). In addition, the average CPU time per alignment is given for each query set and for the entire set of results in table 2. From the SCOP-FSSP set, a sub-set of 5 structure pairs that were not aligned properly by Global ASH are discussed in detail and shown in figures 7, 8, 9, 10, 11. Finally, we present results for the smaller Fischer-Eisenberg data set (see Additional file 2).
Table 2 Timings. The average CPU times for each query from the SCOP-FSSP set using each of the 6 programs is shown. The Meta Gash program CPU can be closely approximated by summing the GASH default, DaliLite, and CE columns.
Query Average CPU (seconds)
GASH Global GASH GASH
ID Nres Default Dalilite CE ASH No Cross High Cross
1bwwA 110 4 4 13 4 3 8
1qgtB 144 7 6 17 10 6 10
1mniA 154 4 4 13 5 4 9
1ab8A 178 5 3 15 6 5 9
1dwuA 214 13 9 18 17 13 20
1u02A 223 13 10 19 15 13 18
1frvA 263 14 10 18 16 13 20
1kyqB 268 15 10 20 20 15 21
1r6fA 273 7 5 17 10 7 11
1obaA 339 27 23 24 31 25 34
1gqeA 355 15 10 22 19 14 20
1sftB 381 25 22 24 28 24 32
1udyA 386 14 12 24 20 13 19
1nvbB 392 25 16 23 30 25 32
1e03L 424 22 22 18 31 21 36
1bgw 680 40 25 45 53 39 48
1bxrA 1074 60 28 61 83 58 69
Figure 7 Myoglobin aligned to Phycocyanobilin. Myoglobin (1mniA, query) aligned to Phycocyanin (1phnB, template). Residues that bind heme in 1mniA and phycocyanobilin in 1phnB are underlined, with matches indicated by a + and the total number of matches reported at the top of each alignment. The color scale used in this figure is identical to that of figure 6. The secondary structure assignments, residue equivalences, and terminal gaps have all been omitted in order to save space.
Figure 8 Carbamoyl phosphate synthetase aligned to methylglyoxal synthase. Carbamoyl phosphate synthetase (1bxrA, query) aligned to methylglyoxal synthase (1egh, template). Conserved residues in the methylglyoxal synthase-like superfamily are underlined, with matches indicated by a + and the total number of matches reported at the top of each alignment. The format used in this figure is identical to that of figure 7.
Figure 9 Alanine Racimase aligned to imidazole glycerol phosphate synthase. Alanine Racimase (1sftB, query) aligned to imidazole glycerol Phosphate synthase (1jvnA, template). A pair of function residues found the TIM barrel are underlined, with matches indicated by a + and the total number of matches reported at the top of each alignment. The format used in this figure is identical to that of figure 7.
Figure 10 Met8p aligned to flavohemoglobin. Met8p (1kyqB, query) aligned to Flavohemoglobin (1cqxA, template). The NAP(p)-binding loop residues are underlined, with matches indicated by a + and the total number of matches reported at the top of each alignment. The format used in this figure is identical to that of figure 7.
Figure 11 Immunoglobulin Light Chain Kappa Variable Domain aligned to antibody for phenobarbital. Immunoglobulin Light Chain Kappa Variable Domain (1bwwA, query) aligned to antibody for phenobarbital (1igyB, template). The characteristic disulfide bond and Thr residues are underlined, with matches indicated by a + and the total number of matches reported at the top of each alignment. The format used in this figure is identical to that of figure 7.
Output files
The raw data resulting from each of the 3,102 structure pairs using 7 different methods is too large to be presented in tabular form here. For this reason, we have prepared a link to the data on our server [28]. The main page summarizes the results for each query and is reproduced in text form in Additional file 1. The HTML table contains links for each query containing individual scores for each structure pair using each of the 7 alignment methods. Within each query page there are links to each alignment and superposition (in PDB format). An example of the alignment between 1bwwA and 1jv5B is given in figure 4.
Figure 4 GASH alignment format. The alignment between 1bwwA and 1jv5B using default GASH is shown. In addition to the total NER score (eqn. 1), the residue-based similarity score (eqn. 2) was evaluated and scaled to integer values between 0 and 9. The distribution of such equivalences is reported at the bottom of the alignment. In order to roughly define the beginning and end of the most important parts of each alignment the first and last set of 5 continuous residues where the average similarity score was 5 or more was located. We refer to this region as the core alignment, and report the number of gaps and aligned residue pairs within the region. Also, the number of residues aligned under the three RMSD cutoffs, N2-6 are indicated. The alignments were written out with the residue pairs and secondary structure color coded by the similarity scale (with red the most and blue the least similar), making it easy to recognize regions of structural similarity.
Correlation between NER score and the number of aligned residues
Although the main focus of this work is on optimization, we first consider whether the NER score is a valid target function for structural alignment by comparing it to a more familiar metric: the number of aligned residues under a given RMSD.
In figure 5 we plot NER4 versus N2, N4, and N6 for all 3,102 structure pairs in the SCOP-FSSP data set using all 7 alignment methods. Since RMSD is an average over a set and NER is a direct sum of normalized values, we do not expect to see an exact agreement; nevertheless, over a broad range of values, the correlation between the different numerical measures is approximately linear. The slope is closest to unity when the smallest RMSD cutoff (2Å) is used (slope 1.2, correlation coefficient .97).
Figure 5 Number of aligned residues under a given RMSD. The correlation between NER4 and the number of aligned residues under three cut-offs is shown. The entire set of alignments from 3,102 structure pairs and 7 alignment methods was used to make this plot. The slope between NER4 and the number of aligned residues under 2Å was 1.2 with a correlation coefficient of .97.
The fact that NER is a direct sum makes it much easier to optimize than the number of aligned residues under a given RMSD cutoff. (For example, it can be maximized directly by dynamic programming or by conjugate gradient optimization.) This fact coupled with the nearly linear agreement between the two measures validates the utility of NER as a target function for structural similarity.
Optimization performance
First we consider the improvement of GASH compared to Global ASH from the summaries given in Additional files 1 and 2. In terms of optimization of the NER score (or any of the other measures), GASH consistently out-performs Global ASH. An improvement of approximately 10 residues is seen in every query average, with the exception of 1e03L, where the two programs agreed on average. The improvement in terms of alignment accuracy is achieved while at the same time decreasing the CPU time per alignment (table 2).
Next we consider the performance compared to DaliLite and CE. GASH consistently aligned an equal or greater number of residues than either DaliLite or CE, independent of the measure used to define accuracy. The improvement relative to CE is particularly dramatic, with an average of 7–19 more residues aligned (depending on the measure used). With DaliLite, the improvement (4–8 residues, on average) was not as dramatic, but it was consistent across most query sets. In terms of CPU usage, CE was the slowest for proteins in the 100–273 residue range, and DaliLite was several seconds faster than GASH in this size range. In the 339–424 residue range, CPU times differ between the four programs by only a few seconds, on average. The greatest difference in CPU usage is seen for the largest structures: DaliLite uses only 25 and 28 seconds, for 1bgw (680 residues) and 1bxrA (1074 residues), but both GASH and CE run for longer times(approximately 40 and 60 seconds, respectively).
It must be emphasized that these results represent use of the DaliLite and CE programs as is, without any changes in the source code for this difficult set of alignment problems. In the hands of the authors, CE or DaliLite might well yield higher NER scores and/or lower CPU times.
Sufficiency of Local ASH initial alignments
We address the question of coverage in the initial set of alignments by comparing the default GASH performance with Meta GASH, where alignments from DaliLite and CE were added to the initial set. Since GASH uses only the crossover operation from the genetic algorithm, not mutation, the only new information it can generate is in the re-alignment step. DaliLite and CE both generate alignments by completely different algorithms from each other and from Local ASH, so if there is not enough information in the initial set of Local GASH alignments, we should see an improvement in the accuracy when alignments from DaliLite and CE are added.
In terms of the NER score, there is a slight average improvement of 2 aligned residues upon using Meta GASH. This improvement can be observed consistently across most query sets, showing that we are not always locating the exact global optimum in the NER score when using default GASH. However, these differences are not great enough to justify using Meta GASH routinely, since the CPU usage would be approximately 3 times that of default GASH.
Necessity and sufficiency of crossover
In order to determine if the crossover is both necessary and sufficient, we compare default GASH to GASH without crossover and to GASH with high crossover. The no-crossover results are not as good as those of default GASH, but the improvement is not very dramatic, on average. In fact, no-crossover GASH is slightly better on average than DaliLite, and at a competitive CPU usage. However, if we consider particular cases, such as the alignment between 1gqeA and 1p32A (figure 6), the difference in NER4 is 16, and in the number aligned under an RMSD of 2Å is 28 residues. Such particular cases, as well as the fact that the difference in CPU between default and no-cross GASH is only a few seconds at most, justifies the use of the crossover operation. In contrast, when we increase the crossover by a factor of 100 we do not see an improvement in any of the similarity measures, on average. This strongly suggests that the extent of crossover in the default program is both necessary and sufficient.
Figure 6 Default GASH vs. no crossover. The default GASH protocol is compared to GASH without crossover for 1gqeA (query) aligned to 1p32A (template). The NER equivalence (eqn. 2) is indicated numerically, on a 0–9 scale, and by color (with red the most and blue the least similar).
Specific examples from SCOP-FSSP set
Here we examine the quality of 5 pair-wise alignments in detail. In addition to numerical measures, such as the number of equivalences, we consider functional information, where available. The examples chosen represent cases where GASH outperformed Global ASH in terms of the NER score, and include all-α, all-β, and mixed α/β folds. In all examples, "GASH" refers to the default GASH method. All sequence identities quoted are obtained by running the entire sequence for both query and template at the University of Southampton SBDS server [29], which uses Lipman and Pearson's algorithm [30].
Myoglobin (1mniA) aligned to phycocyanin (1phnB)
Myoglobin and phycocyanin belong to the same SCOP fold group but have different functions. Alignment of the two sequences yields an identity of 15%. Myoglobin utilizes a heme group to transport oxygen and phycocyanin binds a phycocyanobilin chromophore for light harvesting. The heme group bound by myoglobin and the chromophore bound by phycocyanin are positioned similarly [31]. Although 1mni is a double-mutant form (with two of the binding-site residues switched), the mainchain RMSD from the native is only .25Å [32]. Thus we expect that in the proper structural superposition, the residues responsible for binding the prosthetic groups would be aligned. Figure 7 shows the GASH, DaliLite, CE, and Global ASH alignments between 1mniA and 1phnB. In terms of NER4, GASH (70) and DaliLite (67) perform similarly well, while CE (60) and Global ASH (61) perform similarly poorly. The N2 -4 scores are more varied, but follow the same general trend. There are 22 heme/chromophore binding residues in each structure. Their distribution in sequence is such that a perfect match seems unlikely; nevertheless, the number of matches correlates with the NER score except in the cases with low matches: CE aligns 10 of the functional residues, whereas Global ASH only aligns 7. In the case of DaliLite (14) and GASH (15) we can see that there are only slight differences in the alignment, and that the one pair of functional residues aligned by GASH but not by DaliLite is a borderline case: they are nearly aligned in the DaliLite alignment, and in fact, occur at a point of fairly poor structural superposition in both alignments.
Carbamoyl phosphate synthetase (1bxrA) aligned to methylglyoxal synthase (1egh)
The C-terminal domain of 1bxrA and the entire structure of 1egh both belong to the methylglyoxal synthase fold [10,33-35]. However, the shapes of their active sites, as well as the functions of the two proteins differ significantly. As a result, we cannot use the same approach used to check the 1mniA-1phnB alignments to check the 1bxrA-1egh alignments. Fortunately, even though the overall sequence identity is only 6%, these two proteins contain conserved residues. The set of conserved residues was defined in the following way: each sequence was used as a query to the Conserved Domain Database[36], yielding an alignment to the consensus sequence of the methylglyoxal synthase-like domain; the 1bxrA-1egh alignment was then constructed by aligning the consensus sequences of from each alignment. Conserved residues were defined to be those residues that were aligned and identical to the consensus sequence. This small set of conserved residues are distributed throughout the domain. As figure 8 shows, GASH aligned all of the conserved residues correctly, with two exceptions: The first conserved Lys residue does not superimpose structurally; also, Asp 1025 in 1bxrA should be aligned to Asp 101 in 1egh; however, it is aligned instead to residue 99. Coincidentally, residue 99 happens to be an aspartic acid as well, but this appears to be an accident – the conserved Asp is residue 101. The real problem here lies in the fact that GASH considers only Cα residues in constructing the equivalences used to compute the final alignment. In fact, the side-chains in residues 1025 and 101 are much closer than those of 1025 and 99. In other words, the superposition is fine, but the alignment computed from the superposition is less than optimal, due to the exclusive use of Cα residues in the scoring function. The DaliLite alignment completely misses the C-terminal domain in 1bxrA, and instead aligns 1egh to the N-terminal pre ATP grasp domain. CE produces an alignment that is almost identical to GASH. In terms of the numerical measures, the GASH alignment is slightly better, although CE aligns all of the conserved residues correctly, with the exception of the first Lys, including the aforementioned Asp. Local ASH incorrectly aligns 1egh to the connection domain in 1bxrA.
Alanine racimase (1sftB) aligned to imidazole glycerol phosphate synthase (1jvnA)
Alanine Racimace and imidazole glycerol phosphate synthase share a TIM barrel domain with the active site located at the top of the barrel [10,37-39]. There is only 17.5% sequence identity between the two, and they do not share a common ligand, so we can not find obvious markers as we did in the previous examples. However, both proteins are involved in peptide biosynthesis, and each contains one catalytic residue in the TIM domain: Lys 19 in 1sftB acts as a proton acceptor specifically for D Alanine and Asp 245 in 1jvnA makes hydrogen bonds with imidazole glycerol phosphate, a precursor in the histidine synthetic pathway. As figure 9 shows, GASH and DaliLite yield essentially the same alignment, and align the Lys-Asp pair. Although there is no reason to assume a priori that the functional residue in the two proteins should align, the fact that it does is probably not an accident. The CE alignment is somewhat lower in quality by numerical measures, but aligns the functional residue pair as well. The Global ASH alignment is completely different from the rest, and much lower in quality. In our previous study of Global ASH we also found that correctly pairing the beta strands in TIM barrel structures was non-trivial, due to the 8-fold symmetry of the barrel [6].
Met8p (1kyqB) aligned to flavohemoglobin (1cqxA)
Both 1kyqB and 1jvnA contain a NAD(p)-binding Rossmann domain [31,40], but the sequence identity is low (15.5%), and there are significant differences in the binding pocket as well as the topology of the fold. 1kyqB contains an extra anti-parallel strand at the edge of the sheet. Interestingly, both structures contain a long, extended, and highly charged loop; however, although the loop occupies a similar spatial position in each molecule, the location of this loop in the primary structure is different. It is difficult to find many functional or conserved residues that are paired in the alignment. The NAD binding loop is much longer in 1cqxA than 1kyqB, so we see a distortion in the alignment precisely at this point. However, the general position of the NAD(p)-binding residues can be used to assess the alignments. In terms of the numerical measures, GASH and DaliLite perform similarly, and both align the loop interacting with NAD(p) as well as can be expected. By any measure the CE alignment is not as accurate, and the Global ASH alignment is a failure, aligning a small fragment from two completely different domains (figure 10).
Immunoglobulin light chain kappa variable Domain (1bwwA) aligned to antibody for phenobarbital (1igyB)
Both 1bwwA and 1igyA are immunoglobulins, and contain the characteristic β-sandwich fold. 1bwwA is a single-domain structure, but 1igyB is an intact monoclonal antibody and contains both a variable domain and three constant domains[41]. Thus, the problem of aligning these two structures consists of identifying the best structural match among 4 domain choices. All of the domains in these two structures have the characteristic disulfide bridge and a Trp group located near the bridge. From the standpoint of aligning the correct domain, aligning the Cys and Trp residues, and from the numerical scores, GASH, DaliLite, and CE all succeed and find the exact same solution; Global ASH, on the other hand aligns 1bwwA to one of the constant domains and gets much lower numerical scores. On the other hand, Global ASH does get the functionally conserved residues from the constant domain aligned correctly to those in the variable domain (figure 11).
Fischer-Eisenberg set
In Additional file 2 we summarize 10 structure pairs from the Fischer Eisenberg data set. The structures in this set are generally smaller than those in the SCOP-FSSP data set, so it is not surprising that the differences between methods are not as large. The general trend, in terms of both numbers of aligned residues and NER score is the same as in the SCOP-FSSP results: Default GASH ≥ DaliLite ≥ CE, but the differences are probably not significant. This suggests that the differences between methods only becomes important when there are multiple domains and/or multiple regions of structural similarity.
Conclusion
The primary goal of this study was to design an algorithm that reliably maximizes the NER score for an arbitrary pair of protein structures. The results here indicate that the GASH program is successful in this regard, and that the extent of sampling can easily be increased, if necessary, by adding more initial alignments. Although we have not yet optimized every parameter used in GASH, the results in terms of NER and other scores using default parameters is encouraging. From looking at the dependence on crossover and on the initial alignment set, we can surmise that most of the improvement relative to Global ASH is due to the use of multiple Local ASH alignments.
The Dali algorithm has recently been validated extensively against CATH classifications using receiver operating characteristics in two studies [7,42], however in one of these studies [7], the quality of Dali alignments was found to be inferior to CE. Based on our smaller study Dali alignments appear to be more accurate than CE alignments, by any measure. Perhaps eliminating structure pairs from the test set that could not be aligned by one or more of the programs had some effect on the results. Since CE failed in this regard slightly more often than DaliLite, however, it seems unlikely that this had any effect. GASH has not yet been benchmarked on such a comprehensive test set or against fold classifications, such as CATH or SCOP.
Even with a test set of 3,102 structure pairs, the improvement of GASH relative to Global ASH, both in terms of accuracy and CPU usage, is unambiguous. Moreover, as the specific examples illustrate, a higher NER score correlates well with "correctness" in terms of matching important residue pairs. The one case where GASH misaligns a residue pair found by CE immediately suggests an obvious improvement to the program: using side-chain atoms in addition to Cα atoms to define the equivalence. We intend to incorporate this improvement, as well as to look at a more comprehensive set of structure pairs in the near future.
Availability and requirements
In addition to web access through both a CGI and java interface, we have developed a Simple Object Access Protocol (SOAP) server that allows GASH to be run remotely on the command-line. Sample java and Perl client programs are available for accessing the SOAP server. The java programs require installation of the Apache Axis library [43] and the Perl programs require installation of the SOAP-Lite perl module. All three interfaces are described at the GASH server .
Supplementary Material
Additional File 1
Summary of SCOP-FSSP set results. For each query, the number of residues, number of structural templates (Ntmp), and results from each of the 7 alignment methods, are shown. For each alignment method, the NER4 score (NER), number of gaps in the core alignment (Gp), and number of aligned residues below three RMSD cutoffs (N2,N4,N6) are reported. Each entry in the table represents an average over the number of templates specified in column 3. The averages on the last row refer to the total set of 3,102 structures. In the case of the number of templates, the sum, rather than the average, is given.
Click here for file
Additional File 2
Summary of Fischer-Eisenberg set results. For each structure pair, the PDB ID and number of residues, as well as results from each of the 7 alignment methods, are shown. For each alignment method, the NER4 score (NER), number of gaps in the core alignment (Gp), and number of aligned residues below three RMSD cutoffs (N2,N4,N6) are reported. In the last row, the averages over all ten structure pairs are given.
Click here for file
Acknowledgements
The authors want to thank the anonymous reviewers for their careful reading of the manuscript.
==== Refs
Berman HM Westbrook J Feng Z Gilliland G Bhat TN Weissig H Shindyalov IN Bourne PE The Protein Data Bank Nucleic Acids Res 2000 28 235 242 10592235 10.1093/nar/28.1.235
Holm L Sander C Protein structure comparison by alignment of distance matrices J Mol Biol 1993 233 123 138 8377180 10.1006/jmbi.1993.1489
Holm L Sander C Dictionary of recurrent domains in protein structures Proteins 1998 33 88 96 9741847 10.1002/(SICI)1097-0134(19981001)33:1<88::AID-PROT8>3.0.CO;2-H
Shindyalov IN Bourne PE Protein structure alignment by incremental combinatorial extension (CE) of the optimal path Protein Eng 1998 11 739 747 9796821 10.1093/protein/11.9.739
Gibrat JF Madej T Bryant SH Surprising similarities in structure comparison Curr Opin Struct Biol 1996 6 377 385 8804824 10.1016/S0959-440X(96)80058-3
Standley D Toh H Nakamura H Detecting Local Structural Similarity in Proteins by Maximizing Number of Equivalent Residues Proteins 2004 57 381 391 15340925 10.1002/prot.20211
Kolodny R Koehl P Levitt M Comprehensive evaluation of protein structure alignment methods: scoring by geometric measures J Mol Biol 2005 346 1173 1188 15701525 10.1016/j.jmb.2004.12.032
Daiyasu H Toh H Molecular evolution of the myeloperoxidase family J Mol Evol 2000 51 433 445 11080366
Toh H Introduction of a distance cut-off into structural alignment by the double dynamic programming algorithm Comput Appl Biosci 1997 13 387 396 9283753
Murzin AG Brenner SE Hubbard T Chothia C SCOP: a structural classification of proteins database for the investigation of sequences and structures J Mol Biol 1995 247 536 540 7723011 10.1006/jmbi.1995.0159
Taylor WR Orengo CA Protein structure alignment J Mol Biol 1989 208 1 22 2769748 10.1016/0022-2836(89)90084-3
Orengo CA Taylor WR A local alignment method for protein structure motifs J Mol Biol 1993 233 488 497 8411157 10.1006/jmbi.1993.1526
Smith TF Waterman MS Identification of common molecular subsequences J Mol Biol 1981 147 195 197 7265238 10.1016/0022-2836(81)90087-5
Waterman MS Introduction to Computational Biology 1995 London, Chapman & Hall
Hiroike T Toh H A local structural alignment method that accommodates with circular permutation CBI Journal 2001 1 103 114 10.1273/cbij.1.103
ASH Server
McLachlan AD Gene duplications in the structural evolution of chymotrypsin J Mol Biol 1979 128 49 79 430571 10.1016/0022-2836(79)90308-5
Press WH Teukolsky SA Vettering WT Flannery BP Numerical Recipes in C 1992 2 Cambridge, Cambridge Univ. Press
Gotoh O Alignment of three biological sequences with an efficient traceback procedure J Theor Biol 1986 121 327 337 3795999
May AC Johnson MS Protein structure comparisons using a combination of a genetic algorithm, dynamic programming and least-squares minimization Protein Eng 1994 7 475 485 8029205
May AC Johnson MS Improved genetic algorithm-based protein structure comparisons: pairwise and multiple superpositions Protein Eng 1995 8 873 882 8746725
Gerstein M Levitt M Using iterative dynamic programming to obtain accurate pairwise and multiple alignments of protein structures Proc Int Conf Intell Syst Mol Biol 1996 4 59 67 8877505
Szustakowski JD Weng Z Protein structure alignment using a genetic algorithm Proteins 2000 38 428 440 10707029 10.1002/(SICI)1097-0134(20000301)38:4<428::AID-PROT8>3.0.CO;2-N
Holm L Sander C Touring protein fold space with Dali/FSSP Nucleic Acids Res 1998 26 316 319 9399863 10.1093/nar/26.1.316
FSSP Server
CE FTP Server
Holm L Sander C Removing near-neighbour redundancy from large protein sequence collections Bioinformatics 1998 14 423 429 9682055 10.1093/bioinformatics/14.5.423
GASH Results
SBDS Server
Pearson WR Lipman DJ Improved tools for biological sequence comparison Proc Natl Acad Sci U S A 1988 85 2444 2448 3162770
Ermler U Siddiqui RA Cramm R Friedrich B Crystal structure of the flavohemoglobin from Alcaligenes eutrophus at 1.75 A resolution Embo J 1995 14 6067 6077 8557026
Dou Y Admiraal SJ Ikeda-Saito M Krzywda S Wilkinson AJ Li T Olson JS Prince RC Pickering IJ George GN Alteration of axial coordination by protein engineering in myoglobin. Bisimidazole ligation in the His64-->Val/Val68-->His double mutant J Biol Chem 1995 270 15993 16001 7608158 10.1074/jbc.270.27.15993
Saadat D Harrison DH Mirroring perfection: the structure of methylglyoxal synthase complexed with the competitive inhibitor 2-phosphoglycolate Biochemistry 2000 39 2950 2960 10715115 10.1021/bi992666f
Saadat D Harrison DH The crystal structure of methylglyoxal synthase from Escherichia coli Structure Fold Des 1999 7 309 317 10368300 10.1016/S0969-2126(99)80041-0
Thoden JB Blanchard CZ Holden HM Waldrop GL Movement of the biotin carboxylase B-domain as a result of ATP binding J Biol Chem 2000 275 16183 16190 10821865 10.1074/jbc.275.21.16183
Marchler-Bauer A Bryant SH CD-Search: protein domain annotations on the fly Nucleic Acids Res 2004 32 W327 31 15215404
Chaudhuri BN Lange SC Myers RS Chittur SV Davisson VJ Smith JL Crystal structure of imidazole glycerol phosphate synthase: a tunnel through a (beta/alpha)8 barrel joins two active sites Structure (Camb) 2001 9 987 997 11591353 10.1016/S0969-2126(01)00661-X
Chaudhuri BN Lange SC Myers RS Davisson VJ Smith JL Toward understanding the mechanism of the complex cyclization reaction catalyzed by imidazole glycerolphosphate synthase: crystal structures of a ternary complex and the free enzyme Biochemistry 2003 42 7003 7012 12795595 10.1021/bi034320h
Shaw JP Petsko GA Ringe D Determination of the structure of alanine racemase from Bacillus stearothermophilus at 1.9-A resolution Biochemistry 1997 36 1329 1342 9063881 10.1021/bi961856c
Schubert HL Raux E Brindley AA Leech HK Wilson KS Hill CP Warren MJ The structure of Saccharomyces cerevisiae Met8p, a bifunctional dehydrogenase and ferrochelatase Embo J 2002 21 2068 2075 11980703 10.1093/emboj/21.9.2068
Harris LJ Larson SB Hasel KW Day J Greenwood A McPherson A The three-dimensional structure of an intact monoclonal antibody for canine lymphoma Nature 1992 360 369 372 1448155 10.1038/360369a0
Sierk ML Pearson WR Sensitivity and selectivity in protein structure comparison Protein Sci 2004 13 773 785 14978311 10.1110/ps.03328504
Apache Axis
|
16146579
|
PMC1239909
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Sep 8; 6:221
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-221
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2261615939310.1186/1471-2105-6-226SoftwareAssessing local structural perturbations in proteins Lema Martin A [email protected] Julian [email protected] Universidad Nacional de Quilmes, Roque Sáenz Peña 180, B1876BXD Bernal, Buenos Aires, Argentina2005 13 9 2005 6 226 226 18 8 2004 13 9 2005 Copyright © 2005 Lema and Echave; licensee BioMed Central Ltd.2005Lema and Echave; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Protein structure research often deals with the comparison of two or more structures of the same protein, for instance when handling alternative structure models for the same protein, point mutants, molecule movements, structure predictions, etc. Often the difference between structures is small, restricted to a local neighborhood, and buried in structural "noise" due to trivial differences resulting from experimental artifacts. In such cases, whole-structure comparisons by means of structure superposition may be unsatisfactory and researchers have to perform a tedious process of manually superposing different segments individually and/or use different frames of reference, chosen roughly by educated guessing.
Results
We have developed an algorithm to compare local structural differences between alternative structures of the same protein. We have implemented the algorithm through a computer program that performs the numerical evaluation and allows inspecting visually the results of the structure comparison. We have tested the algorithm on different kinds of model systems. Here we present the algorithm and some results to illustrate its characteristics.
Conclusion
This program may provide an insight into the local structural changes produced in a protein structure by different interactions or modifications. It is convenient for the general user and it can be applied to standard or specific tasks on protein structure research.
==== Body
Background
Localized perturbations in a protein structure can originate from point mutations, chemical modifications, interaction with other molecules, etc. Sometimes, it is necessary to compare alternative structures for the same protein sequence (e.g. different three-dimensional structure predictions, multiple models from NMR studies, etc.). To assess such protein structural perturbations, structures are usually compared in a detailed way, by looking at the position and orientation of individual atoms, residues, or secondary structures (for instance, see [1]). This approach is mandatory on case studies, because it leads to explain how modifications have changed the structure and function of a protein. However, this kind of comparison is usually done by superposing different particular structure elements individually and/or by using different protein-specific frames of reference, which are chosen according to the expertise and personal criteria of the researcher. This complicates establishing generalizations applicable to different proteins and the analysis of large numbers of cases. For such situations, quantitative measures of structural perturbation, such as the root mean square deviation (RMSd) or a derivative function [2-4], are used.
RMSd is a measure that is simple to calculate and to understand, it can be employed to establish comparisons through different structural families, it has been used very widely, and it is familiar to every researcher in the field. However, RMSd is usually referred to whole-molecule superpositions, so that it does not provide information on partial features, such as whether the perturbation is local or distributed throughout the whole molecule. This problem might be overcome by looking at the components of RMSd, the squared distances between pairs points compared. However, there is a problem with this approach, inherent to whole-molecule superpositions: portions of the structure with little or no perturbation may be badly superposed in order to improve the superposition of those portions with important distortions.
Several authors have developed effective methods for sequence alignment based on local structural features [4-13]. Most of these methods do not work with neighborhoods of equal and fixed size, or the local zones are not equally distributed along the molecule. This is not a problem for the task of structural alignment because, once a local zone is defined, the goal is usually to optimize whatever measure is used inside that zone. On the other hand, for the detection and description of structural alterations it is essential that all local zones have the same size, in order to compare among them on a uniform basis, therefore allowing the reliable identification of zones with a perturbation that is significantly different from the average. Moreover, these methods refer (and restrict) their local zones to a secondary structure element or a window of residues along the sequence. These are one-dimensional boundaries, which while being convenient for aligning sequences, are poor to examine three-dimensional structure perturbations, which may involve atoms from residues that are not close in the linear sequence, whose interactions are thus neglected.
Specifically for the recognition and measure of local structural alterations, we propose that it is more suitable to compare structures on a uniform and residue-based approach, and by delimiting the neighborhood of each residue just in terms of distances. As a result, we conceive a unit of comparison integrated by a residue and the group of adjacent atoms within a fixed radius. Therefore, we have developed an algorithm to quantify the degree of structural alteration in the local neighborhood of each residue, when comparing two or more structures, and the means of exploiting this measure not only analytically but also visually.
Implementation
COLORES (Comparison of LOcal Residue Environment Structures) is a program that compares two or more protein structures, by performing an assessment of the local structural alteration in the neighborhood of each residue. The input is a set of protein structure files in PDB (Protein Data Bank) format, a sequence alignment between those structures in GDE format (Genetic Data Environment, Steven Smith, 1994, Version 2.3), and a set of user choices described below. The program generates a log file containing detailed information of each local comparison, a data file containing summaries per alignment position, a structure (PDB-formatted) file and a script file for the RASMOL [14] program, which allows the user to inspect the results visually. COLORES automatically invokes RASMOL to show the results after its job is done.
For each position in the alignment, COLORES calculates two scores, described next:
Truly local score
The algorithm compares protein structures on a residue-by-residue basis. It calculates a score for each alignment position having no gaps on either sequence. The calculation is performed as described next:
For each structure (see Figure 1A), a sphere whose radius is chosen by the user is defined around the residue under consideration. These spheres can be centered either at the alpha carbon atom or at the center of mass of their respective residues, according to the user's choice.
Figure 1 Calculation of the truly local score. This figure explains how the truly-local score is calculated for a single residue in a pairwise comparison (using the default parameters for the program). Previously, a user-provided alignment allows establishing which residues and atoms of one structure are equivalent to which ones in the other. A) On each structure to be compared, a sphere is considered around the alpha carbon atom (red) of the residues corresponding to the alignment position whose truly-local score will be calculated. Two lists including all the backbone atoms falling inside each sphere are prepared (atoms which belong to one list are rendered blue, while the atoms which belong to the other are rendered green). B) Lists are compared to find equivalent atoms. The atoms that were inside one or another sphere are now rendered as balls. The color of atom pairs whose members were one of them inside one sphere and the other outside has been changed to yellow, and these pairs are dismissed from the comparison. C) The remaining atom pairs (whose members were both inside their respective sphere) are taken as a group of fixed points. Then, the alpha carbon atoms being the former centers of the spheres are superposed, and the system is rotated until the root mean square of the atom pair distances reaches a minimum. This minimum value is the score for the residue considered. All the process is iterated for each residue along the sequence. Variations and alternatives on the procedure are discussed in the main text.
Then, two lists are prepared containing all atoms inside each sphere. There are three options regarding the kind of atoms to be included (this is again a user choice defined as "eligible atoms"): all heavy (non-hydrogen) atoms, backbone atoms, or alpha carbon atoms.
Next (see Figure 1B), the algorithm selects which atom pairs will actually be used for calculations (the pairs of equivalent atoms are inferred from the residue equivalence provided in the alignment file). The user option at this stage is to use only atom pairs whose members are both inside their corresponding spheres (i.e. the intersection of the two lists or "intersection set"), or pairs with at least one member inside one of the spheres (the "union set").
After selecting the atom pairs, the sets of points (corresponding to the relative locations of the atom centers) are placed in the same coordinate system, with the centers of both spheres at the origin (see Figure 1C).
Finally, the quaternion method [15] is used to find the rotation of one set of points around the origin that minimizes the RMSd between the two sets of points. The final RMSd value is used as a measure of the perturbation for the region surrounding the residue considered. We will refer to it as the "truly-local score".
In this way a score value is obtained per each non-gapped position of the alignment, for the comparison of two structures. If there are more than two structures to analyze, every one of them is iteratively compared against the rest, and the final score per alignment position is the average of scores obtained from all the pairwise comparisons.
Heuristic penalties: mutations or the use of the "intersection set" (see above) can result in unpaired atoms. In this case, the user can choose either to ignore these atoms or to introduce a heuristic penalty to account for their presence in one neighbor list but not in the other. It can be argued that these unpaired atoms actually reflect a difference in a residue neighborhood from one protein to the other, which should be accounted for. The penalty function is largest when the distance from the unpaired atom to the center of the sphere is zero, and it decreases smoothly to zero when this distance is equal to the sphere radius:
where pmax is a maximum penalty value (set by the user), r stands for the sphere radius, and du is the distance from the unpaired atom to the center of the sphere. The score formula, introduced in Figure 1, is modified to account for the penalty in this way:
where n is the number of paired atoms and m is the number of unpaired atoms.
Pseudo-local score
COLORES also calculates an alternative score, which does not make use of a sphere nor any "neighborhood" concept. A structural superposition of the two whole structures is performed as usual: the complete structures are used, the rotations are centered at the center of mass of the structures, and the RMSd for the whole structure is minimized using the quaternion method.
Then, the distance between equivalent alpha carbon atom pairs (from residues corresponding to the same alignment position) is taken as an alternative "pseudo-local" score. This measure is, in some way, the one used when a researcher superposes two structures and visually analyses the distance between the backbones. This score is not presented as a novelty but as a reference of one of the current ways of looking for "local" structure alteration or conservation, and to show the significant improvement represented by the truly-local score. The pseudo-local score suffers from the drawbacks mentioned before for whole-molecule superpositions, i.e. that portions of the structure with little or no perturbation are superposed badly in order to improve the superposition of those portions with important perturbations.
Output visualization
Along with the detailed log file and the data file containing the scores per alignment position, a structure file is produced whose atom coordinates can be selected to be either:
a) The coordinates of the first protein of the alignment.
b) Average coordinates corresponding to each position: the coordinates of equivalent backbone atoms are averaged after their structures have been superposed. The average structure is only created for visualization purposes; it is not used for the calculation of the scores.
The truly-local and/or the pseudo local score are also saved in the structure file, in the data column corresponding to the b-factor (if both scores are selected for display, two identical structures are saved, each one of them with a different score in the b-factor column). This allows displaying the scores on the structure, by employing different colors and different backbone widths; both means of visualization are used simultaneously to aid the visual inspection of the results. This also allows the user to modify further this display by using one or the other property to show a different specific feature, while still showing the COLORES scores with the remaining property. The three-dimensional presentation of numerical scores and other information as colors or shapes has proven to be a powerful tool for analyzing this kind of data (e.g. see [16]), because it allows to appreciate spatial relationships that are not so evident in a bi-dimensional graph.
A script file for RASMOL is created in order to launch the display automatically, and to spare the user the need to learn how to configure the program (many other programs can also perform a similar display if properly set up by the user).
For a better display, the scores saved in the structure file differ from the original ones (saved in the data file), in the sense that they are internally normalized and corrected to account for gapped positions (as fully explained in the program documentation).
Results and discussion
We have used COLORES to analyze examples taken from different works on protein structure research. Here we provide a detailed description of two cases, for which we compare COLORES with other structural comparison software. Additional examples are available at the COLORES webpage.
Unless stated otherwise, in the examples provided we have used default parameters for the program, which can be found (and modified) in a key file accompanying the executable files distribution. The default values for the more significant options are: (a) the sphere has a radius of 10 A and (b) it is centered at the alpha carbon atom of the residues, (c) the atoms eligible for comparison are all backbone atoms, (d) only atom pairs whose members are inside the spheres in both structures are included in the comparison ("intersection set"), and (e) there is no heuristic penalty to account for unpaired atoms.
Values near 10 A are usually used to define a limit in comparable studies, were the goal is also to reduce the scope of a calculation to the relevant neighborhood, like the cut-offs for atomic interacting forces in molecular mechanics calculations, or the spatial boundary for the Ooi's number [17]. The use of backbone atoms is an appropriate and popular choice for protein structure comparison, although in some cases the alternative possibilities are better suited, for instance when comparing prediction models that are made only of alpha carbon atoms traces, or for the analysis of two very similar structures that may require using all heavy atoms. Regarding the remaining choices, which are considering only atom pairs whose members are inside the spheres in both structures and not using heuristic penalties to account for unpaired atoms, both obey to the criteria of keeping the calculation as simple as possible, in order to make it more transparent to the novel user and avoiding the introduction of additional parameters.
This is a "first approach" and all-purpose set of choices that the user, after an initial test run, in some cases may change to address better his/her specific protein model, goals, and personal criteria.
Protein structure prediction
The assessment of protein structure predictions (models) is an area where our algorithm can make a significant contribution. A three-dimensional structural model of a protein is a powerful asset in the investigation of its biological function (for instance, see [18,19]), but producing such a model through experimental determinations is not always easy or even possible. As a result, powerful programs to produce theoretical predictions are being developed (for example, ROSETTA [20,21]). The different prediction tools are contrasted in the Critical Assessment of Protein Structure Prediction (CASP) a community-wide experiment where sequences of proteins, whose experimental structures will be released soon, are communicated to groups working in the field so that they can make their predictions [22,23]. Original algorithms have been developed and used for the analysis of CASP predictions [24,25]. COLORES is also a valuable tool for this purpose, because it can help to compare both visually and analytically different predictions for a given target.
We will compare the results from COLORES and the RMS/Coverage method. The latter is adequately explained in references [25,26]; in brief, it reports, for a given fraction of the protein residues (coverage), which combination of residues exhibits the best superposition.
The CASP prediction T0030AB807 is analyzed using this method in [25]. The main conclusions of the RMS/Coverage analysis are:
a) For superpositions of up to four residues, the zone around residues 20-23 exhibits the best superposition.
b) For superpositions comprising between 5 and 18 residues, the best superposition primarily involves residues in a hairpin centered on residues 48-49 (from 11 to 18 residues, however, a separated short stretch around residue 26 is also included). For superpositions involving 19 residues or more, a different set of residues comprise the best superposition, this leads to the conclusion that the hairpin structure is well predicted locally but not with respect to the rest of the structure.
c) For superpositions involving 19 residues or more, residue stretches corresponding to four different protein zones integrate the best superposition set. All these stretches grow simultaneously along with the increase in coverage.
In Figure 2 we show the COLORES results from comparing the predicted structure with the experimentally determined target structure. It can be seen that the residue neighborhoods with smallest truly-local scores are those around residues 1, 12, 48, 26-28, 20-23, and 66; three of them correspond to the zones reported in the RMS/Coverage results, and three are new. Zones found by COLORES but not by RMS/Coverage are those still well predicted locally but ranking second to other with the same coverage (because RMS/Coverage only reports the best).
Figure 2 Assessment of a protein structure prediction. (A) COLORES comparison of CASP prediction T0030AB807 (coordinates from [22]) against an average structure from a set of 15 NMR experimental structures (PDB code: 1FGP) of the target sequence. The truly-local score is displayed on the left side and the pseudo-local score on the right side. The higher scores (higher local structural differences) are represented by a thicker backbone trace and colors closer to red in the spectrum, while lower scores are represented by a thinner trace and colors closer to blue. (B) The profile of different scores along the protein sequence: COLORES "truly-local" score (red) using standard values and MOLMOL "local RMS" score (blue). Scores have been normalized for a better contrast.
There are differences in the way that residues are included in the lists for the two programs, and this is reflected in the output. For instance, for the loop around residues 48-49, RMS coverage shows most of it in the best coverage list, but COLORES just marks a low score for residue 48. This is because the neighborhood of residue 48 is composed almost entirely of residues in the loop, while the neighborhoods of the remaining residues of the loop include atoms from other parts of the protein. This means that residue 48 has a relatively better predicted neighborhood, while the neighborhoods of the other residues in the loop include both well predicted and poorly predicted protein regions.
RMS/Coverage and COLORES are similar in one feature: they report results from the superposition of atoms that belong to a list of equivalent atoms. The main difference is that RMS/Coverage atom lists are made from a combination of residues taken from any part of the protein, provided that they exhibits the lower RMSd after superposition; while COLORES lists belong to atoms surrounding a certain residue. The other difference is that if two different protein zones of a similar size are especially well predicted, RMS/Coverage will report just the best superposition, while COLORES will allow noticing the two of them due to their low score.
RMS/Coverage can indicate when a part of the protein is well predicted locally and not with respect to the rest of the protein, because a single zone will have the best RMSd at low coverages but not at larger coverages. On the other hand, the truly-local score of COLORES can show multiple zones that have been well predicted locally, but it does not indicate how these zones have been predicted in the context of the rest of the structure. This can be alleviated partially by looking at the pseudo-local score, as it is based on a whole-molecule superposition. When a neighborhood is well predicted locally (i.e. it has a low truly-local score), if it has a high pseudo-local score it can also be concluded that it is badly predicted with respect to the rest of the structure.
The RMS/Coverage method sometimes reports a single "best superposition" for a given coverage, which if formed by unrelated structure patches (e.g. the two zones around residues 26 and 48 at a coverage of 18 residues, for the present example). These zones are not sequentially close, neither are they near in three-dimensional space, nor belong to the same secondary structure element. What can be deduced from the fact that, when arbitrarily grouped and separated from the rest of the structure, these two structure patches superpose well? Since COLORES reports results of a zone that represents a spatial neighborhood of a residue, its unit of comparison always has an objective interpretation. Besides, when reporting one of these artificial merges, RMS/Coverage may overlook a zone with more structural significance and a good local superposition that does not have the single best RMSd for the same coverage level; this is not likely to happen when using COLORES.
Summing up, COLORES offers two main advantages:
- When two or more zones of similar size have been well predicted, COLORES reports all of them simultaneously.
- COLORES reports a definite score for each residue. In addition, the scores correspond to a protein zone that has a significant meaning (a fixed-size three-dimensional neighborhood of a residue).
A secondary advantage of COLORES versus RMS/Coverage is that COLORES is actually available for download and use, while RMS/Coverage is not available as a software application (just its results on the analysis of past CASP predictions).
It is also worthwhile to compare COLORES with MOLMOL [27], which is a software widely used for structure visualization and comparison. To compare two or more structures, MOLMOL calculates a "local RMSd" by iteratively superposing all combinations of three contiguous residues, and then assigning the RMSd value to the middle one. MOLMOL also calculates a score named "average global displacements" which is the same as the pseudo-local score calculated by COLORES.
When the MOLMOL local RMSd is calculated for our present example (see figure 2b), it can be seen that it detects several three-residue-long windows of low RMSd, being the one of lowest value around residue 20 (as reported independently by the RMS-Coverage method at a coverage level of three residues). But all the other zones which are well predicted and detected by the other methods (like the loop around residue 48) cannot be found using MOLMOL local RMSd. This is because RMS/Coverage and COLORES can take into account bigger sets of atoms. Therefore, MOLMOL shares with COLORES the property of reporting secondary well-predicted zones, but it is restricted to analyze only very low linear stretches of three residues. The idea behind COLORES is to enclose a significant neighborhood, big enough to include atoms that do not necessarily belong to very close and sequentially connected residues.
Macromolecular movements
We have found that COLORES is also especially suited to analyze concerted molecular movements that involve a hinge or shear movement of an entire protein domain [28]. In these cases, standard "whole molecule" superposition is doomed to fail, because there is no global similarity between the two related structures. In contrast, local superposition can sharply differentiate which zones have maintained its local structure and where the structural alteration (allowing the movement) has occurred. We have tested the program against several examples from the Database of Macromolecular Movements [29,30]. Here we detail the example of the calmodulin protein.
The unligated form of Calmodulin is composed of two globular domains connected by a long helix. The protein can bind peptide helices by closing the two domains in a hinge motion, which breaks the long helix in two minor helices with a strand in between.
The standard whole molecule superposition displayed in figure 3a clearly shows the inadequacy of this approach to differentiate portions of the structure with little or no perturbation (i.e. the globular domains) from the connecting helix, which does suffer important perturbations. This is also reflected by the profile of the pseudo-local score on figure 3b (right). On the contrary, the truly-local score shown in figure 3b (left) clearly discriminates these zones.
Figure 3 Macromolecular movement of calmodulin. A) The closed state of calmodulin (left, PDB code: 1MXE), the open state (right, PDB code: 3CLN) and a standard whole molecule superposition (middle). B) The truly-local and pseudo-local COLORES scores displayed on the closed form. On the left: the local score showing two structurally well-conserved regions (thin) and one disturbed hinge (thick); the yellow helix is the bonded peptide. On the right side: the global score, which fails to identify these key structural features (displayed on the same structure, peptide is omitted). C) The profile of different (normalized) scores along the protein sequence: COLORES "truly-local" score (red) using standard values, COLORES "truly-local" score using heuristic penalties (green), and MOLMOL "local RMS" (blue).
This example shows that our algorithm may contribute to discriminate the unaltered domains from hinge or otherwise structurally altered regions, and therefore to detect evidence of this kind of molecular movements. Moreover, this is achieved without employing any particular knowledge about the protein function or structure. Considering that the number of existing protein structures increases exponentially, and concurrently more structures belonging to the same protein but determined (or predicted) under different circumstances are available, COLORES may help to find and even to automate the process of molecular movement detection as a complement of other tools like the Sieve-Fit Procedure [31] or the Multiple Linkage Clustering [32].
Regarding the comparison with other software, as a program for calculating RMS/Coverage is not available, we cannot provide an actual analysis. Nevertheless, it is evident that it would report a single globular domain having the best coverage, but not both of them simultaneously. In Figure 3c we compare COLORES with MOLMOL. It can be seen that MOLMOL local RMSd score reports three highly perturbed residues (79-81), while COLORES reports the 75-84 residue stretch. MOLMOL only detects coarse main chain alterations from one structure to the other, in the center of the hinge where the long helix is broken. On the other hand, COLORES also incorporates the changes in the neighborhood of all the residues along the hinge, so it allows defining this region entirely. Independently, the hinge region has been defined by the residues with the largest torsion angle differences (reported to be located in residues 72 to 82) [30], confirming the better sensitivity of the COLORES analysis.
It is interesting to note, before leaving this example, that when a heuristic penalty is introduced in the truly local score (or when the "union set" is selected), the output is significantly changed. Two zones in each domain increase their score, showing that the neighborhood of the corresponding residues was actually changed; this is due to atoms from the opposite domain, which moved from a nearby position in the closed form to a long distance in the open form.
Conclusion
The comparison of protein structures is an established tool for investigating biological function, macromolecular structure, protein evolution, etc. The superposition of entire structures is the standard approach to initiate this analysis but, in some cases, it can produce misleading results.
A local approach for structural comparison can lead to a better insight and discrimination of perturbed against unchanged portions of the structure. Local comparison has been used mostly in the area of structure-based sequence alignment, by employing approaches suited for that particular purpose, but not for the general description of how each zone of a protein changes between two or more structures.
We have developed an algorithm to describe local perturbations of protein structures in quantitative rather than descriptive terms. The method is applicable to any situation and its results are comparable between very different cases. Besides, a means of analyzing its results visually is provided by the program that implements the algorithm; this is a valuable asset in order to interpret three-dimensional results. The analysis of structural perturbations is not a task that can be done only with numbers and tables; sometimes it is necessary to use visual inspection to discriminate what is relevant, and to search for relationships between structural conservation/perturbation and (bio)chemical modifications, biological function, changes in the crystal contacts, etc. We have compared our results with other existing software to show that the present method offers a useful alternative for the analysis of protein structures.
The examples provided in this article, and others available in the software webpage, show that the program can be used easily to perform standard tasks on protein structural research, like:
a) On experimentally generated structures: the comparison of mutant and wild type proteins, the comparison of enzymes in open/close conformation (or proteins with/without ligand), the comparison of protein structures determined from different sources, and the comparison of structurally similar proteins in general.
b) On computationally generated structures: the assessment of structure alterations after molecular mechanics manipulations like minimization, molecular dynamics simulation, etc.; the comparison of different three-dimensional structure predictions either among them or against the target structure.
COLORES can also be used in less standard applications, like the definition of evolutionarily conserved cores, or the identification of zones implicated in macromolecular movements.
COLORES provides numerical results that can be used for quantitative purposes, and are comparable even across different protein structural families. Its scores are also displayed over the structures under study, in a way that can be interpreted quickly and easily. These characteristics make it both suitable for "first glance" purposes when approaching to a novel system under study, and for more specific analytic tasks. Its local approach facilitates finding which zones have been perturbed and how much, particularly when using the truly-local score.
It is important to note that, in its present implementation, COLORES is only able to calculate trivial alignments for pairwise comparisons, when they involve proteins having the same sequence length, and at most a few point mutations. For comparisons involving multiple structures and/or notoriously different sequences, COLORES does not perform nor improve an alignment between the structures under analysis; but instead it requires the alignment as an input, as described before. We are aware that our comparison method could be developed further as a structural alignment algorithm, and although we may explore that possibility in the future, actually a variety of quite well-developed options are already available for performing the alignment of two protein structures [33,34].
Finally, we would like to consider another possible extension of the method. It has been established that cutoff distances for physical potentials are useful to speed molecular mechanics simulations [35]. In these cases, a spherical limit defines a neighborhood of atoms that effectively interact with the atom whose energetic contribution is being calculated. An energetic local score could be calculated from the difference between single point calculations for all the atoms inside the spheres used in COLORES. This would complement the present geometry-based local score with an energetic one. This, in turn may be useful to assess if a geometrical perturbation has an energetic counterpart. For instance, a conservative mutation may only produce an increment in the geometric score but not in the energetic score, while a non-conservative mutation may have an impact on both.
Availability and requirements
Project name: COLORES software
Project home page: . The files used in the examples described in this paper, among others, can also be downloaded from that Internet address, in order to analyze them interactively.
Operating system(s): Executable files tested for Windows or Linux are available. The source code was written to be platform independent, and it is also available.
Programming language: ANSI standard C.
Other requirements: A software capable of showing temperature factors on protein structures as a scale of variable color and backbone width. It is not required for COLORES to run (nor to analyze results analytically), but just in order to check results graphically. It is convenient to use RASMOL, since COLORES automatically calls the program and execute the proper commands for a better display, with no need of user intervention. Other software (e.g. MOLMOL [27]) can be used instead, but in this case, the user must write the appropriate script file or commands.
License: Free for academic purposes.
Any restrictions to use by non-academics: contact the authors.
Authors' contributions
Both authors developed the method. ML developed and tested the software under the supervision of JE. ML wrote the manuscript and JE edited the final version.
Acknowledgements
ML is a fellow of the Comisión de Investigaciones Científicas de la Provincia de Buenos Aires (CIC); JE is a researcher of the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET). This work was supported by Universidad Nacional de Quilmes, Agencia Nacional de Promoción Científica y Tecnológica, and Fundación Antorchas.
==== Refs
Paoli M Liddington R Jeremy T Wilkinson A Dodson G Crystal Structure of T State Haemoglobin with Oxygen Bound At All Four Haems J Mol Biol 1996 256 775 792 8642597 10.1006/jmbi.1996.0124
Bruschweiler R Efficient RMSd Measures for the Comparison of Two Molecular Ensembles Proteins 2003 50 26 34 12471596 10.1002/prot.10250
Jewett AI Huang CC Ferrin TE Minrms: an Efficient Algorithm for Determining Protein Structure Similarity Using Root-Mean-Squared-Distance Bioinformatics 2003 19 625 634 12651721 10.1093/bioinformatics/btg035
Yang AS Honig B An integrated approach to the analysis and modeling of protein sequences and structures. I. Protein structural alignment and a quantitative measure for protein structural distance J Mol Biol 2000 301 665 78 10966776 10.1006/jmbi.2000.3973
Kobayashi N Nobuhiro G A method to search for similar protein local structures at ligand-binding sites and its application to adenine recognition Eur Biophys J 1997 26 135 144 9232842 10.1007/s002490050065
Koehl P Protein structure similarities Curr Opin Struct Biol 2001 11 348 353 11406386 10.1016/S0959-440X(00)00214-1
Leibowitz N Nussinov R Wolfson HJ MUSTA - a general, efficient, automated method for multiple structure alignment and detection of common motifs: application to proteins J Comput Biol 2001 8 93 121 11454300 10.1089/106652701300312896
Leluk J Konieczny L Roterman I Search for structural similarity in proteins Bioinformatics 2003 19 117 124 12499301 10.1093/bioinformatics/19.1.117
Lehtonen JV Denessiouk K May AC Johnson MS Finding local structural similarities among families of unrelated protein structures: a generic non-linear alignment algorithm Proteins 1999 34 341 55 10024021 10.1002/(SICI)1097-0134(19990215)34:3<341::AID-PROT7>3.0.CO;2-Z
Ochagavia ME Richelle J Wodak SJ Advanced pairwise structure alignments of proteins and analysis of conformational changes Bioinformatics 2002 18 637 40 12016063 10.1093/bioinformatics/18.4.637
Orengo CA Taylor WR A Local Alignment Method for Protein-Structure Motifs J Mol Biol 1993 233 488 497 8411157 10.1006/jmbi.1993.1526
Szustakowski JD Weng Z Protein structure alignment using a genetic algorithm Proteins 2000 38 428 440 10707029 10.1002/(SICI)1097-0134(20000301)38:4<428::AID-PROT8>3.0.CO;2-N
Ye Y Jaroszewski L Li W Godzik A A segment alignment approach to protein comparison Bioinformatics 2003 19 742 9 12691986 10.1093/bioinformatics/btg073
Bernstein HJ Recent changes to RasMol, recombining the variants Trends Biochem Sci 2000 25 453 455 10973060 10.1016/S0968-0004(00)01606-6
Kearsley SK On the Orthogonal Transformation Used for Structural Comparisons Acta Crystallogr A 1989 45 208 210 10.1107/S0108767388010128
Altman RB Hughes C Gerstein M Methods for displaying macromolecular structural uncertainty: application to the globins J Mol Graph 1995 13 142 52 7577841 10.1016/0263-7855(95)00002-N
Nishikawa K Ooi T Radial locations of amino-acid residues in a globular protein – correlation with the sequence J Biochem 1986 100 1043 1047 3818558
Lee SJ Sekimoto T Yamashita E Nagoshi E Nakagawa A Imamoto N Yoshimura M Sakai H Chong KT Tsukihara T Yoneda Y The Structure of Importin-β Bound to SREBP-2: Nuclear Import of a Transcription Factor Science 2003 302 1571 1575 14645851 10.1126/science.1088372
Wedemayer GJ Patten PA Wang LH Schultz PG Stevens RC Structural insights into the evolution of an antibody combining site Science 1997 276 1665 1669 9180069 10.1126/science.276.5319.1665
Bradley P Chivian D Meiler J Misura KM Rohl CA Schief WR Wedemeyer WJ Schueler-Furman O Murphy P Schonbrun J Strauss CE Baker D Rosetta predictions in CASP5: successes, failures, and prospects for complete automation Proteins 2003 53 457 68 14579334 10.1002/prot.10552
Kuhlman B Dantas G Ireton GC Varani G Stoddard BL Baker D Design of a novel globular protein fold with atomic-level accuracy Science 2003 302 1364 8 14631033 10.1126/science.1089427
Tramontano A Morea V Assessment of Homology-Based Predictions in CASP5 Proteins 2003 53 352 368 14579324 10.1002/prot.10543
Protein Structure Prediction Center (CASP website)
Zemla A LGA: a method for finding 3D similarities in protein structures Nucleic Acids Res 2003 31 3370 3374 12824330 10.1093/nar/gkg571
Hubbard TJ RMS/coverage graphs: a qualitative method for comparing three-dimensional protein structure predictions Proteins 1999 3 15 21 10526348 10.1002/(SICI)1097-0134(1999)37:3+<15::AID-PROT4>3.0.CO;2-Z
RMS/Coverage method website
Koradi R Billeter M Wüthrich K MOLMOL: A program for display and analysis of macromolecular structures J Mol Graph 1996 14 51 55 8744573 10.1016/0263-7855(96)00009-4
Krebs WG Tsai J Alexandrov V Junker J Jansen R Gerstein M Tools and databases to analyze protein flexibility; approaches to mapping implied features onto sequences Methods Enzymol 2003 374 544 84 14696388
Database of Macromolecular Movements
Echols N Milburn D Gerstein M MolMovDB: analysis and visualization of conformational change and structural flexibility Nucleic Acids Res 2003 31 478 82 12520056 10.1093/nar/gkg104
Gerstein M Chothia CH Analysis of protein loop closure: two types of hinges produce one motion in lactate dehydrogenase J Mol Biol 1991 220 133 149 2067013 10.1016/0022-2836(91)90387-L
Boutonnet NS Rooman MJ Wodak SJ Automatic analysis of protein conformational changes by multiple linkage clustering J Mol Biol 1995 253 633 47 7473739 10.1006/jmbi.1995.0578
Barton GJ Protein Sequence Alignment Techniques Acta Crystallogr D 1998 54 1139 1146 10089490 10.1107/S0907444998008324
Carugo O Pongor S Recent progress in protein 3D structure comparison Curr Protein Pept Sc 2002 3 441 449 12370006 10.2174/1389203023380530
McCammon JA Harvey SC Dynamics of proteins and Nucleic Acids 1987 Cambridge: Cambridge University Press
|
16159393
|
PMC1239910
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Sep 13; 6:226
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-226
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2271616229610.1186/1471-2105-6-227Research ArticleSystematic survey reveals general applicability of "guilt-by-association" within gene coexpression networks Wolfe Cecily J [email protected] Isaac S [email protected] Atul J [email protected] Children's Hospital Informatics Program and Harvard MIT Division of Health Sciences and Technology, 300 Longwood Avenue, Boston, MA 02115, USA2 Hawaii Institute of Geophysics and Planetology, University of Hawaii at Manoa, 1680 East West Road, Honolulu, HI, 96822, USA3 Stanford Medical Informatics, Department of Medicine and Pediatrics, Stanford University School of Medicine, 251 Campus Drive, Room X-215, Stanford, CA 94305-5479, USA2005 14 9 2005 6 227 227 22 6 2005 14 9 2005 Copyright © 2005 Wolfe et al; licensee BioMed Central Ltd.2005Wolfe et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Biological processes are carried out by coordinated modules of interacting molecules. As clustering methods demonstrate that genes with similar expression display increased likelihood of being associated with a common functional module, networks of coexpressed genes provide one framework for assigning gene function. This has informed the guilt-by-association (GBA) heuristic, widely invoked in functional genomics. Yet although the idea of GBA is accepted, the breadth of GBA applicability is uncertain.
Results
We developed methods to systematically explore the breadth of GBA across a large and varied corpus of expression data to answer the following question: To what extent is the GBA heuristic broadly applicable to the transcriptome and conversely how broadly is GBA captured by a priori knowledge represented in the Gene Ontology (GO)? Our study provides an investigation of the functional organization of five coexpression networks using data from three mammalian organisms. Our method calculates a probabilistic score between each gene and each Gene Ontology category that reflects coexpression enrichment of a GO module. For each GO category we use Receiver Operating Curves to assess whether these probabilistic scores reflect GBA. This methodology applied to five different coexpression networks demonstrates that the signature of guilt-by-association is ubiquitous and reproducible and that the GBA heuristic is broadly applicable across the population of nine hundred Gene Ontology categories. We also demonstrate the existence of highly reproducible patterns of coexpression between some pairs of GO categories.
Conclusion
We conclude that GBA has universal value and that transcriptional control may be more modular than previously realized. Our analyses also suggest that methodologies combining coexpression measurements across multiple genes in a biologically-defined module can aid in characterizing gene function or in characterizing whether pairs of functions operate together.
==== Body
Background
From the very start of the high-throughput microarray expression revolution it was understood [1,2] that guilt-by-association was a powerful heuristic to both explain why genes might have correlated expression in a set of experiments and infer what might be the function of a gene coexpressed with genes of better known function. As gene expression data have increased in numbers and quality, a variety of investigations have been leveraged from this GBA heuristic. Analyses of gene coexpression [3-7] have demonstrated that clusters with similar overall expression are often enriched for genes with similar functions, consistent with the hypothesis of modularly-behaving gene programs, where sets of genes are activated in concert to carry out functions.
GBA has also been exploited highly successfully by investigators who have used a priori determined modules or gene sets and assess if these sets have statistically significant overrepresentation in the genes changed in groups of arrays [8-15]. By exploiting the insight that subtle but coordinated changes in expression can be detected by combining measurements across multiple members of a functional module, these focused studies have successfully found specific modules that are important in diabetes [12], aging [13], and cancer [10,11,14,15], or assigned functions to previously uncharacterized genes in yeast [8,9]. These approaches essentially integrate two frameworks of viewing gene function [16], one framework reflected in module sets that are derived from prior biological knowledge and another framework from the characteristics of gene expression data.
These studies reflect two bidirectional uses of GBA: either using coexpression to define the members of functionally related sets or using sets to define function of coexpressed genes. That is, the first uses prior gene expression data and the second uses prior biological knowledge. We extend these approaches, taking the a priori framework of knowledge available in Gene Ontology (GO) [17] to systematically explore the breadth of GBA across a large and varied corpus of expression data to answer the following questions. 1) To what extent is the GBA heuristic broadly applicable to the transcriptome and GO? 2) In the GBA heuristic, how well does coexpression inform function and vice versa? 3) Which GO heuristics are the most interrelated as measured by a GBA metric?
The testbed for evaluating the extent and organization of GBA were five coexpression networks, constructed using 8341 microarrays representing a variety of tissue types and conditions. For each network we determine whether coordinated coexpression can be detected across multiple genes of each GO-defined module. Our approach is better suited than clustering to systematically examine GBA because it allows for pleiotropy: it does not assign genes to a single function or a single cluster but rather calculates a probabilistic score between each gene and each GO category. This approach better captures complex interrelationships [18], such as genes that code for proteins with multiple functions [19]. We discover that there is a ubiquitous signature of functional association in all of the coexpression networks in that the genes in a module often demonstrate higher-than-expected numbers of coexpressed genes belonging to that same module.
To further illustrate the breadth of GBA, we present the extent of which coexpression implicates members of three sets of genes that are usually thought of as belonging to a very specific biological context: skeletal development, neuropeptide receptor activity, and feeding behavior. We show that these Gene Ontology categories, as well as hundreds of other categories, are associated with coordinated expression patterns across the variety of tissue types and conditions in our data.
Results and discussion
Analysis of coexpression networks
We constructed five different coexpression networks (four single-species networks and one unified multi-species network), which are graphs where genes are nodes and the edges are represented by values reflecting the significance of coexpression between a pair of genes. We selected mammalian organisms for which extensive and diverse microarray data were available on four Affymetrix platforms in the Gene Expression Omnibus (GEO): Homo sapiens (HG-U95A and HG-U133A), Mus musculus (MG-U74A), and Rattus norvegicus (RG-U34A). Orthologs between these organisms were obtained from HomoloGene and from this information, 6624 "metagenes" (hereafter referred to as genes) were defined consisting of sets of orthologous genes across at least two different organisms on the chosen microarray platforms. The multi-species network integrates the 8341 microarrays from all four Affymetrix platforms into a unified coexpression network, using order statistics [6] to assign coexpression P-values (Pc) between all possible pairs of genes. Previous work [6] suggested that by using the signal of evolutionary conservation in a multi-species coexpression network the effect of noise is reduced and the significance of functionally important gene pairs is enhanced, although this approach is only valid when homologous genes share functionality. The four single-species coexpression networks were calculated from Pearson correlation coefficients between genes, in each case using only data from one of the four Affymetrix platforms.
For each network, we next construct a probabilistic score between each gene and each GO category that reflects the tendency for the genes in that GO set to be highly coexpressed with the selected gene. For each gene, a list of all other linked genes was ordered according to most significant coexpression Pc-value (multi-species case) or highest correlation coefficient (single-species cases). For a given GO category, each gene in a coexpression network was analyzed using the hypergeometric distribution to determine if the GO set was overrepresented towards the top of the list of more highly correlated genes (Figure 1a). This process produces a gene set coexpression enrichment P-value (Pe) between each of the genes and each of the GO sets. The Pe-value between a particular gene and GO category is a probabilistic score for that pair, with lower (more significant) Pe-values reflecting greater coexpression enrichment of that GO module. We demonstrate below how these Pe-values have utility in identifying gene function, indicating the ubiquity of GBA across most GO categories, and how they quantify the interrelationships between GO categories (Figure 1b).
Figure 1 Schematic representation of the steps in our analyses. (a) Example flow chart of the different steps for calculating gene set coexpression enrichment Pe values between each of the 6624 genes in the multi-species network and 902 GO sets. For each gene mi we use the hypergeometric distribution to calculate a coexpression enrichment Pe-value (Pe(mi, gj)) for whether GO set gj was significantly overrepresented in the top 250 genes with smallest Pc-values to mi. (b) The four steps in our analyses. 1. A coexpression network is generated with Pc values (multi-species network) or correlation coefficients (single-species network) scoring coexpression between gene pairs. 2. Coexpression enrichment Pe values are calculated between each gene and each GO category, such as between GO category 1 and genes A, B, and C and between GO category 2 and genes A, B, and C. 3. A score reflecting GBA is calculated for each GO category (e.g., GO category 1). 4. The interrelationship between pairs of GO categories is quantified, such as that between GO category 1 and GO category 2, which are sibling categories in a Gene Ontology graph, sharing GO category 3 as a common parent.
Functional relevance of coexpression enrichment values
With a network of coexpression relations computed between pairs of genes, and networks of coexpression enrichment relations computed between all pairs of GO categories and genes, we evaluated how reliably coexpression enrichment Pe-values for a GO category identify genes annotated with that function. Each GO category contains a set of specific Pe-values to score relations to all genes. Taking one GO category at a time, we calculated the true and false positive rates for identifying genes annotated with that GO category at threshold Pe-values throughout its range. We plotted these true and false positive rates on Receiver Operating Characteristic (ROC) curves [20] for each GO category. If there were a threshold Pe-value below which all genes are annotated with the correct function, and above which no genes are annotated with the correct function, then the area under such an ideal ROC curve would be 1. An area of 0 would mean that identifying annotated genes using Pe performs perfectly incorrectly, and an area of 0.5 indicates no overall identification efficiency using Pe (performance equivalent to random chance). Thus the area under an ROC curve for each GO category is a metric for GBA, scoring how well coexpression enrichment Pe-values perform as a "self-diagnostic" for the genes annotated to a category.
For the multi-species network, the ROC curves for the mitochondrion, skeletal development, neuropeptide receptor activity, and feeding behavior are all concave downward and plot above the diagonal (Figures 2a–d) with ROC areas greater than 0.5, indicating GBA for these GO categories. Skeletal development, neuropeptide receptor activity, and feeding behavior are usually thought of as belonging to a very specific biological context, yet genes in these categories are coexpressed across a wide range of samples from 8341 microarrays.
Figure 2 Examples from the multi-species network. (a-d) Self-diagnostic Receiver Operating Characteristic (ROC) curves for the GO categories shown above.
These patterns are typical of most GO categories. Self-diagnostic ROC areas for all of the GO categories in the multi-species network (see additional file 1: self-diagnostic ROC areas and 95% errors from the multi-species network), organized by the three domains of biological process, cellular component, and molecular function (Figures 3a–c), have distributions centered near 0.7, which is above the mean of 0.5 for the case where there would be no useful information in coexpression enrichment. This upward shift in the distributions indicates that for most GO categories, GBA is applicable and coexpression enrichment adds knowledge about gene function. This knowledge is not perfect: the ROC areas are all less than 1, and for many categories the large numbers of false positives at specific Pe-value thresholds would limit the practical application of using this method to identify gene function. But nonetheless, a probabilistic signature of GBA is present. Equivalently, the members of a GO module as a whole tend to have more significant Pe-values for that category than the non-members, because the ROC area also measures the probability that given randomly drawn pairs from two groups, one of members of a GO set and another of nonmembers, Pe(member) <Pe(nonmember) for coexpression enrichment of that set.
Figure 3 Histograms of self-diagnostic ROC areas for the multi-species network. (a) Histogram for biological process GO categories. (b) Histogram for cellular component GO categories. (c) Histogram for molecular function GO categories. (d) Histogram for randomized GO sets. (d) Histogram for a randomized multi-species coexpression network.
The results were tested by taking the multi-species coexpression network and applying the same analysis with randomized GO sets. The population of self-diagnostic ROC areas for the randomized GO sets is centered at 0.5 (Figure 3d). The case of a randomized network, with Pc-values permuted between gene pairs, also yields a distribution that is centered at 0.5 (Figure 3e). Thus the upward shift of the true distributions is unlikely to occur by chance.
We tested whether the ROC areas were correlated with other factors (see additional file 2: supplementary methods), but found that the correlations were not strong, ranging between +/-0.2. We tested whether the type of evidence used to construct a GO set, given in the GO evidence codes, has any relation to the ROC areas; whether there was any correlation between the expression levels in a GO category and the ROC areas; whether there was a correlation between the ROC areas and the average number of GO annotations for the genes in each set.
Interrelations between GO categories
To examine which GO categories are the most interrelated, we test whether coexpression enrichment for one GO set can be used to assign genes to a different GO category ("cross diagnostics"). These analyses score how well different GO modules tend to be coexpressed together, such as whether coexpression enrichment for the mitochondrion module is a characteristic of the oxidative phosphorylation module (the multi-species cross-diagnostic ROC area is 0.94 for this case). In one sense, these scores indicate the strength of coexpression links in a network where the graph nodes are GO categories, rather than genes. However, a complication is that pairs of gene sets may significantly overlap in their annotated genes. Therefore, for the multi-species network we present the systematics between pairs of GO categories that are together on the same graph, where GO relationships are defined and provide additional context for interpreting the results. Gene Ontology organizes biological processes, molecular functions, and cellular components separately on three directed acyclic graphs. A parent GO category has a set of more specific children (from those GO categories just one step below a parent on a graph) and more specific descendents (from all GO categories in the entire subgraph below a parent). To test the results, we apply the same analysis with randomized GO sets that are constructed in a manner that mimics the GO mappings.
For cross-diagnostic tests of whether Pe-values of descendent GO categories can correctly identify genes in parent GO categories (Figure 4a), we find that the distribution is shifted above 0.5 (mean = 0.67). However, descendent sets are subsets of parent sets, so it is consistent that this distribution is similar to the patterns in the self-diagnostic ROC areas. We next examine GO categories that are siblings (children of a common parent), since GO children split a parent into distinct and more specialized categories. For sibling pairs (Figure 4b), the shift above 0.5 is less (mean = 0.57). Yet the populations of ROC areas across sibling pairs and across descendent-parent pairs remain more diagnostic than the population across more distantly-related pairs (Figure 4c), which is centered at the expected mean of 0.5 for the case of no interrelation on average.
Figure 4 Histograms of cross-diagnostic ROC areas for the multi-species network. (a) Histogram of ROC areas for whether descendent Pe-values are diagnostic of parent sets. (b) Histogram of ROC areas for cross pairing of sibling categories. (c) Histogram of ROC areas for all cross pairings of categories (excluding parent-descendent pairs) with distances of 3–16 in a GO graph. GO organizes categories as nodes on a graph and calculates the distance between category pairs on the same graph as the minimum number of arcs needed to traverse from one category node to another on the graph. For example, a parent and its child are separated by a distance of one and siblings are separated by distances of two. (d) Histogram of ROC areas for all cross pairings of categories (excluding parent-descendent pairs) with distances of 3–16 in a GO graph, created using randomized GO sets. (e) Histogram of cross-diagnostic ROC areas between GO category pairs (excluding parent-descendent pairs) in the subgraph below immune response. (f) Histogram of cross-diagnostic ROC areas between GO category pairs (excluding parent-descendent pairs) in the subgraph below cell cycle.
However, the distribution (Figure 4c) does display longer tails than for randomized GO sets (Figure 4d), indicating how there is a nonrandom tendency for some of these modules to either be highly coexpressed together (high areas) or not highly coexpressed together (low areas). (Note that increasing the scale in Figure 4d does not reveal any additional detail in the tails of the distribution.) In addition, some subgraphs of GO show uniformly high cross diagnostics, such as the subgraphs under immune response and cell cycle (Figure 4e–f), where there is a signal that modules from the different sub-categories are often coexpressed together in the types of tissues in our analysis.
Of the 812,702 possible cross diagnostic GO pairings, only a small percentage are related by coexpression (e.g., 5% have ROC areas greater than 0.7). As shown in the above analyses, at least some of the positive relationships are consistent with the known biology reflected the Gene Ontology hierarchy.
Reproducibility across different microarray platforms
The patterns found in the multi-species network are highly reproducible in the single-species networks for each of the 4 different microarray platforms. Self-diagnostic ROC areas derived from single-species networks are strongly correlated with the values derived from the multi-species network with correlation coefficients ranging from 0.8 to 0.9 (Figure 5). However, the ROC areas from single-species networks typically are lower than areas from the multi-species network, plotting below a diagonal straight line of slope one and zero intercept. This shift likely arises because the multi-species coexpression network reduces the effects of noise and enhances the ability of the network to link with more significance gene pairs involved in common function [6]. But for our analysis this enhancement is only minor, illustrating how coexpression from a single organism already captures the signal of GBA. The cross-diagnostic ROC areas are also strongly correlated in the five different networks, with correlation coefficients between single-species and multi-species cross-diagnostic ROC areas ranging from 0.7 to 0.9 (data not shown). The interrelatedness between pairs of GO modules is therefore also reproducible across the different datasets on the different platforms.
Figure 5 Plots of self-diagnostic ROC areas from the multi-species network (x-axis) versus ROC areas from a single-species network (y-axis) for each GO category. Each panel examines one of the single-species networks, created using microarrays from the following Affymetrix platforms: HG-U133A (human), HG-U95A (human), MG-U74A (mouse), and RG-U34A (rat). Correlation coefficients are noted in the upper left corner of the plots.
The observed scores appear to reflect the behavior of the transcriptome rather than being dominated by the mix of samples in each of the networks or the choice of microarray platform. Though the GEO data in our study originated from many laboratories with inhomogeneous protocols, our analyses demonstrate how the extent of GBA for each GO module and the interrelatedness between GO module pairs nonetheless have high reproducibility in the networks. The reproducibility between the entire multi-species and single-species networks is lower than the reproducibility of ROC areas (see additional file 2: supplementary methods), demonstrating how a functionally-based analysis enhances the similarity of the signals between different networks. Our results are consistent with a recent study of expression variability across different platforms and laboratories [21] that found highest reproducibility when the analysis was based on biological themes defined by GO.
Conclusion
Our study provides an investigation of the functional organization of five coexpression networks using data from three mammalian organisms. This method integrates information from two different frameworks of viewing gene function [16], one framework essentially from the manual and subjective curation of evidence in the literature into the Gene Ontology hierarchy and another framework from a probabilistic analysis of expression datasets. Across all five networks, we find a signature that coexpression enrichment predicts coannotation across GO categories, and thus the guilt-by-association heuristic is broadly applicable. Although for gene pairs within a specified GO set the coexpression value may only be weak, by combining coexpression measurements across multiple genes in the module, there is a systematic and reproducible signature of functional association. Because the genes in a particular module demonstrate higher-than-expected numbers of coexpressed genes belonging to that same module, the values for gene set coexpression enrichment tend to be predictive of gene function.
It was unexpected that a simple test based on coexpression would have value in assigning genes to so many different types of GO categories. While some GO annotations may themselves have been defined on the basis of expression, there are also many GO annotations that did not necessarily employ expression results, such as the annotations in the cellular component domain, where the population of ROC areas still displays better-than-random ability to correctly identify the genes annotated to GO categories. That some GO categories score better than others likely reflects the characteristics of underlying biological behavior, as the scores of GO categories are reproducible across all of the coexpression networks. This study demonstrates how using coexpression enrichment to assign a probabilistic score between genes and functions can add information about gene function. We note that an analogous data mining approach to ours was previously applied by Lamb et al. [11] to discover that C/EBPβ was a mechanism of cyclin D1 action, using a single module gene set of cyclin D1 target genes. Our more comprehensive study of 902 GO module gene sets suggests this type of approach should also be successful for other biological systems. Our results are in agreement with a recent study [22] that used a support vector machines method on mouse coexpression data and found that genes in many GO biological process categories could be identified as being in those categories. Our results disagree with low degree of GBA found by Clare and King [23], who clustered yeast microarray data and found the clusters did not in generally agree with functional annotation classes. One explanation for this disagreement may be that the use of clustering by Clare and King [23] does not reveal the more subtle signal of GBA that we discover using gene set coexpression enrichment. Another difference may be that our larger and more comprehensive dataset (8341 Affymetrix mammalian microarrays) is better suited to identify GBA.
Our strategy demonstrates that the functions of a cell operate on an exquisitely coordinated level and that the modular character of cell biology [24] is evident across the biologically variable microarray data in our analysis. Within the large scope of the considered GEO samples and GO categories, we find that the guilt-by-association identification of gene function on the basis of expression has universal value. This result provides optimism that high-throughput measurements of gene expression and community-based gene annotation efforts will continue to demonstrate synergy in the collective investigations of cellular physiology and understanding of human diseases.
Methods
Assignment of metagenes
Genes from one organism were associated with their orthologous counterparts in other organisms using HomoloGene (downloaded on June 22, 2004). 6624 "metagenes "were defined as sets of orthologs across at least two organisms with available microarray probes, using cases where no more than one gene was found for each organism. Microarray probes for orthologs were assigned into metagene probe groups. Because some genes have multiple probes on an array, for each of the 6624 metagenes, we considered all combinations of probes across the microarray platforms (see additional file 2: supplemental methods).
Generation of coexpression networks
Microarray data consisted of 8341 arrays from 4 different platforms downloaded from NCBI Gene Expression Omnibus. 2179 arrays were Affymetrix HG-U95A (version 2), 2438 arrays were Affymetrix HG-U133A, 2216 arrays were Affymetrix MG-U74A (version 2), and 1508 arrays were Affymetrix RG-U34A. We normalized expression data on each array by converting values to rank percentile. For the multi-species network, for each probe group we computed Pearson correlation coefficients between other probes on a platform and then ranked these other probes according to their correlations. For each distinct pair of metagene probe groups, a probabilistic method based on order statistics was used to evaluate the probability of observing the ranks by chance (see additional file 2: supplemental methods). This generates coexpression P-values (Pc(mi, mj)) between pairs of metagenes. A unique Pc-value between metagene pairs is selected based on lowest Pc-value obtained from all of the analyzed probe groups, with the philosophy that when coexpression is present, more significant Pc-values will be associated with more accurate probes. Single-species coexpression networks for each of the four different platforms were calculated from Pearson correlation coefficients between gene pairs, limited to those genes also analyzed in the multi-species network and selecting the highest correlation coefficient obtained for the cases where multiple probes are available for gene pairs.
GO gene sets
For each network, gene sets were compiled for 902 GO categories with at least 20 genes in the multi-species network. The graph relationships were obtained from the Gene Ontology MySQL database, downloaded on September 24, 2004. The annotations of genes to GO categories were taken from LocusLink, downloaded on September 27, 2004. Gene Ontology organizes biological processes, molecular functions, and cellular components separately on three directed acyclic graphs, with more general parent categories having subgraphs of more specific descendent categories. The GO true path rule is that annotation to a category implies annotation to all parents and gene products are conventionally annotated just to the most specific levels of the ontology. We associate a gene to a GO set if it is annotated with that GO category in human, mouse, or rat or if it is annotated with a descendent of that GO category. See additional file 2: supplemental methods, for the construction of the randomized GO sets.
Statistical significance of coexpression enrichment of a GO set
For each gene mi, all other linked genes are ranked by the most significant value obtained for coexpression. We use the hypergeometric distribution to calculate a coexpression enrichment P-value (Pe(mi, gj)) for whether GO set gj was significantly overrepresented in the top 250 genes with most significant coexpression values to mi (Figure 1). Similar results were obtained for cases where the number of top ranked genes selected was different (decreased to 100 or increased to 500) or where an enrichment score was based on a normalized Kolmogorov-Smirnov statistic [12].
Receiver Operating Characteristic (ROC) curves
A self-diagnostic ROC curve tests whether the Pe(mi, gj)-values for GO set gj can distinguish genes associated to gj. An ROC curve is constructed for a range of closely spaced Pe(mi, gj)-value cutoffs. At a given cutoff, the true-positive rate (sensitivity) is calculated as the number of genes associated to GO set gj with Pe(mi, gj)-values below the cutoff divided by the total number associated to gj; the false-positive rate (1-specificity) is calculated as the number of genes not associated to a GO set gj with Pe(mi, gj)-values below the cutoff divided by the total number not associated to a GO set. The area is estimated by trapezoidal integration and 95% confidence intervals are also calculated [20]. Cross-diagnostic ROC areas are calculated as above, except we test whether the Pe(mi, gj)-values for gj can distinguish genes associated to a different GO set gk.
Authors' contributions
This study was conceived by AJB and CJW. CJW carried out the analyses, with advice and guidance from AJB and ISK. All authors fully participated in the interpretation of results and the writing of the manuscript.
Supplementary Material
Additional File 1
Self-diagnostic ROC areas and 95% errors from the multi-species network. This file lists ROC areas for the 902 GO categories.
Click here for file
Additional File 2
Supplemental Methods. This file provides supplemental information on the methods used in the analyses.
Click here for file
Acknowledgements
This work as supported by grants from the the NIH National Center for Biomedical Computing (U54 LM008748), National Library of Medicine (2TA5 LM07092-11 and 5T15 LM07092), National Institute of Diabetes and Digestive and Kidney Diseases (K12 DK63696 and R01 DK62948), the Harvard-MIT Division of Health Sciences and Technology, and the Lawson Wilkins Pediatric Endocrine Society.
==== Refs
Brown PO Botstein D Exploring the new world of the genome with DNA microarrays Nature Genet 1999 21 33 37 9915498 10.1038/4462
Quackenbush J Computational analysis of microarray data Nat Rev Genet 2001 2 418 427 11389458 10.1038/35076576
Eisen MB Spellman PT Brown PO Botstein D Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 1998 95 14,863 14,868 9843981 10.1073/pnas.95.25.14863
Tavazoie S Hughes JD Campbell MJ Cho RJ Church GM Systematic determination of genetic network architecture Nature Genet 1999 22 281 285 10391217 10.1038/10343
Wu LF Hughes TR Davierwala AP Robinson MD Stoughton R Altschuler SJ Large-scale prediction of Saccharomyces cerevisiae gene function using overlapping transcriptional clusters Nature Genet 2002 31 255 265 12089522 10.1038/ng906
Stuart JM Segal E Koller D Kim SK A gene-coexpression network for global discovery of conserved genetic modules Science 2003 302 249 255 12934013 10.1126/science.1087447
Lee HK Hsu AK Sajdak J Qin J Pavlidis P Coexpression analysis of human genes across many microarray data sets Genome Res 2004 14 1085 1094 15173114 10.1101/gr.1910904
Hughes TR Marton MJ Jones AR Roberts CJ Stoughton R Armour DC Bennett HA Coffey E Dai H He YD Kidd MJ King AM Meyer MR Slade D Lum PY Stepaniants SB Shoemaker DD Gachotte D Chakraburtty K Simon J Bard M Friend SH Functional discovery via a compendium of expression profiles Cell 2000 102 109 125 10929718 10.1016/S0092-8674(00)00015-5
Ihmels J Friedlander G Bergmann S Sarig O Ziv Y Barkai N Revealing modular organization in the yeast transcriptional network Nature Genet 2002 31 370 377 12134151
Draghici S Khatri P Martins RP Charles Ostermeier G Krawetz SA Global functional profiling of gene expression Genomics 2003 81 98 104 12620386 10.1016/S0888-7543(02)00021-6
Lamb JA Ramaswamy S Ford HL Contreras B Martinez RV Kittrell FS Zahnow CA Patterson N Golub TR Ewen ME A mechanism of cyclin D1 action encoded in the patterns of gene expression in human cancer Cell 2003 114 323 334 12914697 10.1016/S0092-8674(03)00570-1
Mootha VK Lindgren CM Eriksson KF Subramanian A Sihag S Lehar J Puigserver P Carlsson E Ridderstrale M Laurila E Houstis N Daly MJ Patterson N Mesirov JP Golub TR Tamayo P Spiegelman B Lander ES Hirschhorn JN Altshuler D Groop LC PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes Nature Genet 2003 34 267 273 12808457 10.1038/ng1180
McCarroll SA Murphy CT Zou S Pletcher SD Chin CS Jan YN Kenyon C Bargmann CI Li H Comparing genomic expression patterns across species identifies shared transcriptional profile in aging Nature Genet 2004 36 197 204 14730301 10.1038/ng1291
Segal E Friedman N Koller D Regev A A module map showing conditional activity of expression modules in cancer Nature Genet 2004 36 1090 1098 15448693
Sweet-Cordero A Mukherjee S Subramanian A You H Roix JJ Ladd-Acosta C Mesirov J Golub TR Jacks T An oncogenic KRAS2 expression signature identified by cross-species gene-expression analysis Nature Genet 2004 37 48 55 15608639
Fraser AG Marcotte EM A probabilistic view of gene function Nature Genet 2004 36 559 564 15167932 10.1038/ng1370
Ashburner M Ball CA Blake JA Botstein D Butler H Cherry JM Davis AP Dolinski K Dwight SS Eppig JT Harris MA Hill DP Issel-Tarver L Kasarskis A Lewis S Matese JC Richardson JE Ringwald M Rubin GM Sherlock G Gene ontology: tool for the unification of biology. The Gene Ontology Consortium Nat Genet 2000 25 25 29 10802651 10.1038/75556
Bittner M Meltzer P Trent J Data analysis and integration: of steps and arrows Nature Genet 1999 22 213 215 10391202 10.1038/10265
Jeffrey C Moonlighting proteins Trends Biochem Sci 1999 24 8 11 10087914 10.1016/S0968-0004(98)01335-8
Hanley JA McNeil BJ The meaning and use of the area under a receiver operating characteristic (ROC) curve Radiology 1982 143 29 36 7063747
Members of the Toxicogenomics Research Consortium Standardizing global gene expression analysis between laboratories and across platforms Nat Methods 2005 2 351 365 15846362 10.1038/nmeth754
Zhang W Morris GQ Chang R Shai O Bakowski MA Mitsakakis N Mohammad N Robinson MD Zirngibl R Somogvi E Laurin N Eftekharpour E Sat E Grigull J Pan Q Peng WT Krogan N Greenblatt J Fehlings M van der Kooy D Aubin J Bruneau GB Rossant J Blencowe BJ Frey BJ Hughes TR The functional landscape of mouse gene expression J Biol 2004 3 21 15588312 10.1186/jbiol16
Clare A King RD How well do we understand the clusters found in microarray data? In Silico Biol 2002 2 511 22 12611631
Hartwell LH Hopfield JJ Leibler S Murray AW From molecular to modular cell biology Nature 1999 402 C47 C52 10591225 10.1038/35011540
|
16162296
|
PMC1239911
|
CC BY
|
2021-01-04 16:27:27
|
no
|
BMC Bioinformatics. 2005 Sep 14; 6:227
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-227
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2291617152710.1186/1471-2105-6-229SoftwareProbeMaker: an extensible framework for design of sets of oligonucleotide probes Stenberg Johan [email protected] Mats [email protected] Ulf [email protected] Department of Genetics and Pathology, Rudbeck Laboratory, Uppsala University, Se-751 85, Uppsala, Sweden2005 19 9 2005 6 229 229 17 6 2005 19 9 2005 Copyright © 2005 Stenberg et al; licensee BioMed Central Ltd.2005Stenberg et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Procedures for genetic analyses based on oligonucleotide probes are powerful tools that can allow highly parallel investigations of genetic material. Such procedures require the design of large sets of probes using application-specific design constraints.
Results
ProbeMaker is a software framework for computer-assisted design and analysis of sets of oligonucleotide probe sequences. The tool assists in the design of probes for sets of target sequences, incorporating sequence motifs for purposes such as amplification, visualization, or identification. An extension system allows the framework to be equipped with application-specific components for evaluation of probe sequences, and provides the possibility to include support for importing sequence data from a variety of file formats.
Conclusion
ProbeMaker is a suitable tool for many different oligonucleotide design and analysis tasks, including the design of probe sets for various types of parallel genetic analyses, experimental validation of design parameters, and in silico testing of probe sequence evaluation algorithms.
==== Body
Background
Increasing numbers of methods are being developed for parallel nucleic acid analyses for different purposes. Many of these methods employ sets of oligonucleotide probes or probe pairs that hybridize to the sequences targeted for analysis, allowing the probe sequences to be acted upon by one or more enzymes, creating new molecular species that reflect the presence or nature of the different target sequences. The reaction products generally contain identifying sequences or other features that allow the separation of signals originating from different targets. This is the case in methods such as the multiplex oligonucleotide ligation assay (OLA) [1], the multiplex ligation-dependent probe amplification assay (MLPA) [2], the RNA- and cDNA-mediated annealing, selection, extension and ligation assays (RASL, DASL) [3,4], the GoldenGate genotyping assay [5], multiplex minisequencing [6], and the padlock or molecular inversion probe assay [7,8]. The latter method has been used to genotype more than 10,000 single nucleotide polymorphisms (SNPs) in multiplex. Another method that utilizes sets of oligonucleotide probes for multiplex processing of nucleic acid molecules is the selector amplification technique. This technique uses partially double-stranded oligonucleotides, called selectors, to circularize a selection of restriction fragments from total genomic DNA, and it incorporates a general sequence motif that allows parallel amplification of all circularized fragments using a single primer pair [9].
With molecular solutions to many tasks of highly parallel genetic analysis now at hand, other factors become limiting, such as the design and the synthesis of reagents. In the work presented here, we address the problem of large-scale probe design. When large numbers of probes are combined, the risk for unintended interactions between probes and targets must be considered. This risk places strict requirements on the design of sets of probes to be used together. In particular, it is important that probes do not contain sequences that result in the production of detectable signal from any probe in the absence of its cognate target molecule, or that otherwise interfere with the activity of other probes in the set. Due to these and other constraints and the many possible alternative probe sequences to evaluate, the difficulty of designing probe sets increases rapidly with the size of the probe sets.
Many computer programs exist for the design of oligonucleotide probes such as PCR primers [10-12], microarray probes [13,14], and more [15]. These programs define algorithms to evaluate the risk of primer or probe sequences being involved in undesired interactions such as probe homo- or heterodimer formation, cross-hybridization, false priming, etc. However, the available programs are generally limited in scope, and are not applicable to the task of designing sets of complex probes containing multiple sequence elements.
The ProbeMaker software presented herein is a framework for computer-assisted design and analysis of sets of oligonucleotide probe sequences composed of several functional sequence elements. As the composition of probes and the constraints imposed on sets of probes vary between applications, this framework has been constructed to support the design of different types of probes using application-specific constraints, as defined by the user. ProbeMaker takes as input a set of target sequences and a number of sets of so-called 'tag' sequences. These tag sequences may serve as targets for restriction digestion, as binding sites for amplification primers or fluorescent detection probes, or as identification codes for individual amplification products that are decoded by hybridization to oligonucleotide arrays [16]. Probes are designed for each target by construction of target-specific sequences and addition of tag sequences according to rules specified by the user. Different combinations of sequence elements are evaluated for each probe, and a set of probe sequences is created that satisfies user-defined criteria.
Implementation
The main objectives in the development of ProbeMaker were to provide a framework that is flexible, in the sense that it should support design of oligonucleotide probes for different purposes, and extensible, in that it should be possible to add support for designing new types of probes and to add new types of design constraints. Furthermore, the software should be adaptable to new applications, and it should have the potential to import sequence data from a variety of sources.
The flexibility is provided by the target and probe sequence data structures used. Each target defines two template sequences that are used to construct target-specific sequences (TSSs) to use in the corresponding probe. Each probe is made up of two such TSSs and a number of tag sequences, which may be located 5' of, between, or 3' of the TSSs. As TSSs may be of zero length, this system allows the design of many different types of probes. Support for more than two TSSs per probe was not deemed necessary as this is not used in any current methods. Furthermore, targets may be grouped, allowing the program to perform selection of tag sequences based on the relations of target sequences, for example variants of the same polymorphic sequence.
The extensibility is realized by using an extension mechanism for much of the functionality. Extensions are constructed in the form of Java classes that implement defined interfaces and may be loaded into the framework at run-time. This mechanism allows the addition of new target types and support for different formats for sequence input and output, as well as design constraints and acceptor schemes, the function of which will be described below.
ProbeMaker may be run through a graphical user interface or from the command line. For the graphical user interface, a set of target sequences and sets of tag sequences are provided as input by the user. Application-specific parameters for probe design and evaluation are set through the user interface. When running ProbeMaker from the command line, a project file defining all sequences and parameters is used as input.
The potential for supporting different file formats is provided by using the sequence input system of the MolTools Java library [17]. A combination of components for sequence file parsing, sequence notation conversion, and post-import modifications are used to allow creation of sets of any type of target from a variety of sequence file formats, with the possibility to carry out other operations on the imported data, such as selecting which position within the target sequence to design probes for, or to group or sort sequences based on some particular property.
Results
For a given set of targets, and a number of sets of tag sequences, ProbeMaker performs two tasks (Figure 1A). Firstly, TSSs are constructed for each target as determined by the target type in use, forming the basis for a probe for that target. Secondly, tag sequences are added to each probe sequentially in a pattern specified by the user. During this procedure, different combinations of tags are evaluated for each probe in order to find one that satisfies specified design constraints.
Figure 1 Schematic description of the probe set design procedure. A) Target-specific sequences are first designed for all targets. Tags are then added to each TSS pair in sequence to form complete probes. B) Prior to tag allocation, each TSS pair is evaluated using selected constraints and the current acceptor scheme. Accepted TSS pairs are used to create a series of probe candidates using each valid combination of tags in turn. This procedure is stopped if an acceptable candidate is found, or when all candidates have been tested. A probe is then selected from the list of accepted or temporarily accepted candidates, using the current selector scheme.
Target-specific sequence construction
The TSSs of each probe are constructed to be complementary to the template sequences defined by the target, with sequence length chosen within a specified length interval to yield a melting temperature (Tm) as close as possible to a specified preferred Tm value. The Tm calculations are performed using a nearest neighbor model [18-20]. The model implementation used here does not take into account the influence of dangling ends or possible stacking interactions between the two probe ends, as these are not always known at this stage of the design. It is possible to use target types that strictly determine the TSS length, which is useful e.g. if other software has been used to find suitable sequences for probe-target hybridization.
Tag selection
After TSS construction, each probe is designed in turn by generation, evaluation, and selection of probe candidates as follows (Figure 1B).
1) The target-specific sequences of the probe are evaluated according to selected design constraints. If found acceptable the process proceeds, otherwise the probe is skipped and reported as a failure.
2) A probe candidate is generated by allocating one tag from each set of tag sequences. This candidate is evaluated according to selected design constraints and ranked on a three-level scale. Based on this rank, the candidate is either accepted, stored in a temporary list, or rejected. This step is iterated, generating a new candidate each time, until all possible tag combinations have been tried or until a candidate has been accepted.
3) One probe is selected from the list of temporarily accepted candidates and any finally accepted candidate.
Probe candidates are constructed by the selection of tags from the provided tag sets based on the selection mode of each tag set. There are five selection modes available.
• A unique tag for every probe
• A common tag for all probes
• A tag common for probes within a group, but unique among groups
• A different tag for each probe in a group, same set of tags used for all groups
• Any tag, regardless of use in other probes
Optionally, a spacer tag may be included to extend any probe that is shorter than a specified length, if probes of identical lengths are desired. Several possible tag combinations may exist for each probe, depending on the selection mode and what tags have been used previously in the probe set. Also, during candidate testing, certain tags may be found unsuitable for use in a particular probe and be excluded from the selection procedure in order to reduce the number of candidates that need to be tested for that probe.
For testing and evaluation of target-specific sequences and probe candidates, the user selects tests that are suitable for the type of probes currently being designed. These tests are incorporated into the framework as extensions. Typically, tests will check for potential unwanted base-pairing interactions within a probe, between a probe and its target, between probes, or between a probe and the targets of other probes. Each test may generate warnings or errors for a candidate; these are then used to rank that candidate. Candidates are by default of the highest rank, warnings reduce this to the intermediate level while errors result in the lowest rank.
Criteria for accepting probe candidates and for choosing among stored candidates are specified by the user by selecting an acceptor scheme and a selector scheme. The acceptor schemes provided with the program include one that will temporarily store all candidates of intermediate or highest rank, and one that will accept candidates of the highest rank, while temporarily storing those of intermediate rank. When probes are designed in groups of two, an exhaustive tag selection mode is available. In this mode, the first probe of a group is not finally determined until an acceptable candidate has been found for the second probe. Both probes are skipped if acceptable candidates cannot be found.
Limitations
Probe design is limited by the amount of available memory, and the amount of time required. Using 500 MB of RAM, it is possible to design probes for at least 20,000 targets. However, when using tests for inter-probe interactions the design time grows exponentially with the number of targets and quickly becomes more limiting than memory. The time required for ProbeMaker to complete a design job is influenced by many factors and is difficult to model and predict. Briefly, the design time depends on the total number of candidates that are generated and the time required for the selected tests to be performed on each generated candidate. The maximum number of candidates generated depends on the size and selection mode of the tag sets used in the design, while the time required for testing of each candidate depends on the tests that are performed and the number of targets/probes being designed.
To illustrate actual time requirements, we set up a design of allele-specific pairs of padlock probes for 1000 random SNP target sequences, allocating to each probe one common primer, either of two allele-specific primers, and one target-specific hybridization tag from a set of 1000 random 20-mers. Without constraints, this design required 8 seconds to complete on a desktop computer system (Intel Pentium 4, 2.5 GHz). When testing for the risk of false ligation using an adaptation of the false-priming algorithm described by Kaderali et al. [15], the same design required 10.5 minutes to complete.
Discussion
A number of recently developed methods for nucleic acid analyses allow large sets of oligonucleotide probes to be used in parallel for simultaneous interrogation of many qualities of a sample. These methods require design of large numbers of oligonucleotide probes. Computer programs commonly used to design various types of oligonucleotides [10-15] define a repertoire of criteria, and algorithms to evaluate oligonucleotides based on these criteria. However, the available programs are mainly dedicated for the design of amplification or sequencing primers or microarray probes, and most programs can not readily be modified for new uses.
In this work, we present a framework for computer-assisted design and analysis of sets of oligonucleotide probes. The ProbeMaker software allows the design of sets of any type of probes with up to two elements that are complementary to the target sequence and that include a number of other sequence elements. Furthermore, ProbeMaker is equipped with an extension mechanism that allows the incorporation of new design criteria as well as criteria described in previous works. Similarly, it is possible to define new types of targets, which will allow the design of new types of probes, including probes for non-nucleic acid targets, such as pairs of oligonucleotides to be attached to antibodies or other affinity reagents for protein analyses by proximity ligation [21,22].
Conclusion
ProbeMaker enables constraint-based design of large sets of probes. Besides facilitating the deployment of large-scale assays, this can be used to systematically vary design criteria in order to experimentally optimize design parameters. Furthermore, the flexibility and extensibility of this framework makes it suitable for in silico comparison and evaluation of different oligonucleotide analysis algorithms, and it could act as a common platform for further development within the field.
Availability and requirements
Project name: ProbeMaker
Project home page:
Operating system(s): Platform independent
Programming language: Java
Other requirements: Java 1.4 or higher, MolTools and AppTools libraries (provided with ProbeMaker and available under the GNU LGPL License from )
License: GNU GPL
Any restrictions to use by non-academics: No
Authors' contributions
JS designed and implemented the software and drafted the manuscript. MN and UL conceived of and supervised the work. All authors read and approved the final manuscript.
Acknowledgements
This work was supported by grants from the Wallenberg Consortium North for functional genomics, the EU framework program 6 integrated project MolTools, and the Swedish Research Council's Scientific Councils for natural and engineering sciences, and for medicine.
==== Refs
Grossman PD Bloch W Brinson E Chang CC Eggerding FA Fung S Iovannisci DM Woo S Winn-Deen ES High-density multiplex detection of nucleic acid sequences: oligonucleotide ligation assay and sequence-coded separation Nucleic Acids Res 1994 22 4527 4534 7526344
Schouten JP McElgunn CJ Waaijer R Zwijnenburg D Diepvens F Pals G Relative quantification of 40 nucleic acid sequences by multiplex ligation-dependent probe amplification Nucleic Acids Res 2002 30 e57 12060695 10.1093/nar/gnf056
Yeakley JM Fan JB Doucet D Luo L Wickham E Ye Z Chee MS Fu XD Profiling alternative splicing on fiber-optic arrays Nat Biotechnol 2002 20 353 358 11923840 10.1038/nbt0402-353
Fan JB Yeakley JM Bibikova M Chudin E Wickham E Chen J Doucet D Rigault P Zhang B Shen R McBride C Li HR Fu XD Oliphant A Barker DL Chee MS A versatile assay for high-throughput gene expression profiling on universal array matrices Genome Res 2004 14 878 885 15123585 10.1101/gr.2167504
Oliphant A Barker DL Stuelpnagel JR Chee MS BeadArray technology: enabling an accurate, cost-effective approach to high-throughput genotyping Biotechniques 2002 Suppl 56 8, 60-1 12083399
Fan JB Chen X Halushka MK Berno A Huang X Ryder T Lipshutz RJ Lockhart DJ Chakravarti A Parallel genotyping of human SNPs using generic high-density oligonucleotide tag arrays Genome Res 2000 10 853 860 10854416 10.1101/gr.10.6.853
Banér J Isaksson A Waldenstrom E Jarvius J Landegren U Nilsson M Parallel gene analysis with allele-specific padlock probes and tag microarrays Nucleic Acids Res 2003 31 e103 12930977 10.1093/nar/gng104
Hardenbol P Yu F Belmont J Mackenzie J Bruckner C Brundage T Boudreau A Chow S Eberle J Erbilgin A Falkowski M Fitzgerald R Ghose S Iartchouk O Jain M Karlin-Neumann G Lu X Miao X Moore B Moorhead M Namsaraev E Pasternak S Prakash E Tran K Wang Z Jones HB Davis RW Willis TD Gibbs RA Highly multiplexed molecular inversion probe genotyping: over 10,000 targeted SNPs genotyped in a single tube assay Genome Res 2005 15 269 275 15687290 10.1101/gr.3185605
Dahl F Gullberg M Stenberg J Landegren U Nilsson M Multiplex amplification enabled by selective circularization of large sets of genomic DNA fragments Nucl Acids Res 2005 33 e71 15860768 10.1093/nar/gni070
Rozen S Skaletsky H Primer3 on the WWW for general users and for biologist programmers Methods Mol Biol 2000 132 365 386 10547847
Chen SH Lin CY Cho CS Lo CZ Hsiung CA Primer Design Assistant (PDA): A web-based primer design tool Nucleic Acids Res 2003 31 3751 3754 12824410 10.1093/nar/gkg560
Vallone PM Butler JM AutoDimer: a screening tool for primer-dimer and hairpin structures Biotechniques 2004 37 226 231 15335214
Kaderali L Schliep A Selecting signature oligonucleotides to identify organisms using DNA arrays Bioinformatics 2002 18 1340 1349 12376378 10.1093/bioinformatics/18.10.1340
Rouillard JM Zuker M Gulari E OligoArray 2.0: design of oligonucleotide probes for DNA microarrays using a thermodynamic approach Nucleic Acids Res 2003 31 3057 3062 12799432 10.1093/nar/gkg426
Kaderali L Deshpande A Nolan JP White PS Primer-design for multiplexed genotyping Nucleic Acids Res 2003 31 1796 1802 12626722 10.1093/nar/gkg267
Shoemaker DD Lashkari DA Morris D Mittmann M Davis RW Quantitative phenotypic analysis of yeast deletion mutants using a highly parallel molecular bar-coding strategy Nat Genet 1996 14 450 456 8944025 10.1038/ng1296-450
Stenberg J The MolTools Java library
Sugimoto N Nakano S Yoneyama M Honda K Improved thermodynamic parameters and helix initiation factor to predict stability of DNA duplexes Nucleic Acids Res 1996 24 4501 4505 8948641 10.1093/nar/24.22.4501
Owczarzy R Vallone PM Gallo FJ Paner TM Lane MJ Benight AS Predicting sequence-dependent melting stability of short duplex DNA oligomers Biopolymers 1997 44 217 239 9591477 10.1002/(SICI)1097-0282(1997)44:3<217::AID-BIP3>3.0.CO;2-Y
SantaLucia JJ Hicks D The thermodynamics of DNA structural motifs Annu Rev Biophys Biomol Struct 2004 33 415 440 15139820 10.1146/annurev.biophys.32.110601.141800
Fredriksson S Gullberg M Jarvius J Olsson C Pietras K Gustafsdottir SM Ostman A Landegren U Protein detection using proximity-dependent DNA ligation assays Nat Biotechnol 2002 20 473 477 11981560 10.1038/nbt0502-473
Gullberg M Gustafsdottir SM Schallmeiner E Jarvius J Bjarnegard M Betsholtz C Landegren U Fredriksson S Cytokine detection by antibody-based proximity ligation Proc Natl Acad Sci U S A 2004 101 8420 8424 15155907 10.1073/pnas.0400552101
|
16171527
|
PMC1239912
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Sep 19; 6:229
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-229
|
oa_comm
|
==== Front
BMC Cardiovasc DisordBMC Cardiovascular Disorders1471-2261BioMed Central London 1471-2261-5-271616474310.1186/1471-2261-5-27Research ArticleA common variant of endothelial nitric oxide synthase (Glu298Asp) is associated with collateral development in patients with chronic coronary occlusions Lamblin Nicolas [email protected] François J [email protected] Nicole [email protected] Jean [email protected] Jean-Marc [email protected] Philippe [email protected] Christophe [email protected] Belle Eric [email protected] Hôpital Cardiologique, Centre Hospitalier Universitaire de Lille, Place de Verdun, 59037 Lille cedex, France2 INSERM U508, Institut Pasteur de Lille, 1 rue Calmette, 59019 Lille cedex, France2005 15 9 2005 5 27 27 7 3 2005 15 9 2005 Copyright © 2005 Lamblin et al; licensee BioMed Central Ltd.2005Lamblin et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Experimental studies support an important role for endothelial nitric oxide synthase (eNOS) in the regulation of angiogenesis. In humans, a common polymorphism exists in the eNOS gene that results in the conversion of glutamate to aspartate for codon 298. In vitro and in vivo studies have suggested a decreased NOS activity in patients with the Asp298 variant. We hypothesized that a genetic-mediated decreased eNOS activity may limit collateral development in patients with chronic coronary occlusions.
Methods
We selected 291 consecutive patients who underwent coronary angiography and who had at least one chronic (>15 days) total coronary occlusion. Collateral development was graded angiographically using two different methods: the collateral flow grade and the recipient filling grade. Genomic DNA was extracted from white blood cells and genotyping was performed using previously published techniques.
Results
Collateral development was lower in patients carrying the Asp298 variant than in Glu-Glu homozygotes (collateral flow grade: 2.64 ± 0.08 and 2.89 ± 0.08, respectively, p = 0.04; recipient filling grade: 3.00 ± 0.08 and 3.24 ± 0.07, respectively, p = 0.04). By multivariable analysis, three variables were independently associated with the collateral flow grade: female gender, smoking, and the Asp298 variant (p = 0.03) while the Asp298 variant was the sole variable independently associated with the recipient filling grade (p = 0.03).
Conclusion
Collateral development is lower in patients with the Asp298 variant. This may be explained by the decreased NOS activity in patients with the Asp298 variant. Further studies will have to determine whether increasing eNOS activity in humans is associated with coronary collateral development.
==== Body
Background
In spite of recent advances in the techniques used for myocardial revascularization, chronic total coronary occlusions are frequently observed in patients with coronary artery disease. This could lead to symptoms of angina, quality of life impairment, left ventricular dysfunction, and prognosis worsening. In the case of severe stenosis or total occlusion of a coronary artery, the collateral circulation may be an alternative source of blood supply to the myocardium at risk [1,2]. Although some factors, such as the duration of myocardial ischemic symptoms, have been associated with the extent of collateralization, coronary collateral development remains difficult to anticipate and there is considerable inter-individual variability in this process [3]. One emerging concept in cardiovascular diseases, which could explain this variability, is the possible interaction between genetic determinants and the pathophysiological responses to cardiac injury.
Among candidate genes that may be implicated in collateral development is the endothelial nitric oxide synthase (eNOS) gene. Experimental studies support an important role for eNOS in the regulation of angiogenesis [4]: mice lacking eNOS gene have severely reduced angiogenesis in response to tissue ischemia [5,6] while eNOS overexpression enhances angiogenesis [7-9]. In humans, different common polymorphisms exist in the eNOS gene and among them one that results in the conversion of glutamate to aspartate for codon 298. In vitro studies have suggested that the Asp298 variant may be functional and associated with a decreased of eNOS activity [10]. In vivo studies have documented an increased reactivity to alpha-adrenergic stimulation in patients with the Asp298 variant suggesting a decreased NOS activity [11].
In the present study, we hypothesized that a genetic-mediated decreased eNOS activity may limit collateral development in patients with chronic coronary occlusions. We studied 291 patients with chronic coronary occlusions in whom collateral development was graded angiographically. We show that patients with the Asp298 variant have significantly less collateral vessel formation than Glu-Glu homozygotes.
Methods
Study population
Between May 2000 and October 2001, 2050 consecutive patients who underwent a coronary angiography at our institution were enrolled in a registry. All patients gave informed consent and had blood and serum samples that were stored at -80°C until further analysis. The baseline clinical and angiographic characteristics were prospectively recorded by trained physicians.
For the purpose of this study, we selected all patients who had at least one chronic (>15 days) total occlusion of a major coronary vessel. The patients with a history of coronary artery bypass graft were excluded. Two hundred and ninety one patients were thus selected to form the study population.
Angiography procedure and coronary collaterals grading
Selective coronary angiography was performed in multiple orthogonal projections using the Judkins technique. In case of significant lesion (stenosis or total occlusion), there was an intracoronary nitrates infusion. Collateral development was graded using two different methods by two independent observers. These methods have been previously validated [12].
The collateral flow grade evaluates the flow in the collateral: 0 = no flow in the collateral; 1 = the collateral is barely apparent; dye is not visible throughout the cardiac cycle but is present in at least 3 consecutive frames; 2 = the collateral is moderately opaque but is present throughout at 75% of the cardiac cycle; 3 = the collateral is well opacified and the column of dye is well defined but is < 0.7 mm wide throughout the majority of its length; 4 = the collateral is well opacified, fills antegrade, and is very large.
The second method was the recipient filling grade: 0 = no angiographically apparent collaterals; 1 = apparent collaterals extend into a region of myocardium with no angiographically apparent recipient vessel; 2 = minimal recipient filling by collaterals is manifested by minor side branch filling and no epicardial artery or epicardial side branch filling; 3 = Moderate recipient filling by collaterals is manifested by complete filling of epicardial side branches and partial filling of a major epicardial artery; 4 = there is complete filling of a major epicardial segment.
Genetic analysis
Blood samples were collected at the time of coronary angiography. Genomic DNA was extracted from white blood cells by a « salting out » procedure as previously described [13]. DNA fragment – including the G/T translation in the exon 7 – amplification was performed by Polymerase Chain Reaction (PCR). Primers and PCR conditions used for eNOS have been reported previously [14]. The products was digested by the Ban II enzyme for genotyping as previously described [14] and the results of the genotyping were tested for the Hardy-Weinberg equilibrium (p > 0.05).
Statistical analysis
Patients were grouped on the basis of the presence or absence of the Asp298 variant (Asp298 homozygotes and heterozygotes were combined and compared to the homozygotes Glu298) as previously reported [11,14-16]. For continuous variable, distributions were first tested before analyses. Since the distribution were normal, they were presented as mean ± SEM and were compared with use of the bilateral unpaired Student' t test. Qualitative variables were compared with use of the Pearson chi-square test or the Fisher exact test when necessary. Multivariable analysis was performed with use of a general linear model (GLM) adjusted for age, gender, smoking, hypertension, hypercholesterolemia and diabetes mellitus. Statistical analysis were performed with the SAS software, version 8 (SAS Institute Inc., Cary, NC, USA).
Results
The baseline characteristics of the study population are shown in table 1. There were no statistically significant differences between patients with the Asp298 variant and Glu-Glu homozygotes. Most patients were male with a mean age of 63 ± 11 years and with a high prevalence of cardiovascular risk factors. Notably, 77% of the patients were current or past smokers and 36% were diabetics. Cardiovascular medications did not differ between the two groups. The angiographic severity of coronary atherosclerosis was similar in the two groups. In most of the cases, only one coronary artery was totally occluded.
Table 1 Demographics and medical therapy at baseline by genotype
Asp298 (n = 168) Glu-Glu (n = 123) All patients (n = 291) p
Age, years 63 ± 11 62 ± 11 63 ± 11 0.92
Female gender, % 18 20 19 0.72
Body mass index, kg/m2 28.3 ± 4.7 27.8 ± 4.6 28.1 ± 4.6 0.32
Risk factors, %
Smoking 76 78 77 0.71
Hypercholesterolemia 76 81 78 0.29
Hypertension 54 55 55 0.86
Diabetes mellitus 35 37 36 0.80
Familial history of CAD 37 35 36 0.18
Clinical symptoms, %
Stable 73 78 75 0.35
Unstable 27 22 25
Angiographic data:
No. of vessels with > 50% stenosis, %
1 vessel 24 25 25
2 vessels 38 34 36 0.92
3 vessels 38 41 39
No. of vessels with total occlusion, %
1 vessel 85 83 84 0.61
2 vessels 15 17 16
Occluded LAD, % 38 29 34 0.12
Occluded Cx, % 17 22 19 0.32
Occluded RCA, % 60 66 62 0.27
Cardiovascular medications, %
ASA 78 78 78 0.96
ACE inhibitors 55 52 54 0.57
ARB 6 10 8 0.24
Beta-blockers 63 58 61 0.34
Nitrates 58 48 54 0.09
Calcium antagonists 24 31 27 0.20
Statins 63 58 61 0.34
Data are presented as percent of patients or mean value ± SD
CAD = coronary artery disease; MI, myocardial infarction
LAD = left anterior descending artery; Cx = circumflex; RCA = right coronary artery
ASA = acetylsalicylic acid; ACE = angiotensin-converting enzyme; ARB = angiotensin 2 receptor blockers.
In the overall study population, the mean collateral flow grade was 2.75 ± 0.06 and the mean recipient filling grade was 3.10 ± 0.06. By univariable analysis, angiographic evidence of collateral development was lower in patients carrying the Asp298 variant than in Glu-Glu homozygotes (collateral flow grade: 2.64 ± 0.08 and 2.89 ± 0.08, respectively, p = 0.04; recipient filling grade: 3.00 ± 0.08 and 3.24 ± 0.07, respectively, p = 0.04). When patients were classified into 3 groups (Glu-Glu homozygotes, Glu-Asp heterozygotes, Asp-Asp homozygotes), respective values for collateral flow grade were 2.84 ± 0.08, 2.63 ± 0.09, and 2.69 ± 0.23, while respective values for recipient filling grade were 3.24 ± 0.07, 3.01 ± 0.08, and 2.94 ± 0.23. Independent predictors of collateral development were then determined by multivariable analysis (table 2). When considering the collateral flow grade, three variables were independently associated with an impaired collateral development: female gender, smoking, and the Asp298 variant; there was also a strong trend for a deleterious effect of diabetes mellitus on collateral development. When considering the recipient filling grade, the Asp298 variant was the sole variable independently associated with an impaired collateral development; again, the presence of diabetes mellitus was associated with a trend for a deleterious effect. The presence of the Asp298 variant was therefore the sole independent predictor of both the collateral flow grade and the recipient filling grade.
Table 2 Predictors of collateral development: multivariable analysis
Collateral flow grade Recipient filling grade
β p β p
Age ≥ 63 - 0.04 0.72 + 0.03 0.77
Female gender - 0.49 0.009 - 0.22 0.23
Smoking - 0.34 0.05 - 0.12 0.47
Hypertension - 0.04 0.72 - 0.04 0.74
Hypercholesterolemia - 0.01 1.00 - 0.16 0.23
Diabetes mellitus - 0.23 0.06 - 0.20 0.09
Asp298 variant - 0.26 0.03 - 0.24 0.03
Since collateral vessel development has been associated with a preserved left ventricular function in the case of total occlusion of a coronary artery [17], we compared the left ventricular ejection fraction in both groups. A recent evaluation of left ventricular ejection fraction was available in 162 (96%) of patients with the Asp298 variant and in 120 (98%) of Glu-Glu homozygotes. The mean ( ± SD) left ventricular ejection fraction was 53 ± 16% in the overall study population and was 50 ± 16% in patients with the Asp298 variant versus 54 ± 16% in Glu-Glu homozygotes (p = 0.05).
Discussion
In the present study, we found that patients carrying the Asp298 variant of eNOS gene had significantly less angiographic evidence of collateral vessel formation in response to total coronary occlusion. Multivariable analysis showed that this effect was independent of other factors that influence collateral vessel formation.
Nitric oxide (NO), constitutively produced by endothelial nitric oxide synthase (eNOS), plays critical roles in vascular biology, including regulation of vascular tone and blood pressure. In addition to its vasodilatory properties, NO has been implicated in the modulation of angiogenesis [4]. Ziche et al. suggested that NO may play a role in angiogenesis elicited by VEGF but not by FGF [18]. Murohara et al. demonstrated that angiogenesis developing in response to limb ischemia was severely reduced in mice lacking eNOS gene [5]. Moreover, eNOS over expression in transgenic mice [8] or using gene transfer strategies [7,9] enhances angiogenesis in response to tissue ischemia. There is thus strong experimental evidence to support an important role for eNOS in the regulation of angiogenesis in animal models, however the implication of eNOS activity in collateral vessel formation in response to myocardial ischemia in humans remains unknown
ENOS is encoded by a 26-exon gene located on chromosome 7 [19]. In view of the physiological and pathophysiological importance of NO, the potential role of eNOS in the pathogenesis of various human diseases has been examined using its polymorphic variants as potential disease markers. Different common polymorphisms exist and among them one in nucleotide 894 (G-T) that results in the conversion of glutamate to aspartate for codon 298. A study by Philip et al. has documented in vivo an increased reactivity to alpha-adrenergic stimulation in patients with the Asp298 variant suggesting a decreased NOS activity [11]. In recent clinical studies, the Asp298 variant has been implicated as a risk factor for coronary artery disease [20], hypertension [16], or has been associated with a poorer event-free survival in patients with congestive heart failure [14]. In a meta-analysis of 26 studies involving 23028 subjects, homozygosity for Asp298 was associated with increased risk of ischemic heart disease by 31% [15]. When taken together with the results of the present study, the above described literature suggests that a decreased NOS activity in coronary vessels of patients with the Asp298 variant may explain the decreased collateral vessel formation observed in this subgroup of patients. The mechanism by which the Asp298 variant could decrease the eNOS activity remains unclear. The Asp298 variant has been associated with an increased susceptibility to enzymatic cleavage [10] but it has been suggested that the increased susceptibility to proteolytic cleavage of NOS could result from sample preparation [21]. In addition, other recent studies failed to find any association between the Asp298 variant and eNOS activity [22,23]. An other possible explanation is that the variant Asp298 may simply be a genetic marker. For example, the Asp298 variant is in linkage disequilibrium with an other polymorphism resulting in a nucleotide substitution in the promoter region (T/C -786) of the eNOS gene [24,25]. It has been shown that the rarer variant (C) suppresses eNOS transcription by approximately 50%. The Asp298 could be a marker of the occurrence of the unfavorable "C" allele in the promoter region, the latter responsible for the decreased of eNOS activity.
In the present study, collateral vessel formation was assessed using angiographic criteria. Angiographically visible collaterals represent only a fraction of the total collateral vessels because collaterals are angiographically demonstrable only when they reach 200 μm [12]. Several studies have shown that assessing the collateral circulation by intracoronary Doppler flow or pressure wires may be an interesting alternative to determine collateral blood flow in humans [26,27]; however, the invasive nature of this method which imply to cross the occlusion site by a guide wire would limit its interest in a genetic association study like the present one in which inclusion of consecutive and as much as possible unselected cases is mandatory to provide unbiased results. In an attempt to provide a rigorous, systematic analysis of human coronary angiogenesis by angiography, we used two criteria recently reviewed by Gibson et al. [12]. The collateral flow grade focuses on the development of the collateral network itself while the recipient filling grade is adapted from the Rentrop grade [28] and provides information on how the recipient vessel is filled by the collaterals. In the present study, the fact that the deleterious impact of the Asp298 variant was evident with both criteria reinforces our findings. An important question in the present study is to know whether the better angiographic criteria in Glu-Glu homozygotes are a result of the vasodilation properties of eNOS or are due to an increase in blood vessel growth. Such a question is beyond the scope of this clinical study but a direct effect of eNOS on angiogenesis has been documented in the above described experimental studies [7-9]. Moreover, the trend for higher left ventricular ejection fraction observed in Glu-Glu homozygotes suggests a beneficial effect of the collaterals on myocardial function. Finally, our analysis was based on a single time point; further assessment of collateral development by repeated angiographic follow-up would be of interest but was not performed due to the invasive nature of coronary angiography.
Conclusion
In conclusion, this investigation is the first study to show the relationship between the 894 (G-T) eNOS polymorphism and coronary collaterals in humans. It demonstrates that collateral development is poorer in patients with the Asp298 variant. This may be explained by a decreased NOS activity in patients with the Asp298 variant. Further studies will have to determine whether increasing eNOS activity in humans is associated with coronary collateral development.
Abbreviations
eNOS: endothelial nitric oxyde synthase
FGF: fibroblast growth factor
PCR: polymerase chain reaction
VEGF: vascular endothelial growth factor
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
NL, FJC, and EVB participated in the clinical organisation including patients inclusion and angiographic analyses of the collateral development. NL, and NH carried out the genetic analyses and the validation of the analyses. NL performed the interpretation and the statistical analyses of the data. JML, JD, and EVB participated in the conception and the design of the study. PA participated in the interpretation and the analyses of the data. NL, CB, and EVB participated in the conception and design of the study and drafted the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Helfant RH Vokonas PS Gorlin R Functional importance of the human coronary collateral circulation N Engl J Med 1971 284 1277 81 5576437
Fuster V Frye RL Kennedy MA Connolly DC Mankin HT The role of collateral circulation in the various coronary syndromes Circulation 1979 59 1137 44 436205
Koerselman J van der Graaf Y de Jaegere PP Grobbee DE Coronary collaterals: an important and underexposed aspect of coronary artery disease Circulation 2003 107 2507 11 12756191 10.1161/01.CIR.0000065118.99409.5F
Cooke JP NO and angiogenesis Atheroscler Suppl 2003 4 53 60 14664903 10.1016/S1567-5688(03)00034-5
Murohara T Asahara T Silver M Bauters C Masuda H Kalka C Kearney M Chen D Symes JF Fishman MC Huang PL Isner JM Nitric oxide synthase modulates angiogenesis in response to tissue ischemia J Clin Invest 1998 101 2567 78 9616228
Zhao X Lu X Feng Q Deficiency in endothelial nitric oxide synthase impairs myocardial angiogenesis Am J Physiol Heart Circ Physiol 2002 283 H2371 8 12388304
Smith RS JrLin KF Agata J Chao L Chao J Human endothelial nitric oxide synthase gene delivery promotes angiogenesis in a rat model of hindlimb ischemia Arterioscler Thromb Vasc Biol 2002 22 1279 85 12171788 10.1161/01.ATV.0000026613.18742.67
Amano K Matsubara H Iba O Okigaki M Fujiyama S Imada T Kojima H Nozawa Y Kawashima S Yokoyama M Iwasaka T Enhancement of ischemia-induced angiogenesis by eNOS overexpression Hypertension 2003 41 156 62 12511546 10.1161/01.HYP.0000053552.86367.12
Namba T Koike H Murakami K Aoki M Makino H Hashiya N Ogihara T Kaneda Y Kohno M Morishita R Angiogenesis induced by endothelial nitric oxide synthase gene through vascular endothelial growth factor expression in a rat hindlimb ischemia model Circulation 2003 108 2250 7 14568906 10.1161/01.CIR.0000093190.53478.78
Tesauro M Thompson WC Rogliani P Qi L Chaudhary PP Moss J Intracellular processing of endothelial nitric oxide synthase isoforms associated with differences in severity of cardiopulmonary diseases: cleavage of proteins with aspartate vs. glutamate at position 298 Proc Natl Acad Sci USA 2000 97 2832 5 10717002 10.1073/pnas.97.6.2832
Philip I Plantefeve G Vuillaumier-Barrot S Vicaut E LeMarie C Henrion D Poirier O Levy BI Desmonts JM Durand G Benessiano J G894T polymorphism in the endothelial nitric oxide synthase gene is associated with an enhanced vascular responsiveness to phenylephrine Circulation 1999 99 3096 8 10377070
Gibson CM Ryan K Sparano A Moynihan JL Rizzo M Kelley M Marble SJ Laham R Simons M McClusky TR Dodge JT Jr Angiographic methods to assess human coronary angiogenesis Am Heart J 1999 137 169 79 9878950
Miller SA Dykes DD Polesky HF A simple salting out procedure for extracting DNA from human nucleated cells Nucleic Acids Res 1988 16 1215 3344216
McNamara DM Holubkov R Postava L Ramani R Janosko K Mathier M MacGowan GA Murali S Feldman AM London B Effect of the Asp298 variant of endothelial nitric oxide synthase on survival for patients with congestive heart failure Circulation 2003 107 1598 602 12668492 10.1161/01.CIR.0000060540.93836.AA
Casas JP Bautista LE Humphries SE Hingorani AD Endothelial nitric oxide synthase genotype and ischemic heart disease: meta-analysis of 26 studies involving 23028 subjects Circulation 2004 109 1359 65 15007011 10.1161/01.CIR.0000121357.76910.A3
Miyamoto Y Saito Y Kajiyama N Yoshimura M Shimasaki Y Nakayama M Kamitani S Harada M Ishikawa M Kuwahara K Ogawa E Hamanaka I Takahashi N Kaneshige T Teraoka H Akamizu T Azuma N Yoshimasa Y Yoshimasa T Itoh H Masuda I Yasue H Nakao K Endothelial nitric oxide synthase gene is positively associated with essential hypertension Hypertension 1998 32 3 8 9674630
Khaja F Sabbah HG Brymer JF Stein PD Influence of coronary collaterals on left ventricular function in patients undergoing coronary angioplasty Am Heart J 1988 116 1174 80 2973212 10.1016/0002-8703(88)90436-X
Ziche M Morbidelli L Choudhuri R Zhang HT Donnini S Granger HJ Bicknell R Nitric oxide synthase lies downstream from vascular endothelial growth factor-induced but not basic fibroblast growth factor-induced angiogenesis J Clin Invest 1997 99 2625 34 9169492
Marsden PA Heng HH Scherer SW Stewart RJ Hall AV Shi XM Tsui LC Schappert KT Structure and chromosomal localization of the human constitutive endothelial nitric oxide synthase gene J Biol Chem 1993 268 17478 88 7688726
Hingorani AD Liang CF Fatibene J Lyon A Monteith S Parsons A Haydock S Hopper RV Stephens NG O'Shaughnessy KM Brown MJ A common variant of the endothelial nitric oxide synthase (Glu298--> Asp) is a major risk factor for coronary artery disease in the UK Circulation 1999 100 1515 20 10510054
Fairchild TA Fulton D Fontana JT Gratton JP McCabe TJ Sessa WC Acidic hydrolysis as a mechanism for the cleavage of the Glu(298)--> Asp variant of human endothelial nitric-oxide synthase J Biol Chem 2001 276 26674 9 11331296 10.1074/jbc.M103647200
Golser R Gorren AC Mayer B Schmidt K Functional characterization of Glu298Asp mutant human endothelial nitric oxide synthase purified from a yeast expression system Nitric Oxide 2003 8 7 14 12586536 10.1016/S1089-8603(02)00131-3
McDonald DM Alp NJ Channon KM Functional comparison of the endothelial nitric oxide synthase Glu298Asp polymorphic variants in human endothelial cells Pharmacogenetics 2004 14 831 9 15608562 10.1097/00008571-200412000-00006
Tanus-Santos JE Desai M Flockhart DA Effects of ethnicity on the distribution of clinically relevant endothelial nitric oxide variants Pharmacogenetics 2001 11 719 25 11692081 10.1097/00008571-200111000-00011
Marroni AS Metzger IF Souza-Costa DC Nagassaki S Sandrim VC Correa RX Rios-Santos F Tanus-santos JE Consistent interethnic differences in the distribution of clinically relevant endothelial nitric oxide synthase genetic polymorphisms Nitric Oxide 2005 12 177 82 15797845 10.1016/j.niox.2005.02.002
Pohl T Seiler C Billinger M Herren E Wustmann K Mehta H Windecker S Eberli FR Meier B Frequency distribution of collateral flow and factors influencing collateral channel development. Functional collateral channel measurement in 450 patients with coronary artery disease J Am Coll Cardiol 2001 38 1872 8 11738287 10.1016/S0735-1097(01)01675-8
Werner GS Bahrmann P Mutschke O Emig U Betge S Ferrari M Figulla HR Determinants of target vessel failure in chronic total coronary occlusions after stent implantation. The influence of collateral function and coronary hemodynamics J Am Coll Cardiol 2003 42 219 25 12875755 10.1016/S0735-1097(03)00624-7
Rentrop KP Cohen M Blanke H Phillips RA Changes in collateral channel filling immediately after controlled coronary artery occlusion by an angioplasty balloon in human subjects J Am Coll Cardiol 1985 5 587 92 3156171
|
16164743
|
PMC1239913
|
CC BY
|
2021-01-04 16:30:06
|
no
|
BMC Cardiovasc Disord. 2005 Sep 15; 5:27
|
utf-8
|
BMC Cardiovasc Disord
| 2,005 |
10.1186/1471-2261-5-27
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-1091610721010.1186/1471-2164-6-109Research ArticleDual activation of pathways regulated by steroid receptors and peptide growth factors in primary prostate cancer revealed by Factor Analysis of microarray data Lozano Juan Jose [email protected] Marta [email protected] Raquel [email protected] David [email protected] Pedro L [email protected] Timothy M [email protected] Angel R [email protected] Bioinformatics Unit, Centro de Biología Molecular "Severo Ochoa" (CSIC-UAM), Universidad Autónoma de Madrid, Cantoblanco, 28049 Madrid, Spain2 Department of Physiology and Biophysics, Mount Sinai School of Medicine, One Gustave Levy Pl., New York, NY 10029, USA3 Instituto de Biología Molecular, Consejo Superior de Investigaciones Científicas, c. Jordi Girona 18–26, 08034 Barcelona, Spain4 Departament de Anatomía Patològica, Hospital Clínic, and Institut de Investigacions Biomèdiques August Pi i Sunyer, c. Villarroel 170, 08036 Barcelona, Spain5 Center for Genome Regulation, Barcelona (Spain)2005 17 8 2005 6 109 109 17 1 2005 17 8 2005 Copyright © 2005 Lozano et al; licensee BioMed Central Ltd.2005Lozano et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We use an approach based on Factor Analysis to analyze datasets generated for transcriptional profiling. The method groups samples into biologically relevant categories, and enables the identification of genes and pathways most significantly associated to each phenotypic group, while allowing for the participation of a given gene in more than one cluster. Genes assigned to each cluster are used for the detection of pathways predominantly activated in that cluster by finding statistically significant associated GO terms. We tested the approach with a published dataset of microarray experiments in yeast. Upon validation with the yeast dataset, we applied the technique to a prostate cancer dataset.
Results
Two major pathways are shown to be activated in organ-confined, non-metastatic prostate cancer: those regulated by the androgen receptor and by receptor tyrosine kinases. A number of gene markers (HER3, IQGAP2 and POR1) highlighted by the software and related to the later pathway have been validated experimentally a posteriori on independent samples.
Conclusion
Using a new microarray analysis tool followed by a posteriori experimental validation of the results, we have confirmed several putative markers of malignancy associated with peptide growth factor signalling in prostate cancer and revealed others, most notably ERRB3 (HER3). Our study suggest that, in primary prostate cancer, HER3, together or not with HER4, rather than in receptor complexes involving HER2, could play an important role in the biology of these tumors. These results provide new evidence for the role of receptor tyrosine kinases in the establishment and progression of prostate cancer.
==== Body
Background
The phenotype of a cell is determined by its transcriptional repertoire, a result of combinations of transcriptional programs partly set during lineage determination and partly activated in response to intrinsic and extrinsic stimuli. Microarray hybridization experiments permit a quantitative analysis of this transcriptional repertoire in response to defined experimental conditions. A particularly interesting case of study is given by the transcriptional repertoire of human tumors. Here, the objective is usually the search for cancer subtypes for individualized prognosis and/or therapy. The questions most frequently asked are whether samples can be automatically grouped, in the absence of additional information, into biologically relevant phenotypes; and whether transcriptional programs can be unveiled that can explain such phenotypes. It must be noted that this situation (sample clustering and relevant gene extraction) is difficult mainly due to three reasons [1]: the sparsity of the data (samples), the high dimensionality of the feature (gene) space, and the fact that many features are irrelevant or redundant (low signal-to-noise ratio). It has been pointed out that, due to the low signal-to-noise ratio, the quality and reliability of clustering may degrade when using standard hierarchical clustering algorithms or similar approximations [2]. Similarly, model-based clustering methods encounter problems due to the sparsity of the set and its high dimensionality, leading to overfitting during the density estimation process [3]. Additional difficulties are encountered during the selection of features (genes) relevant to the sample cluster structure, since most clustering methods produce non-overlapping gene clusters. This behaviour may distort the extraction of biologically relevant genes in cases where expression patterns overlap several classes of samples or experimental conditions, a reflection of the dependence of the expression of most genes on multiple signals and their participation in more than one regulatory network.
Three main strategies have been taken in sample-based clustering: unsupervised gene selection, interrelated clustering and biclustering [1]. The first views gene selection and sample clustering as basically independent processes, the second dynamically uses the relationship between gene and sample spaces to iteratively apply a clustering and selection engine, while the third tries to cluster both genes and samples at the same time in a reduced space. For the first one, principal components analysis (PCA)[4] has been proposed. PCA, a well known dimensionality reduction technique, has been criticized because the sample projection in the low-dimensional space is not guaranteed to yield optimal sample partitions, particularly when the fraction of relevant genes specific to each cluster is small. As for the second approach, several novel methods have been proposed recently based on various greedy filtering techniques (for a review see [1]), but it has been suggested that they may group the data based on local decisions [1]. Finally, different biclustering methods have also been applied to this situation [5-8], but a difficulty with most biclustering tools is that they generate non-overlapping partitions.
Here we apply Factor Analysis (FA) [9], a multivariate tool related to PCA, coupled to clustering algorithms in sample space, t-test scores in gene space and data mining procedures. Q-mode (i.e. in sample space) FA is a latent variable modelling tool [9] that assumes that the observed gene expression levels are the result of a linear combination of an unknown number of independent underlying global transcriptional programs, called latent variables or factors (Figure 1). The contribution of each factor to the expression levels of the genes in each sample is given by the elements of the loadings matrix (arrows in Figure 1). Each sample contains, in addition, a given amount of expression that cannot be modelled by the latent variables, for example due to the presence of noise. FA models the covariance of a data matrix, as opposed to PCA, which attempts to summarize the total variance. Covariance in the mRNA expression levels has been shown to occur in proteins involved in related pathways and functions, as well as in proteins co-locating to the same organuli in the cell, and may be indicative of common regulatory mechanisms at the expression level[10]. By contrast, the specific variance in the expression of a given gene, not associated with the rest of the genes in the sample, is most likely related to artefacts in the chip or in data handling. We couple FA dimensionality reduction to clustering algorithms [11] to obtain clusters in sample space. For gene extraction, a multiple-testing corrected t-test (the so-called q-value) is employed. Finally, the genes assigned to each cluster are used for the detection of pathways predominantly activated in that cluster by finding statistically significant the GO [12] or GenMAPP terms associated to each cluster.
Figure 1 Graphical overview of Q-mode Factor Analysis (FA) [9]. Each sample is described by a vector xi, containing the expression levels for all genes in the chip. The complete expression for all samples is contained in the matrix X = {xi}. The expression levels of each sample are assumed to be generated by a linear combination of a small number of underlying transcriptional programs, the latent (non-observable) variables, contained in the set of vectors {Fi}, forming matrix F. The relative contribution of each program is given by the thickness of the arrows connecting factors and samples, stored in variables lij, altogether forming the loading matrix L. Each lij element can be understood as the correlation coefficient between the expression levels of the sample and the corresponding latent variable. Residuals are kept in vectors {εi}, giving rise to matrix E. Note that small loadings connecting a given sample (i.e., X4 with the factor model implies large residuals.
We first tested the approach by using a published dataset of microarray experiments in yeast [13], and then applied it to the analysis of human prostate cancer samples [14]. The yeast dataset is particularly relevant because the biochemistry of S. cereviseae is relatively well understood in comparison with other eukaryots, and the data set has been previously analyzed with other clustering techniques. From the application to the prostate cancer dataset, a number of significant gene outcomes highlighted by the algorithm have been corroborated experimentally a posteriori by expression analysis on an independent set of samples. The biological interpretation of the results lead us to propose that two major pathways are predominantly activated in organ-confined, non-metastatic prostate cancer: those regulated by androgen receptor and by receptor tyrosine kinases. We close this paper by discussing the implications of these findings.
Results and Discussion
Testing FADA with the yeast expression dataset
Our procedure is coded in a software package, FADA. We first tested FADA by analyzing the dataset of Gasch et al [13], who studied the transcriptional responses of S. cerevisiae to a variety of stress stimuli. The main results are in Table 1, and the genes most significantly associated to the clusters are in Table 1 of the Supporting Information. We discuss in what follows only the most salient features of the analysis for this set, corresponding to clusters 1, 2, 6, 8 and 16, from a biological viewpoint.
Table 1 Results of the analysis of the yeast dataset [13]. The different clusters found by FADA are shown, together with the significant GO terms associated to them. The samples belonging to each one of the clusters are also shown. The first column shows the cluster number; the second shows the conditions associated to that cluster; columns 3 to 5 show the Z-score of the GO terms associated to the cluster (see Methods) at the Cellular Component (CC), Biological Process (BP) and Molecular Function (MF) levels; columns 6 to 8 show the corresponding GO terms.
C CONDITIONS Z(CC) Z(BP) Z(MF) GO(CC) GO(BP) GO(MF)
1 Heat_Shock_05_minutes_hs.1, Heat_Shock_10_minutes_hs.1
Heat_Shock_15_minutes_hs.1, Heat_Shock_20_minutes_hs.1
Heat_Shock_30_minutes_hs.1, Heat_Shock_40_minutes_hs.1
Heat_Shock_60_minutes_hs.1, Heat_Shock_80_minutes_hs.1
Heat_Shock_015_minutes_hs.2, Heat_Shock_030minutes_hs.2
heat_shock_17_to_37._20_minutes, heat_shock_21_to_37._20_minutes heat_shock_25_to_37._20_minutes, heat_shock_29_to_37._20_minutes heat_shock_33_to_37._20_minutes 29C_to_33C_._5_minutes, 29C_to_33C_._15_minutes
29C_.1M_sorbitol_to_33C_._1M_sorbitol_._5_minutes
29C_.1M_sorbitol_to_33C_._1M_sorbitol_._15_minutes
29C_.1M_sorbitol_to_33C_._.NO_sorbitol_._5_minutes
dtt_240_min_dtt.2 1M_sorbitol_._5_min
1M_sorbitol_._15_min 1M_sorbitol_._30_min
1M_sorbitol_._45_min_ DBY7286_37degree_heat_._20_min
DBYmsn2.4._37degree_heat_._20_min
DBYmsn2.4_.real_strain._._37degrees_.20_min.
DBYyap1._37degree_heat_._20_min_.redo.
DBYyap1_._37degree_heat_.repeat., DBYyap1_._0.32_mM_H2O2_.20_min. Msn2_overexpression_.repeat.
Msn4_overexpression 3.25 5.39 1.16 nucleolus (325/88;0.81E-006) ribosome biog. & ass. (271/75; 0.57E-007) response to stress (214/52; 0.26E-003)
2 constant_0.32_mM_H2O2_.10_min._redo
constant_0.32_mM_H2O2_.20_min._redo
constant_0.32_mM_H2O2_.30_min._redo
constant_0.32_mM_H2O2_.40_min._rescan
constant_0.32_mM_H2O2_.50_min._redo
constant_0.32_mM_H2O2_.60_min._redo
1.5_mM_diamide_.5_min. 1.5_mM_diamide_.10_min.
1.5_mM_diamide_.20_min. 1.5_mM_diamide_.30_min.
1.5_mM_diamide_.40_min. 1.5_mM_diamide_.50_min.
1.5_mM_diamide_.60_min. 1.5_mM_diamide_.90_min.
DBY7286_._0.3_mM_H2O2_.20_min.
DBYmsn2msn4_.good_strain._._0.32_mM_H2O2
DBYmsn2.4_.real_strain._._0.32_mM_H2O2_.20_min.
DBYyap1._._0.3_mM_H2O2_.20_min. 0.39 9.71 11.40 protein catabolism (114/28; 0.74E-017) cell homeostasis (54/8; 0.10E-003) peptidase activity (125/25; 0.17E-012) oxidored. Act. (263/30; 0.36E-008)
3 2.5 mM_DTT_045_min_dtt.1 2.5 mM_DTT_060_min_dtt.1
2.5 mM_DTT_090_min_dtt.1 2.5 mM_DTT_120_min_dtt.1
2.5 mM_DTT_180_min_dtt.1 dtt_120_min_dtt.2 6.55 2.53 1.86 endoplasmic ret. (353/27; 0.11E-008)
4 constant_0.32_mM_H2O2_.80_min._redo
constant_0.32_mM_H2O2_.100_min._redo
constant_0.32_mM_H2O2_.120_min._redo
constant_0.32_mM_H2O2_.160_min._redo 1.25 0.63 1.58
5 37_deg_growth_ct.1 NA NA NA
6 Nitrogen_Depletion_8_h Nitrogen_Depletion_12_h
Nitrogen_Depletion_1_d Nitrogen_Depletion_2_d
Nitrogen_Depletion_3_d Nitrogen_Depletion_5_d 7.11 5.30 0.61 plasma membrane (197/16; 0.86E-005)
extracellular region (19/4; 0.85E-004) transcription (225/15 0.57E-003)
7 diauxic_shift_timecourse_18.5_h
diauxic_shift_timecourse_20.5_h YPD_6_h_ypd.2
YPD_8_h_ypd.2 YPD_10_h_ypd.2 YPD_12_h_ypd.2
YPD_1_d_ypd.2 YPD_2_d_ypd.2 YPD_3_d_ypd.2
YPD_5_d_ypd.2 YPD_stationary_phase_12_h_ypd.1
YPD_stationary_phase_1_d_ypd.1
YPD_stationary_phase_2_d_ypd.1
YPD_stationary_phase_3_d_ypd.1
YPD_stationary_phase_5_d_ypd.1
YPD_stationary_phase_7_d_ypd.1
YPD_stationary_phase_13_d_ypd.1
YPD_stationary_phase_22_d_ypd.1
YPD_stationary_phase_28_d_ypd.1
ethanol_vs._reference_pool_car.1
YP_ethanol_vs_reference_pool_car.2 9.91 #### 5.51 ribosome (368/126; 0.20E-010)
peroxisome (52/22; 0.66E-004) protein biosynthesis (493/168; 0.12E-012)
vitamin metabolism (48/20; 0.27E-003) structural mol act (359 /119; 0.13E-006)
8 aa_starv_0.5_h aa_starv_1_h aa_starv_2_h aa_starv_4_h
aa_starv_6_h Nitrogen_Depletion_30_min.
Nitrogen_Depletion_1_h Nitrogen_Depletion_2_h
Nitrogen_Depletion_4_h 3.60 #### 4.66 peroxisome (52/6; 0.14E-003)
plasma membrane (197/17; 0.35E-006) aminoacid metab (173/42; 0.24E-031) transporter act (343/27; 0.12E-004)
lyase activity (97/12; 0.24E-004)
9 33C_vs._30C_._90_minutes dtt_480_min_dtt.2
steady_state_36_dec_C_ct.2
steady_state_36_dec_C_ct.2_.repeat_hyb._ 0.12 -1.42 -0.24
10 dtt_060_min_dtt.2 YP_galactose_vs_reference_pool_car.2
YP_raffinose_vs_reference_pool_car.2 1.43 0.12 -1.53
11 Diauxic_Shift_Timecourse_._0_h
diauxic_shift_timecourse_9.5_h diauxic_shift_timecourse11.5_ 2.86 1.16 0.44 vacuole (140/6; 0.63E-003)
12 YPD_stationary_phase_2_h_ypd.1
YPD_stationary_phase_4_h_ypd.1 -0.39 3.11 2.06 electron transport (14/1; 0.91E-003)
13 diauxic_shift_timecourse_13.5_h
diauxic_shift_timecourse_15.5_h
YPD_stationary_phase_8_h_ypd.1 -0.51 -0.27 -0.20
14 1M_sorbitol_._60_min 1M_sorbitol_._90_min
1M_sorbitol_._120_min 1.21 2.46 1.33 cell cycle (115/4; 0.11E-003)
15 YPD_2_h_ypd.2 YPD_4_h_ypd.2 YAP1_overexpression -0.34 0.20 0.56
16 1_mM_Menadione_.10_min.redo
1_mM_Menadione_.20_min._redo
1_mM_Menadione_.30_min._redo
1mM_Menadione_.40_min._redo
1_mM_Menadione_.50_min.redo
1_mM_Menadione_.80_min._redo
1_mM_Menadione_.105_min._redo
1_mM_Menadione_.120_min.redo
1_mM_Menadione_.160_min._redo 4.71 6.25 2.67 mitochondrion (732/88; 0.49E-003)
Golgi apparatus (90/17; 0.63E-003) vesicle-med. Transp. (190/31; 0.84E-004)
17 Heat_Shock_000_minutes_hs.2
Heat_Shock_000_minutes_hs.2.1
Heat_Shock_000_minutes_hs.2.2
37C_to_25C_shock_._15_min 37C_to_25C_shock_._30_min
37C_to_25C_shock_._45_min 37C_to_25C_shock_._60_min
37C_to_25C_shock_._90_min dtt_000_min_dtt.2
dtt_015_min_dtt.2 dtt_030_min_dtt.2
steady_state_21_dec_C_ct.2 steady_state_25_dec_C_ct.2
steady_state_29_dec_C_ct.2 5.33 #### 5.49 ribosome (368/145; 0.26E-007) nucleolus (325/138; 0.15E-009) ribosome biog & ass (271/121; 0.87E-011)
RNA metabolism (382/148; 0.14E-007)
protein biosynthesis (493/194 0.12E-010) structural mol. act. (359/123; 0.11E-003)
RNA binding (268/96; 0.93E-004)
18 YP_fructose_vs_reference_pool_car.2
YP_glucose_vs_reference_pool_car.2
YP_mannose_vs_reference_pool_car.2
YP_sucrose_vs_reference_pool_car.2 2.12 0.31 0.32 mitochondrial membr (136/7; 0.59E-004)
19 Hypo.osmotic_shock_._15_min
Hypo.osmotic_shock_._30_min
Hypo.osmotic_shock_._45_min
Hypo.osmotic_shock_._60_min 3.08 5.06 3.67 bud (59/6; 0.40E-003)
nucleolus (325/21; 0.525E-005) ribosome biog & ass (271/24; 0.12E-007)
RNA metabolism (382/22; 0.86E-004
20 Heat_Shock_060_minutes_hs.2 NA NA NA
21 17_deg_growth_ct.1 21_deg_growth_ct.1 25_deg_growth_ct.1
29_deg_growth_ct.1 1.67 1.54 -1.34
22 steady_state_15_dec_C_ct.2 steady_state_17_dec_C_ct.2 0.51 -1.14 0.48
23 2.5mM_DTT_005_min_dtt.1 2.5mM_DTT_015_min_dtt.1
2.5mM_DTT_030_min_dtt.1 0.03 1.28 0.32
24 galactose_vs._reference_pool_car.1
glucose_vs._reference_pool_car.1
mannose_vs._reference_pool_car.1
raffinose_vs._reference_pool_car.1
sucrose_vs._reference_pool_car.1 steady_state_33_dec_C_ct.2 8.69 5.93 6.57 cell cortex (39/7; 0.97E-003)
cytoplasmic vesicle (52/11; 0.15E-004)
bud (59/10; 0.27E-003)
endomembrane syst (76/15; 0.21E-005) cytokinesis (52/10; 0.25E-003)
nuclear org & biog (105/16; 0.20E-003)
vesicle-med transp (190/23; 0.55E-003) helicase activity (71/13; 0.21E-004)
25 29C_to_33C_._30_minutes
29C_.1M_sorbitol_to_33C_._1M_sorbitol_._30_minutes
29C_.1M_sorbitol_to_33C_._.NO_sorbitol_._15_minute
29C_.1M_sorbitol_to_33C_._.NO_sorbitol_._30_minute
Hypo.osmotic_shock_._5_min steady.state_1M_sorbitol 0.39 2.96 3.27 DNA metabolism (221/8; 0.78E-003) DNA binding (146/7; 0.21E-003)
26 Heat_Shock_005_minutes_hs.2 NA NA NA
(a) GO terms discussed in the text are shown in bold. Together with each GO term, we show the number of genes corresponding to that term; the number of genes of that term in the cluster; and the corresponding P-value, according to the hipergeometric distribution.
(b) NA: Data Not Available. No significant genes (according to the q-value cutoff) could be found.
Cluster 1 encompasses responses to heat shock, DTT (late), sorbitol (early response), stationary culture (late), and overexpression of Msn2p and Msn4p. The significant GO terms [12] automatically detected by FADA indicate that this grouping is related to a common environmental stress response (Table 1) (ESR or CER in the case of S. cerevisae, CESR in the case of S. pombe [13,15,16]). Inspection of the top selected genes (Table 1, supplementary material) confirms this assignment, as most of them are known to be transcriptionally regulated through the stress response element (STRE), recognized by Msn2p and Msn4p [17,18]. Thus, Cluster 1 corresponds largely to a "core" ESR, induced by a variety of stimuli, including "early" time points of osmotic stress and "late" time points of DTT treatment and stationary culture. A relatively late induction of ESR by DTT has been noted previously, with suggestions that ESR could be a secondary response to the exposure of this reducing agent [13]. Conversely, hyperosmotic shock is known to induce a rapid and strong expression of ESR [13,15].
Clusters 2 and 16 correspond to responses to three oxidizing agents: hydrogen peroxide, which generates peroxides and hydroxyl radicals; menadione, a generator of superoxide; and diamide, a thiol reducing agent. FADA groups together responses to H2O2 and diamide (Cluster 2), while defining a distinct group for responses to menadione (Cluster 16). It is well known that several organisms use distinct sensing and response systems to discriminate among different degrees of oxidative injury. In S. pombe, reponses to low concentrations of H2O2 and to diamide depend on the b-Zip transcription factor pap1, while reponses to higher concentrations of H2O2 utilize a different transcription factor, atf1 [19]. The S. cerevisiae homologue of S. pombe pap1 is Yap1p, a transcription factor regulated by oxidation, formation of intramolecular disulfide bonds at its carboxy terminus and nuclear translocation upon exposure to H2O2 and diamide [20-22]. On the other hand, of the b-ZIP proteins in S. cerevisiae, Yap3p is most similar to atf1 in its carboxy terminus, suggesting that both atf1 and Yap3p could be subject to a similar redox regulation. Interestingly, YAP1 is upregulated in Cluster 2 (H2O2 and diamide), but not in Cluster 8 (menadione), while Yap3 is upregulated in the latter Cluster (Table 1 of Supplementary Data). Moreover, several of the genes most relevant to Cluster 2 are known to respond to mild oxidative stress, and are controlled by Yap1p [23]. The statistically significant GO-terms selected are related to "oxidoreductase" and "peptidase" activities. This includes genes regulating the thioredoxin and glutathione biosynthesis, genes for heat shock proteins, and a large number of genes involved in proteasome function and ubiquitin-dependent protein degradation (Table 1 of Supplementary Material).
Cluster 6 includes cultures at late times of nitrogen starvation. Many of the relevant genes in this group code for enzymes for the utilization and enhanced transport of poor nitrogen sources, such as allantoin or urea (Table 1 of Supplementary Material). Other upregulated genes include those required for different stages of meiosis (chromosome pairing, recombination and segregation; anaphase; or nucleokinesis), sporulation, autophagy, or genes that regulate vesicle and peroxisome structure and dynamics. Among these genes are also transcriptional regulators with major roles in the control of several of these processes, such as UGA3, DAL81 (allantoin metabolism), or IME1, RIM101 and SPO1 (meiosis and sporulation). This is consistent with the development of a classical response to nitrogen starvation in the absence of fermentable carbon sources, which leads to meiosis and sporulation [24-27]. FADA also suggests that this response to nitrogen starvation becomes most prominent at relatively late times, when it can be distinguished from the early, relatively non-specific response to nutrient deprivation [25,26]. In fact, FADA finds "transcription" and "sporulation" as significant GO-terms (Table 1).
Cluster 8 aggregates samples from early stages of both early response to aminoacid and nitrogen starvation. FADA finds a significant overrepresentation of genes for amino acid biosynthetic pathways (Table 1), consistent with the fact that deprivation of nutrients, including nitrogen and carbon sources, is recognized by several sensing systems regulating rapamycin-sensitive TOR kinase [28]. This lipid-dependent kinase derepresses translation of the GATA transcription factor Gcn4p [29,30], which controls expression of many genes, including enzymes involved in amino acid biosynthesis [31]. Thus, the selection of genes in Cluster 8 is consistent with known Gcn4p-dependent responses to nutrient and nitrogen starvation [31].
Altogether, these results indicate that the automatic analysis provided by FADA yields results consistent with the known biochemistry of yeast.
Application of FADA to the prostate cancer dataset
We next applied FADA to the dataset published by Welsh et al. [14], for the analysis of transcripts associated with prostate cancer. Samples were classified into two major branches: samples from cultured cells, and samples from tissues, which in turn could be further bifurcated into two well-supported branches, one corresponding to samples enriched for carcinomatous cells and one for non-neoplastic prostate cells (Figure 2). The first-level grouping into cultured vs. non-cultured samples most likely reflects the profound impact of culturing procedures on the transcriptional profiles of the different cell types. Within the cultured cells subgroup, samples were generally clustered according to cell type, with haematopoietic cell lines forming well-clustered groups and epithelial and fibroblastic prostate-derived cells clustering together with endothelial cells. A separate cluster was formed by the androgen-sensitive epithelial cell line LNCaP, the prostate cancer cell lines included in the study. The genes most significantly contributing to each sample cluster were analyzed for their participation in the pathways contained both in GenMAPP [32], and GO (Tables 2 and 3). Since pathway categorization is a difficult problem, as partition of the global interaction network in "parts" inevitably introduces artefacts, we also proceeded to a detailed, gene-by-gene inspection of the most discriminative genes based on inspection of literature data.
Figure 2 Dendrogram for the Welsh dataset [14]. The dashed line indicates the thresholding used to define the clusters.
Table 2 Results of the analysis of the Welsh dataset for up-regulated genes. The different sample clusters found by FADA are shown, together with the significant GO and GenMAPP terms associated to them. The first column shows the cluster number; the second shows the samples associated to that cluster; columns 3 and 4 show the z-score of the GenMAPP and GO terms associated to the cluster (see Methods); columns 5 to 8 show the corresponding GenMAPP and GO terms selected.
C SAMPLES Z(GM) Z(GO) GENMAPP GO(MF) GO(BP) GO(CC)
1 LNCaP_A, LNCaP_B, LNCaP_+_DHT 7.22 5.09
-0.39
2.59 RNA_transcription_React. (2.40e-03)
Electron_Transport_Chain (5.74e-10) oxidoreductase activity (1.58e-04)
carrier activity (1.87e-04)
ATPase activity, coupled to transmembrane movement of substances (4.23e-03)
transcriptional activator activity (2.09e-03) intracellular organelle (5.48e-04)
2 CAF_1598, BPHF_1598
CAF_1303, CAF_1852
PrSC_A, PrSC_B, CAF_2585, Du145, PC3, HUVEC_A, HUVEC_B, hPr1, PrEC 3.83 4.66
2.74
5.13 Hypertrophy_model (8.97e-03)
Proteasome_Degradation (1.44e-08)
Cell_cycle_KEGG (7.86e-03)
Pentose_Phosphate_Pathway (4.89e-03) Enzyme inhibitor activity (4.96e-04)
hydrolase activity (9.26e-03)
small protein conjugating enzyme activity (3.54e-04)
structural constituent of cytoskeleton (2.73e-03)
nucleotide binding (1.17e-05) regulation of cellular process (4.31e-03)
cellular physiological process (1.93e-04) signalosome complex (7.27e-03)
membrane coat adaptor complex (4.24e-03)
tubulin (5.06e-06)
proteasome complex (sensu Eukaryota)(3.69e-07)
Arp2/3 protein complex (4.19e-05)
3 B_CELLS_A, B_CELLS_B B_CELLS_C, MOLT4, HL60 4.27 0.42
0.18
-0.04 mRNA_processing_React. (2.10e-03)
G1_to_S_cell_cycle_React. (9.25e-05)
Cell_cycle_KEGG (1.94e-04)
Small_ligand_GPCRs (5.52e-03)
Ovarian_Infertility_Genes (5.19e-03)
GPCRDB_Class_C_Metabotropic_glutamate_pheromone (4.72e-04)
4 T4, T7 T3, T5, T1, T27, T10, T9, T13A, T13B, T22, T12, T29, T8, T31, T30, T26, T19, T16, T23, T6, T24, T21, T11, T17 2.06 1.93
0.39
0.76 Fatty_Acid_Degradation (8.48e-03)
Hypertrophy_model (3.34e-03)
Eicosanoid_Synthesis (3.56e-03) steroid binding (7.57e-03)
isomerase activity (7.99e-04)
vitamin binding (6.99e-04)
5 N2, N1, N5, N3, N9, N8, N7, N10, N4 6.86 2.52
-0.31
-0.40 Smooth_muscle_contraction (6.55e-03)
Calcium_reg_in_card_cells (3.97e-03) channel or pore class transporter activity (1.75e-03)
structural constituent of cytoskeleton (2.75e-04)
(a) GO or GenMAPP terms are discussed in the text. Together with each term, we show the corresponding P-value, according to the hipergeometric distribution (see Methods).
(b) The Z(GO) column shows the Z-scores corresponding to Molecular Function (MF), Biological Process (BP), and Celular Component (CC), respectively, obtained for each cluster. The Z(GM) column refers to the Z-score corresponding to the GenMAPP terms.
Table 3 Results of the analysis of the Welsh dataset for down-regulated genes. The different sample clusters found by FADA are shown, together with the significant GO and GenMAPP terms associated to them. The first column shows the cluster number; the second shows the samples associated to that cluster; columns 3 and 4 show the Z-score of the GenMAPP and GO terms associated to the cluster (see Methods); columns 5 to 8 show the corresponding GenMAPP and GO terms selected.
C SAMPLES Z(GM) Z(GO) GENMAPP GO(MF) GO(BP) GO(CC)
1 LNCaP_A, LNCaP_B, LNCaP_+_DHT 0.36 -0.10
-0.54
4.09 extracellular space (3.74e-04)
MHC protein complex (4.71e-04)
2 CAF_1598, BPHF_1598, CAF_1303, CAF_1852, PrSC_A, PrSC_B, CAF_2585, Du145, PC3, HUVEC_A, HUVEC_B, hPr1, PrEC 2.33 2.43
0.68
-0.50 Hs_GPCRDB_Other (1.96e-03)
Hs_Ribosomal_Proteins (1.03e-08) structural constituent of ribosome (2.57e-05)
SH3/SH2 adaptor activity (6.48e-03)
nucleic acid binding (7.71e-04)
oxygen transporter activity (6.93e-03)
nucleobase, nucleoside, nucleotide and nucleic acid transporter activity (6.93e-03)
3 B_CELLS_A, B_CELLS_B, B_CELLS_C, MOLT4, HL60 0.91 0.85
-0.65
-0.91
4 T4, T7, T3, T5, T1, T27, T10, T9, T13A, T13B, T22, T12, T29, T8, T31, T30, T26, T19, T16, T23, T6, T24, T21, T11, T17 4.30 1.04
1.86
7.71 G1_to_S_cell_cycle_React (7.76e-04)
Glycolysis_and_Gluconeogenesis (4.02e-04)
Cell_cycle_KEGG (7.03e-08)
DNA_replication_Reactome (2.01e-03) signalosome complex (5.23e-04)
intracellular (4.66e-04)
tubulin (2.97e-03)
proteasome complex (sensu Eukaryota) (2.47e-04)
proton-transporting ATP synthase complex (9.01e-04)
Arp2/3 protein complex (5.23e-04)
5 N2, N1, N5, N3, N9, N8, N7, N10, N4 -0.21 -0.65
3.83
-0.53 metabolism (3.90e-04)
(a) GO or GenMAPP terms are discussed in the text. Together with each term, we show the corresponding P-value, according to the hipergeometric distribution (see Methods).
(b) The Z(GO) column shows the Z-scores corresponding to Molecular Function (MF), Biological Process (BP), and Cellular Component (CC), respectively, obtained for each cluster. The Z(GM) column refers to the Z-score corresponding to the GenMAPP terms.
Cluster 1 corresponds to the LNCaP cluster. It is placed in a branch distinctly separated from the rest of the cultured prostatic cells. LNCaP cells were originally derived from metastatic prostate cells, presumably of epithelial origin [33] and respond to androgens through its cognate receptor [34]. FADA found significant overrepresentation of upregulated genes coding for proteins that participate in electron transport and ATP generation, both when using GenMAPP and GO annotations (Figure 3, 4 and Table 2). Other sets of genes likely relevant to the LNCaP cluster, but not highlighted in the pathway mapping protocol, are those for proteins in steroid metabolism and signalling, such UDP glycosyltransferases B15 (Table 2 of the Supporting Information). Cluster 2 includes mesenchimal, epithelial, and endothelial cells. This cluster shows a bias for genes and pathways involved in ubiquitin and proteasome-dependent protein degradation, cell cycle regulation, inflammatory responses and cell-matrix interaction. Cluster 3 (hematopoietic cells) showed a significant bias in genes and pathways involved in cell cycle regulation and RNA processing. The selected genes included known markers of differentiation of B cell, T cell or myelomonocytic lineages. Examples are genes for immunoglobulin, histocompatibility antigens, haematopoietic-specific cytokines and their receptors, and regulatory proteins known to play significant roles in such lineages in processes such as signal transduction or cytoskeletal dynamics.
Figure 3 Expression levels for the 20 most relevant genes selected in each cluster for the Welsh dataset. Gene descriptions can be found in Table 2 of the Supporting Information. A) (See Figure 3) Up-regulated; B) Down-regulated. (See Figure 4)
Regarding Cluster 4 (prostate tumor tissue), GenMAPP mapping finds significant overexpression of enzymes related to fatty acid metabolism (Table 2). Other genes and KEGG pathways with a significantly biased association with cluster 4 are those for ribosomal function and fatty acid synthesis (Table 2 of the Supporting Information). The upregulation of these two functions in prostate cancer has been noted previously [14,35]. In addition, GO mapping finds overrepresentation of genes for proteins directly involved in steroid receptor recognition, including androgen receptor and estrogen receptor β. This is confirmed by a survey of the list of selected genes, where one can find a number of proteins involved in steroid signalling, including the coactivators GRIP1 and NRIP1, and genes that have been described as transcriptional targets of these pathways [36], such as the secreted proteases KLK2 and KLK3, and protein IQGAP, involved in cytoskeletal dynamics [37], or the enzymes fatty acid CoA-ligase or androgen-regulated short chain dehydrogenase (Table 2 of the Supporting Information). A second group of genes significantly contributing to this cluster are those for cell surface polypeptide growth factor receptors, associated signalling molecules and regulators, and known transcriptional targets for these pathways. These include the receptor tyrosine kinase partner ERBB3 (HER3), the calmodulin-dependent kinase activator CAMKK2, or the signalling modulators RAPGA1 and PDE3B (Table 2 of the Supporting Information).
Finally, the highest ranking genes for samples from normal prostate tissue (Cluster 5) correspond, according to GO, to proteins involved in the control of cytoskeletal architecture and dynamics in muscle cells (Table 2). GenMAPP finds a significant overrepresentation of muscle-associated functions. The implication is that, in these experiments, normal prostate tissue samples possibly are strongly enriched for muscle cells. This strong overrepresentation of genes corresponding to a smooth muscle phenotype suggests that the non-neoplastic tissues used correspond to areas of prostate hyperplasia or adenoma derived from the transition zone, in which smooth muscle cells are often major contributors [38]. In practical terms, this suggests that these experiments may be used with caution in the comparison of tumor epithelial cells with corresponding normal epithelial counterparts.
In recent years, several transcriptional profiling studies have been performed in prostate cancer, aimed at the identification of novel tumor markers [14,39-41] or prognostic signatures [42-44]. So far, only one study has systematically searched for overrepresented biochemical pathways in a meta-analysis of four previously published prostate cancer transcriptional profiling studies [45]. This study used KEGG as reference pathway database, which is biased towards metabolic pathways [46]. Our study, however, focuses on GenMapp and GO terms, and therefore on the identification of signalling pathways.
Signalling pathways in prostate cancer and their experimental validation
In order to validate the pathways found to be overrepresented in prostate tumor samples, we used real-time RT-PCR. We chose for our analysis the genes for hepsin, KLK3 (PSA), ERBB3 (HER3), IQGAP2, and POR/ARFAPTIN2. Hepsin was found to be overexpressed in most tumor samples, and validated by immunohistochemical analysis [14]. This gene has been shown to be overexpressed in prostate cancer by several other groups. KLK3 (PSA) is the marker par excellence of prostate epithelial activity and cellular bulk, and detection of its serum protein levels is the best available marker for monitoring prostate cancer [47]. HER3 is a receptor for the paracrine growth factor neregulin-1, and a transmembrane protein that tethers the ligand to its dimerization partners, the receptor tyrosine kinases HER2 and HER4 [48], and known to play important roles in the development and progression of the malignant phenotype in breast cancer [49]. The abnormal expression and activity of HER2 has been studied extensively in the context of prostate cancer [50], being found overexpressed in advanced tumors, either metastatic or homone-independent, but infrequently in primary, organ-confined tumors. More controversial is the information available on the role of HER3, with reports of its overexpression in prostate cancer together with HER2, HER4, or both [51,52], but also of its overexpression only in metastatic tumors, in particular of a truncated form corresponding to the extracellular domains of HER3 [53]. Furthermore, several transcriptional profiling analyses have found overexpression of this gene in prostate cancer. IQGAP2 is a calmodulin-binding protein that participates in cell signalling and modulation of cytoskeletal dynamics [37], and its activity has been reported to be positively [54] and negatively associated with neoplastic phenotype. POR1/ARFAPTIN2, a Rac1-interacting protein [55], is a regulator of cytoskeletal dynamics that so far has not been associated with any particular type of neoplasia.
The results of semiquantitative real-time RT-PCR on our samples indicate that hepsin is significantly overexpressed in 14 out of 14 cases, IQGAP2 in 8 of 14, and HER3 in 10 of 14 cases (Figure 5). Other genes analyzed, such as KLK3 (PSA), HER2 or the steroid receptors androgen receptor, estrogen receptor α or estrogen receptor β are less frequently overexpressed in these tumors. Levels of desmin transcripts were determined as an index of the contribution of stromal cells, suggesting that the overexpression of the analyzed genes are detected in tumor samples even in the presence of substantial stromal contamination (Figure 5). Of particular interest is the observed upregulation of HER3 in prostate tumor tissues relative to normal tissues. The HER3/ErbB3 protein has impaired intrinsic kinase activity [56], and it appears to function in signal transduction by tethering the ligand to other members of the HER family of receptors, with preference for HER2/ErbB2 [57]. Increased levels of expression of HER3 are seen in many tumors that express HER2 [58], and it is widely assumed that the signalling and/or oncogenic functions reside in the corresponding heterodimer, rather than in either individual receptor [59,60]. Recent experimental evidence further highlights the importance of HER3 in conferring a malignant phenotype and a hormone-refractory state to prostate epithelial cells [61]. Thus, whenever HER3 is expressed it is reasonable to expect co-expression of at least one other member of the HER family. Therefore, we determined by real-time RT-PCR the relative expression in our prostate tissue samples of the genes for all four members of the HER family of receptor tyrosine kinases. Our results show that HER4 is expressed at increased levels in 10 of 14 prostate tumor samples (Fig. 5A, B), whereas HER2/ErbB2 and EGFR are overexpressed in 3 of the 14 samples analyzed. Seven samples simultaneously overexpressed HER3 and HER4, of which 2 overexpressed all four members of the HER family (Fig. 5A, B). None of the samples overexpressed the pairs HER3 and HER2, or HER3 and EGFR, without overexpressing at the same time one of the other members of the family (Fig. 5A, B).
Figure 4 Figure 5 Validation of genes selected by FADA from the Welsh et al. dataset [14] as overexpressed in prostate cancer. (A) RT-PCR was applied to 14 paired prostate tumor – normal prostate samples to determine the expression levels of a selection of genes shown by FADA as significantly overrepresented in prostate cancer (HPN, KLK3, IQGAP2, POR1 and HER3), and additional genes relevant to this tumor (genes for the receptor tyrosine kinases EGFR, HER2, HER4, and genes for the steroid hormone receptors AR, ERα and ERβ). The expression values for each gene, previously normalized with respect to the S14r expression level in each sample, are shown as ratios of the normalized values in prostate cancer vs. values in the matching normal prostate tissue. Quantitation of desmin expression levels was used to assess the degree of contribution of stromal components in the samples analyzed. Values equal to or above 100-fold are shown as 100. (B) Heatmap representation of the same data (color scale as shown below). (C) Real-time PCR analysis for HER3 transcript levels of laser microdissected tumor and normal samples, compared with relative transcript levels in enriched (non-microdissected) tissues from the same cases. (D) Immunohistochemical analysis of HER3 on paraffin-embedded prostate tissue sections arranged in tissue microarrays (see Methods). Left, low magnification image (×100) of one case, with weak staining for HER3 in normal glands (n), and a strong staining in tumor epithelial cells (t). Right, higher magnification (×400) of a second case.
As mentioned in the Methods section, both tumor and normal tissues were carefully chosen to have similar representation of epithelial compartment. However, to further ensure that the observed expression of HER3 was not due to a dilution effect of normal epithelial cells by stroma, we performed real-time PCR analysis of laser microdissected samples. For this, we selected four samples that had shown overexpression of HER3 in the enriched tumor samples described above, and two that had levels that did not differ significantly from non-tumor containing (normal) matched tissues. Of the four samples in which the enriched tumor tissue had shown increased levels of HER3 transcript, three microdissected samples overexpressed HER3 (Fig. 5C). In two of the microdissected samples, HER3 transcript levels were equal in normal and tumor microdissected epithelia, and this also corresponded to samples in which HER3 levels did not differ significantly between enriched tumor and normal prostate tissues (Fig. 5C). This analysis showed that overexpression of HER3 in prostate tumor tissues is not due to simple enrichment of epithelial cells in comparison with non-tumor tissues. To further confirm the cell type expressing HER3 in prostate tissues, immunohistochemical analysis with a monoclonal antibody to HER3 was performed on 16 prostate samples, arranged in duplicate 1-mm diameter cores in tissue microarrays, in which both tumor and normal glands were present. HER3 protein was found clearly overexpressed in tumor epithelia in 13 of the 16 cases (81.2%), showing juxtamembrane and finely granular cytoplasmic patterns (Fig. 5D). In all cases, normal epithelia showed weak reactivities for HER3 (Fig. 5D).
In summary, our transcriptome re-analysis, validated by real-time RT-PCR of non-microdissected and microdissected samples and by immunohistochemical analysis, significantly reinforces previous immunohistochemical studies that reported high levels of expression of HER3 and HER4 in primary prostate cancer [51,52].
Conclusion
We have shown that the method presented here for the analysis of expression microarray data permits the classification of samples into meaningful categories and, simultaneously, to identify a subset of genes and their assignment to pathways most significantly contributing to the corresponding phenotypes, while allowing for a given gene to participate as significant in more than one cluster of samples. The analysis of the yeast dataset validates the approach. Our results are consistent with biochemical pathways known to be activated in the different stress conditions analyzed, and the clustering of samples reflects the underlying similarity of the biochemical responses. In the application to the prostate cancer dataset, we have found that two pathways, one modulated by androgen receptor and a second one by signals that originate from cell surface growth factor receptors, are prominently active in the organ-confined, non-metastatic prostate cancer samples analyzed. The latter pathway has been reported to be spuriously active in at least a subset of prostate tumors that have progressed to invasive and hormone-independent states [62]. Our results suggest that such altered activation may already be present in primary tumors. Although a prevailing model for prostate tumor progression is that acquisition of the capacity for metastatic and hormone independent growth proceeds through selection of rare populations of cells concealed among primary tumor cells, there is also evidence that a transcriptional program for metastasis may already be present in the bulk of primary tumors at the time of diagnosis [63,64]. Our analysis would be more consistent with the latter model.
Finally, we have unveiled and validated several markers highlighted by the analysis of the prostate cancer dataset. While several of these genes were identified in the original analysis of the data [14], others are revealed here, notably HER3, IQGAP2 and POR1, the biologically most relevant being HER3. With an external dataset, we have found that prostate cancer samples frequently co-overexpress HER3 and HER4, accompanied less frequently by increased expression of EGFR or HER2. Overexpression of HER2 and consequent increased signalling have been associated with advanced prostate cancer, development of hormone independent state and poor prognosis [65,66], but is infrequently observed in primary tumors [67,68]. On the other hand, our results suggest that, in primary prostate cancer, HER3, together or not with HER4, rather than receptor complexes involving HER2, could play important roles in the biology of these tumors.
Materials and methods
Datasets
The S. cereviseae dataset consists of transcriptional responses of the yeast S. cerevisiae to environmental stress [13]. It originally consists of spotted array measurements of 6152 genes in 173 experimental conditions that include temperature shocks, hyper and hypoosmotic shocks, exposure to various agents such as peroxide, menadione, diamide, dithiothreitol, amino acid starvation, nitrogen source depletion and progression into stationary phase. Log-ratios were preprocessed following several steps: first data from genes with missing values were filtered out, and their missing values estimated with LSimpute [69] using the 'Adaptive' method. Next, ratios were computed from the log-ratios and quantile-normalized (experiment-wise) using the normalizeQuantile function from the R package [70], so that all experiments had the same average sample distribution. Finally, ratios were log transformed again.
The prostate cancer dataset chosen is described in [14]. It was originally obtained by hybridizations on Affymetrix U95A oligonuleotide arrays with probes for a total of 55 samples. Intensity values were preprocessed following several steps: first intensity data were thresholded, with intensities below 10 fixed at 10 and values above 16000 fixed at 16000. The thresholded values were log-transformed and then centered by the median of all experiments. Finally, genes were subjected to z-transformation (per gene basis).
Determination of genotypically coherent groups of samples
Q-mode Factor Analysis (FA) [9] seeks to find an underlying orthogonal factor model of an original X-matrix nxm (where n are the number of samples and m the number of mRNA levels measured) of the form:
X = LF + E
L is the loadings matrix of size nxk, where k is the number of factors, and F the scores matrix of size kxm, while E is the residual matrix, which contains both the specific variance of the individual genes and the errors in the model (see Figure 1). We used the so-called principal factor solution to solve this factor model. Specifically, in a first step, and based on the correlation matrix R derived from X, communalities (i.e. the proportion of the variance explained by common factors) were computed from the multiple squared correlation coefficient between the ith variable and the rest. These communalities replaced the diagonal entries of the correlation matrix, which was subjected to diagonalization. New communalities were computed from the loadings at the chosen dimensionality, obtained by scaling the eigenvector matrix (P), as follows:
L = P Λ1/2
The new communalities again replaced the diagonal entries, and the process was iterated until convergence. Finally, we proceeded to rotate the factor loadings by means of a varimax rotation [9]. The effect of this rotation is to maximally align each of the samples with one factor in order to simplify the factor model and make it more readily interpretable. Phyletic trees were derived by clustering samples in loadings space at the optimal dimensionality using average linkage [4,11]. When needed, bootstrap values were computed by selecting random subsets of 90% of the genes [71]. Distribution of trees and frequency of each branch in the original tree were recorded using CONSENSE, program included in the PHYLIP package [72].
Selecting genes associated to each cluster
Once sample clusters are defined, these are used to identify groups of genes contributing heavily to the specific character of different groups. Each gene on the list is subjected to a Student's t-test that measures the differential expression of the gene in the cluster as compared with the rest of the samples. t-test scores were transformed to q-values, which include multiple testing correction. The q-value is similar to the well known P-value, except that it is a measure of significance in terms of the false discovery rate, rather than the false positive rate [73]. Genes with a q-value < 10-4 were taken as differentially expressed for that particular cluster.
Assigning pathways to gene clusters
The association between selected genes and biological functions was established by determining the hypergeometric distribution of genes on the annotation databases GO [12] or GenMapp [32]. With this distribution we computed the probability that at least x genes annotated within a given biological function according to GO (or GenMapp) in a cluster of size n (the total number of genes per cluster selected in the previous step) can be obtained by chance, given a population of N genes under consideration and given A, the total number of genes within N with that particular annotation. These P-values are obtained according to:
An aggregated score for each cluster from the significant P-values (i.e., those below 10-2) is computed as follows:
s0 = ∑-ln p(x; N, A, n)
The significance of this score is established by simulation. We randomly selected 100 samples of size n genes each (the number of genes per cluster selected according to the q-value) and computed a new s-score (sr) for each one. The Z-score is finally computed as:
z = (so - <sr>)/σr
Z-scores > 2.0 are taken as indicative of significant association between the samples in the cluster and the set of pathways uncovered.
We should emphasize that in spite of the apparent intricacy of the computational procedure, the computational complexity is similar to other biclustering methods, and operates within a highly constrained parameter space: in the factor analysis part of the program only the percentage of variance employed should be set, yielding a reduced number of dimensions or latent variables, usually below 5; the number of clusters is automatically determined in this space from the c-index, and has no free parameters, and the selection of genes relevant for each cluster only depends on the cutoff employed in the q-value.
Real-time RT-PCR
We used RT-PCR with either TaqMan probes or by SYBRGreen incorporation to determine the expression levels of selected genes on samples unrelated to the original study by Welsh et al. In each instance, the tumor sample and its matching normal counterpart were obtained from the same case, upon removal by radical prostatectomy. Serial sections from all normal counterparts to the tumor tissues were stained and analyzed to confirm that normal prostate glands and epithelial cells were present in near-normal patterns, and that they contained less than 1% of cells or structures with carcinomatous appearance. In addition, samples were chosen such that the tumor and normal counterparts in each case had approximately equal representations of the epithelial compartment, as assessed microscopically. RNA was isolated from corresponding frozen serial sections, and controlled for quality on a 2100 BioAnalyzer instrument (Agilent, Palo Alto, CA). For each sample, 0.5 μg of total RNA was reverse transcribed by priming with random hexamers at 42°C for 50 minutes, followed by treatment with RNase at 37°C for 20 min. The resulting cDNAs were used as templates in PCR reactions with gene-specific primers. Real-time PCR was performed on ABI PRISM 7700 (Applied Biosystems, Foster City, CA) or DNA Engine Opticon (MJ Research, Waltham, MA) instruments. TaqMan probes and their corresponding primer sets were obtained from Applied Biosystems. Thermal cycler conditions were 95°C for 10 min and 40 cycles of 95°C for 15 sec and 60°C for 1 minute for TaqMan assays. In the case of SYBRGreen reactions, the conditions were 95°C for 15 min, and 40 cycles of 95°C for 15 sec, 55°C for 30 sec and 72°C for 30 sec. All determinations were performed in triplicate and in at least two independent experiments. Since the relative amplification efficiencies of target and reference samples were found to be approximately equal, the ΔΔCt method was applied to estimate relative transcript levels. Levels of ribosomal S14r amplification were used as an endogenous reference to normalize each sample value of Ct (threshold cycle) and normal tissues were used as calibrators for their tumoral counterparts in each case. The final results, expressed as n-fold differences in target gene expression were calculated as follows:
nTARGET = 2-[(Ct target - Ct reference)TUMORAL - (Ct target - Ct reference)NORMAL]
Laser capture microdissection
Prostate tissues were obtained by punch sections of radical prostatectomies and snap-frozen in isopentane at -50°C embedded in OCT-containing cryomolds. 8 μM cryosections were mounted onto plastic membrane-covered glass slides (PALM Mikrolaser Technology, Bernried, Germany), fixed for 3 minutes in 70% ethanol, stained with Mayer's hematoxilin, dehydrated, air-dried for 10 minutes and stored at -80°C until used. Laser catapulting microdissection was performed with a PALM MicroBeam Systems instrument. 2 to 5 × 104 normal or carcinomatous epithelial cells were collected and estimated to be >99% homogeneous by microscopic visualization.
Total RNA from microdissected samples was isolated using the PicoPure RNA Kit (Arcturus Engineering, Santa Clara, CA), with an additional DNase I digestion step (Qiagen, Valencia, CA).
Immunohistochemistry
Sixteen paraffin embedded prostate samples were evaluated for HER3 expression by immunohistochemistry on a tissue microarray. The cases were represented in duplicated 1-mm diameter cores and always included normal prostatic glands adjacent to neoplastic foci in at least one of the cores. Three μM sections of the microarray were deparaffinized, rehydrated and subjected to antigen retrieval in a pressure cooker with citrate buffer at pH 6.0 for 5 min. Slides were cooled for 15 min, washed in water and incubated overnight at 4°C with anti-HER3 mouse monoclonal antibody (Upstate Biotechnology, Lake Placid, New York). Endogenous peroxidase was quenched and slides were incubated for 30 minutes with secondary antibody (Envision, DAKO, Gostrup, Denmark). Reactions were detected after development with diaminobencidine and H2O2 for 3 min. Slides were counterstained with Harri's hematoxilin, dehydrated and mounted. As a negative control, the primary antibody was substituted for isotype-matched mouse IgG.
Access to the program
The complete procedure has been coded in a Fortran-77 program, called FADA. Remote access to the program has been enabled by setting up a web-server where the program can be executed [74].
Authors' contributions
JJL and DA implemented the software and carried out the analysis; MS and RB performed the RT-PCR and Inmunohistochemistry experiments; PLF obtained the prostate samples and carried out the laser capture microdissections; TMT and ARO coordinated the work and wrote the manuscript.
Supplementary Material
Additional File 1
List of genes significantly associated to each cluster in the yeast dataset (q-value < 10-3).
Click here for file
Additional File 2
List of genes significantly associated to each cluster in the prostate cancer dataset (q-value < 10-3).
Click here for file
Acknowledgements
We thank I. Nayach for procurement and processing of prostate tissue samples and R. Muñoz for the help with the web server. JJL was partly supported by a NATO postdoctoral fellowship. This work has been facilitated by an institutional grant from Fundación Ramón Areces to the CBMSO, by grants GEN2001-4856-C13-10, GEN2001-4856-C13-07 and SAF2001-1969 from MCYT, and by grant PI020231 from FIS.
==== Refs
Jiang D Tang C Zhang A Cluster Analysis for Gene Expression Data: A Survey IEEE Transactions on Knowledge and Data Engineering 2004 16 1370 1386 10.1109/TKDE.2004.68
Xing EP Karp RM CLIFF: clustering of high-dimensional microarray data via iterative feature filtering using normalized cuts Bioinformatics 2001 17 Suppl 1 S306 15 11473022
Yoshida R Higuchi T Imoto S A Mixed Factors Model for Dimension Reduction and Extraction of a Group Structure in Gene Expression Data 2004
Johnson RA Wichern DW Applied Multivariate Statistical Analysis. 1992 Upper Saddle River, NJ, Prentice Hall
Sheng Q Moreau Y De Moor B Biclustering microarray data by Gibbs sampling Bioinformatics 2003 19 Suppl 2 II196 II205 14534190
Kluger Y Basri R Chang JT Gerstein M Spectral biclustering of microarray data: coclustering genes and conditions Genome Res 2003 13 703 716 12671006 10.1101/gr.648603
Tanay A Sharan R Shamir R Discovering statistically significant biclusters in gene expression data Bioinformatics 2002 18 Suppl 1 S136 44 12169541
Cheng Y Church GM Biclustering of expression data Proc Int Conf Intell Syst Mol Biol 2000 8 93 103 10977070
Reyment RJ Joreskog KG Applied Factor Analysis in the Natural Sciences 1996 Cambridge, Cambridge University Press
Bar-Joseph Z Gerber GK Lee TI Rinaldi NJ Yoo JY Robert F Gordon DB Fraenkel E Jaakkola TS Young RA Gifford DK Computational discovery of gene modules and regulatory networks Nat Biotechnol 2003 21 1337 1342 14555958 10.1038/nbt890
Hartigan JA Clustering algorithms. 1975 New York, NY, Wiley & Sons
Consortium GO The Gene Ontology (GO) database and informatics resource Nucl Acids Res 2004 32 D258 D261 14681407 10.1093/nar/gkh036
Gasch AP Spellman PT Kao CM Carmel-Harel O Eisen MB Storz G Botstein D Brown PO Genomic expression programs in the response of yeast cells to environmental changes Mol Biol Cell 2000 11 4241 4257 11102521
Welsh JB Sapinoso LM Sn AI Kern SG Wang-Rodriguez J Moskaluk CA Frierson HFJ Hampton GM Analysis of gene expression identifies candidate markers and pharmacological targets in prostate cancer. Cancer Res 2001 61 5974 5978 11507037
Causton HC Ren B Koh SS Harbison CT Kanin E Jennings EG Lee TI True HL Lander ES Young RA Remodeling of yeast genome expression in response to environmental changes Mol Biol Cell 2001 12 323 337 11179418
Chen D Toone WM Mata J Lyne R Burns G Kivinen K Brazma A Jones N Bahler J Global transcriptional responses of fission yeast to environmental stress Mol Biol Cell 2003 14 214 229 12529438 10.1091/mbc.E02-08-0499
Moskvina E Schuller C Maurer CT Mager WH Ruis H A search in the genome of Saccharomyces cerevisiae for genes regulated via stress response elements Yeast 1998 14 1041 1050 9730283 10.1002/(SICI)1097-0061(199808)14:11<1041::AID-YEA296>3.0.CO;2-4
Treger JM Schmitt AP Simon JR McEntee K Transcriptional factor mutations reveal regulatory complexities of heat shock and newly identified stress genes in Saccharomyces cerevisiae J Biol Chem 1998 273 26875 26879 9756934 10.1074/jbc.273.41.26875
Quinn J Findlay VJ Dawson K Millar JB Jones N Morgan BA Toone WM Distinct regulatory proteins control the graded transcriptional response to increasing H(2)O(2) levels in fission yeast Schizosaccharomyces pombe Mol Biol Cell 2002 13 805 816 11907263 10.1091/mbc.01-06-0288
Kuge S Jones N Nomoto A Regulation of yAP-1 nuclear localization in response to oxidative stress Embo J 1997 16 1710 1720 9130715 10.1093/emboj/16.7.1710
Delaunay A Pflieger D Barrault MB Vinh J Toledano MB A thiol peroxidase is an H2O2 receptor and redox-transducer in gene activation Cell 2002 111 471 481 12437921 10.1016/S0092-8674(02)01048-6
Delaunay A Isnard AD Toledano MB H2O2 sensing through oxidation of the Yap1 transcription factor Embo J 2000 19 5157 5166 11013218 10.1093/emboj/19.19.5157
Dumond H Danielou N Pinto M Bolotin-Fukuhara M A large-scale study of Yap1p-dependent genes in normal aerobic and H2O2-stress conditions: the role of Yap1p in cell proliferation control in yeast Mol Microbiol 2000 36 830 845 10844671 10.1046/j.1365-2958.2000.01845.x
Herskowitz I Life cycle of the budding yeast Saccharomyces cerevisiae Microbiol Rev 1988 52 536 553 3070323
Chu S DeRisi J Eisen M Mulholland J Botstein D Brown PO Herskowitz I The transcriptional program of sporulation in budding yeast Science 1998 282 699 705 9784122 10.1126/science.282.5389.699
Primig M Williams RM Winzeler EA Tevzadze GG Conway AR Hwang SY Davis RW Esposito RE The core meiotic transcriptome in budding yeasts Nat Genet 2000 26 415 423 11101837 10.1038/82539
Mata J Lyne R Burns G Bahler J The transcriptional program of meiosis and sporulation in fission yeast Nat Genet 2002 32 143 147 12161753 10.1038/ng951
Thomas G Hall MN TOR signalling and control of cell growth Curr Opin Cell Biol 1997 9 782 787 9425342 10.1016/S0955-0674(97)80078-6
Cherkasova VA Hinnebusch AG Translational control by TOR and TAP42 through dephosphorylation of eIF2alpha kinase GCN2 Genes Dev 2003 17 859 872 12654728 10.1101/gad.1069003
Kubota H Obata T Ota K Sasaki T Ito T Rapamycin-induced translational derepression of GCN4 mRNA involves a novel mechanism for activation of the eIF2 alpha kinase GCN2 J Biol Chem 2003 278 20457 20460 12676950 10.1074/jbc.C300133200
Natarajan K Meyer MR Jackson BM Slade D Roberts C Hinnebusch AG Marton MJ Transcriptional profiling shows that Gcn4p is a master regulator of gene expression during amino acid starvation in yeast Mol Cell Biol 2001 21 4347 4368 11390663 10.1128/MCB.21.13.4347-4368.2001
Dahlquist KD Salomanis N Vranizan K Lawlor SC Conkin BR GenMAPP, a new tool for viewing and analyzing microarray data on biological pathways Nature Gen 2002 31 19 20 10.1038/ng0502-19
Horoszewicz JS Leong SS Chu TM Wajsman ZL Friedman M Papsidero L Kim U Chai LS Kakati S Arya SK Sandberg AA The LNCaP cell line - a new model for studies on human prostatic carcinoma Prog Clin Biol Res 1980 37 115 132 7384082
Horoszewicz JS Leong SS Kawinski E Karr JP Rosenthal H Chu TM Mirand EA Murphy GP LNCaP model of human prostatic carcinoma. Cancer Res 1983 43 1809 1818 6831420
Baron A Migita T Tang D Loda M Fatty acid synthase: a metabolic oncogene in prostate cancer? J Cell Biochem 2004 91 47 53 14689581 10.1002/jcb.10708
DePrimo SE Diehn M Nelson JB Reiter RE Matese J Fero M Tibrishani R Brown PO Brooks JD Transcriptional programs activated by exposure of human prostate cancer cells to androgen Genome Biol 2002 3 research 0032.1 32.12 12184806 10.1186/gb-2002-3-7-research0032
Briggs MW Sacks DB IQGAP proteins are integral components of cytoskeletal regulation EMBO Rep 2003 4 571 574 12776176 10.1038/sj.embor.embor867
Foster CS Pathology of benign prostatic hyperplasia Prostate Suppl 2000 9 4 14 11056496 10.1002/1097-0045(2000)45:9+<4::AID-PROS3>3.0.CO;2-Q
Magee JA Araki T Patil S Ehrig T True L Humphrey PA Catalona WJ Watson MA Milbrandt J Expression profiling reveals hepsin overexpression in prostate cancer Cancer Res 2001 61 5692 5696 11479199
Luo J Duggan DJ Chen Y Sauvageot J Ewing CM Bittner ML Trent JM Isaacs WB Human prostate cancer and benign prostatic hyperplasia: molecular dissection by gene expression profiling Cancer Res 2001 61 4683 4688 11406537
Ernst T Hergenhahn M Kenzelmann M Cohen CD Bonrouhi M Weninger A Klaren R Grone EF Wiesel M Gudemann C Kuster J Schott W Staehler G Kretzler M Hollstein M Grone HJ Decrease and gain of gene expression are equally discriminatory markers for prostate carcinoma: a gene expression analysis on total and microdissected prostate tissue Am J Pathol 2002 160 2169 2180 12057920
Singh D Febbo PG Ross K Jackson DG Manola J Ladd C Tamayo P Renshaw AA D'Amico AV Richie JP Lander ES Loda M Kantoff PW Golub TR Sellers WR Gene expression correlates of clinical prostate cancer behavior Cancer Cell 2002 1 203 209 12086878 10.1016/S1535-6108(02)00030-2
LaTulippe E Satagopan J Smith A Scher H Scardino P Reuter V Gerald WL Comprehensive gene expression analysis of prostate cancer reveals distinct transcriptional programs associated with metastatic disease Cancer Res 2002 62 4499 4506 12154061
Stuart RO Wachsman W Berry CC Wang-Rodriguez J Wasserman L Klacansky I Masys D Argen K Goodison S McCelland M Wang Y Sawyers A Kalcheva I Tarin D Mercola D In silico dissection of cell-type-associated patterns of gene expression in prostate cancer. Proc Natl Acad Sci USA 2004 101 615 620 14722351 10.1073/pnas.2536479100
Rhodes DR Barrette TR Rubin MA Ghosh D Chinnaiyan AM Meta-analysis of microarrays: interstudy validation of gene expression profiles reveals pathway dysregulation in prostate cancer Cancer Res 2002 62 4427 4433 12154050
Kanehisa MGSKSOYHM Goto S Kawashima S Okuno Y Hattori M The KEGG resource for dephiphering the genome Nucl Acids Res 2004 32 D277 D280 14681412 10.1093/nar/gkh063
Brawer MK Prostate-specific antigen Semin Surg Oncol 2000 18 3 9 10617891 10.1002/(SICI)1098-2388(200001/02)18:1<3::AID-SSU2>3.0.CO;2-I
Schlessinger J Ligand-induced, receptor-mediated dimerization and activation of EGF receptor. Cell 2002 110 669 672 12297041 10.1016/S0092-8674(02)00966-2
Menard S Tagliabue E Campiglio M Pupa SM Role of HER2 gene overexpression in breast carcinoma J Cell Physiol 2000 182 150 162 10623878 10.1002/(SICI)1097-4652(200002)182:2<150::AID-JCP3>3.0.CO;2-E
Agus DB Akita RW Fox WD Lofgren JA Higgins B Maiese K Scher HI Sliwkowski MX A potential role for activated HER-2 in prostate cancer Semin Oncol 2000 27 76 83; discussion 92-100 11236032
Myers RB Srivastava S Oelschlager DK Grizzle WE Expression of p160erbB-3 and p185erbB-2 in prostatic intraepithelial neoplasia and prostatic adenocarcinoma J Natl Cancer Inst 1994 86 1140 1145 7913137
Lyne JC Melhem MF Finley GG Wen D Liu N Deng DH Salup R Tissue expression of neu differentiation factor/heregulin and its receptor complex in prostate cancer and its biologic effects on prostate cancer cells in vitro Cancer J Sci Am 1997 3 21 30 9072304
Vakar-Lopez F Cheng CJ Kim J Shi GG Troncoso P Tu SM Yu-Lee LY Lin SH Up-regulation of MDA-BF-1, a secreted isoform of ErbB3, in metastatic prostate cancer cells and activated osteoblasts in bone marrow J Pathol 2004 203 688 695 15141384 10.1002/path.1568
Li S Wang Q Chakladar A Bronson RT Bernards A Gastric hyperplasia in mice lacking the putative Cdc42 effector IQGAP1 Mol Cell Biol 2000 20 697 701 10611248 10.1128/MCB.20.2.697-701.2000
Van Aelst L Joneson T Bar-Sagi D Identification of a novel Rac1-interacting protein involved in membrane ruffling Embo J 1996 15 3778 3786 8670882
Guy PM Platko JV Cantley LC Cerione RA Carraway KL Insect cell-expressed p180erbB3 possesses an impaired tyrosine kinase activity Proc Natl Acad Sci U S A 1994 91 8132 8136 8058768
Daly JM Jannot CB Beerli RR Graus-Porta D Maurer FG Hynes NE Neu differentiation factor induces ErbB2 down-regulation and apoptosis of ErbB2-overexpressing breast tumor cells Cancer Res 1997 57 3804 3811 9288791
Naidu R Yadav M Nair S Kutty MK Expression of c-erbB3 protein in primary breast carcinomas Br J Cancer 1998 78 1385 1390 9823984
Holbro T Beerli RR Maurer F Koziczak M Barbas CF Hynes NE The ErbB2/ErbB3 heterodimer functions as an oncogenic unit: ErbB2 requires ErbB3 to drive breast tumor cell proliferation Proc Natl Acad Sci U S A 2003 100 8933 8938 12853564 10.1073/pnas.1537685100
Yarden Y Slikowski MX Untangling the ErbB signalling network. Nature Rev Mol Cell Biol 2001 2 127 137 11252954 10.1038/35052073
Mellinghoff IK Vivanco I Kwon A Tran C Wongvipat J Sawyers CL HER2/neu kinase-dependent modulation of androgen receptor function through effects on DNA binding and stability. Cancer Cell 2004 6 517 527 15542435 10.1016/j.ccr.2004.09.031
Feldman BJ Feldman D The development of androgen-independent prostate cancer Nat Rev Cancer 2001 1 34 45 11900250 10.1038/35094009
Ramaswamy S Ross KN Lander ES Golub TR A molecular signature of metastasis in primary solid tumors. Nat Genet 2003 33 49 54 12469122 10.1038/ng1060
Wigelt B Glas AM Wessels LFA Witteveen AT Peterse JL van't Veer LJ Gene expression profiles of primary breast tumors maintained in distant metastases Proc Natl Acad Sci USA 2003 100 15901 15905 14665696 10.1073/pnas.2634067100
Signoretti S Montironi R Manola J Altimari A Tam C Bubley G Balk S Thomas G Kaplan I Hlatky L Hahnfeldt P Kantoff P Loda M Her-2-neu expression and progression toward androgen independence in human prostate cancer J Natl Cancer Inst 2000 92 1918 1925 11106683 10.1093/jnci/92.23.1918
Osman I Scher HI Drobnjak M Verbel D Morris M Agus D Ross JS Cordon-Cardo C HER-2/neu (p185neu) protein expression in the natural or treated history of prostate cancer Clin Cancer Res 2001 7 2643 2647 11555574
Savinainen KJ Saramaki OR Linja MJ Bratt O Tammela TL Isola JJ Visakorpi T Expression and gene copy number analysis of ERBB2 oncogene in prostate cancer Am J Pathol 2002 160 339 345 11786427
Lara PNJ Meyers FJ Gray CR Edwards RG Gumerlock PH Kauderer C Tichauer G Twardowski P Doroshow JH Gandara DR HER-2/neu is overexpressed infrequently in patients with prostate carcinoma. Results from the California Cancer Consortium Screening Trial Cancer 2002 94 2584 2589 12173324 10.1002/cncr.10526
Bo TH Dysvik B Jonassen I LSimpute: accurate estimation of missing values in microarray data with least squares methods Nucleic Acids Res 2004 32 e34 14978222 10.1093/nar/gnh026
Durbin R Eddy S Krogh A Mitchison G Biological sequence analysis. Probabilistics models of proteins and nucleic acids. 1998 Cambridge, UK, Cambridge University Press
Felsenstein J Inferring phylogenies from protein sequences by parsimony, distance, and likelihood methods. Methods Enzymol 1996 266 418 427 8743697
Storey JD Tibshirani R Statistical significance for genomewide studies Proc Natl Acad Sci U S A 2003 100 9440–9445 12883005 10.1073/pnas.1530509100
|
16107210
|
PMC1239914
|
CC BY
|
2021-01-04 16:32:48
|
no
|
BMC Genomics. 2005 Aug 17; 6:109
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-109
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-1221616227710.1186/1471-2164-6-122Research ArticleLaterally transferred elements and high pressure adaptation in Photobacterium profundum strains Campanaro Stefano [email protected] Alessandro [email protected] Nicola [email protected] Federico M [email protected]'Angelo Michela [email protected] Francesca [email protected] Alessandro [email protected] Giorgio [email protected] Giulio [email protected] Giorgio [email protected] Douglas H [email protected] CRIBI Biotechnology Centre and Dept. of Biology, University of Padova, Via U. Bassi 58/B, 35131 Padova, Italy2 Scripps Institution of Oceanography, University of California San Diego, La Jolla CA, 92093-0202, USA3 Department of Histology, Microbiology and Medical Biotechnology, University of Padova, Via A. Gabelli 63, 35121 Padova, Italy2005 14 9 2005 6 122 122 14 6 2005 14 9 2005 Copyright © 2005 Campanaro et al; licensee BioMed Central Ltd.2005Campanaro et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Oceans cover approximately 70% of the Earth's surface with an average depth of 3800 m and a pressure of 38 MPa, thus a large part of the biosphere is occupied by high pressure environments. Piezophilic (pressure-loving) organisms are adapted to deep-sea life and grow optimally at pressures higher than 0.1 MPa. To better understand high pressure adaptation from a genomic point of view three different Photobacterium profundum strains were compared. Using the sequenced piezophile P. profundum strain SS9 as a reference, microarray technology was used to identify the genomic regions missing in two other strains: a pressure adapted strain (named DSJ4) and a pressure-sensitive strain (named 3TCK). Finally, the transcriptome of SS9 grown under different pressure (28 MPa; 45 MPa) and temperature (4°C; 16°C) conditions was analyzed taking into consideration the differentially expressed genes belonging to the flexible gene pool.
Results
These studies indicated the presence of a large flexible gene pool in SS9 characterized by various horizontally acquired elements. This was verified by extensive analysis of GC content, codon usage and genomic signature of the SS9 genome. 171 open reading frames (ORFs) were found to be specifically absent or highly divergent in the piezosensitive strain, but present in the two piezophilic strains. Among these genes, six were found to also be up-regulated by high pressure.
Conclusion
These data provide information on horizontal gene flow in the deep sea, provide additional details of P. profundum genome expression patterns and suggest genes which could perform critical functions for abyssal survival, including perhaps high pressure growth.
==== Body
Background
Piezophilic microbes have been isolated from a variety of abyssal and hadal deep-sea environments and include several psychrophilic or psychrotolerant proteobacteria, and several high temperature Euryarchaeota and Crenarchaeota [1]. While the study of these extremophiles is still in its infancy, both physiological and structural adaptations appear to be important for high-pressure life.
One moderately piezophilic, gamma-proteobacterial isolate, Photobacterium profundum strain SS9, has been the subject of a number of studies addressing the nature and regulation of genes important for pressure-sensing and high pressure adaptation, owing to the relative ease of its cultivation as well as its genetic tractability [1]. Here we make use of another important P. profundum feature, namely the availability of multiple closely related strains which differ in their pressure and temperature optima. Strain SS9 was isolated from an amphipod in the Sulu Trough at a depth of 2551 m and displays optimum growth at 28 MPa and 15°C [2]. P. profundum strain DSJ4 was recovered from a sediment sample obtained from the Ryukyu Trench (Japan) at a depth of 5110 m and displays its optimum growth at 10 MPa (with little change in growth at pressures up to 50 MPa) and a temperature optimum of 10°C [3]. P. profundum strain 3TCK was isolated from a shallow sediment sample obtained from San Diego Bay (California, U. S. A.) and exhibits optimal growth at atmospheric pressure and a broad temperature span for growth from below 0°C to above 20°C.
Recently, the complete genome sequence of strain SS9 was obtained [4]. This achievement has enabled the scaling up of the study of piezophily and, more generally, of adaptations to the deep sea (i.e., low temperature, low nutrient input, absence of sunlight), at the genomic level. In this study a microarray covering nearly the complete SS9 genome was used to investigate both the flexible gene pool (genes whose presence is variable due to insertion/deletion events) and high pressure adaptation by means of two different post-genomic approaches:
1-Using the SS9 genome as a reference, comparative genomic hybridization experiments were performed with DNA extracted from the other two P. profundum strains (DSJ4 and 3TCK) to identify the flexible gene pool in SS9. To determine if these genes were obtained from lateral gene transfer events or, conversely, from genomic reduction events in the other strains, their GC content, codon usage and genomic signature was analyzed.
2-Transcriptome analyses were performed as a function of pressure (0.1, 28 and 45 MPa at 16°C) and temperature (4°C vs. 16°C at 0.1 MPa). Although we have recently presented preliminary data on SS9 expression at 0.1 and 28 MPa, in this study temperature effects on gene regulation were compared with pressure effects since increasing pressure exerts some common effects with decreasing temperature in terms of membrane microviscosity and with increasing temperature in terms of protein stability [5]. Moreover the transcriptional changes identified in the 0.1 MPa vs. 28 MPa and 28 MPa vs. 45 MPa experiments were compared in order to reveal expression changes in a piezophilic bacterial species grown at supra-optimal pressure.
Finally, the results obtained from comparative genomic analyses and expression profiling experiments were combined to identify genes shared among the P. profundum piezophiles, absent from the piezosensitive strain, and up-regulated at high pressure. This allowed a few genes to be selected from a pool of approximately 6,000 genes whose distribution and expression characteristics suggest possible function in high pressure adaptation and thus present themselves as candidates for future genetic investigation.
Results
Comparison of three P. profundum strains
Amplification and analysis of the 16S rDNA from strain 3TCK revealed that it is 97.7% identical at the 16S level with Photobacterium profundum SS9 and 98.7% identical to P. profundum DSJ4, suggesting that they are all members of the same species. Figure 1 shows a 16S rRNA-based phylogenetic tree demonstrating the relationship among selected Photobacterium species, including the three P. profundum strains selected for this study. Growth curves showed that 3TCK is psychrotolerant and piezo-sensitive. However it grows at higher temperatures than P. profundum SS9 and DSJ4 (data not shown) and has faster growth rates at 0.1 MPa.
Figure 1 Phylogenetic tree showing the relationship of 16S rRNA gene sequence within the Photobacterium genus using the neighbor joining method. The scale represents the average number of nucleotide substitutions per site. Bootstrap values are from 1,000 replicates and shown for frequencies above the threshold of 50%. The phylogenetic tree was created using E. coli and V. cholerae as outgroups.
Genomic comparison between different P. profundum strains
The first question that arises from the genomic comparison between SS9 and the other two strains is: "how many SS9 genes are missing or highly divergent in the 3TCK and DSJ4 genomes?". 544 ORFs were determined to be absent in 3TCK strain genome, 313 (9.1% of the ORFs located on chr1) belong to the SS9 chr1 and 231 (11.5% of the ORFs located on chr2) to chr2. 562 ORFs are absent in the DSJ4 genome, 292 (8.5%) are located on SS9 chr1 and 270 (13.5%) on chr2.
An interesting aspect of these data is that for both strains the percentage of missing/divergent regions is higher on chr2 than chr1. This indicates that chr2 (Figure 2) contains a proportionally larger flexible gene pool and that it has been the target of more gene transfer events for its size (~2.2 Mbp) than chr1 (~4.1 Mbp) that contains the most "established" genes. This is also true for other Vibrionaceae genomes [6].
Figure 2 Genomic organization of the three P. profundum strains compared with expression level and differentially expressed genes obtained from microarray experiments. Form the outside inward circles represent: 1, 2) predicted protein-coding ORFs on the plus and minus strands of SS9 genome (colours were assigned according to the colour code of the COG functional classes); 3) transposon-related proteins (black bars) and genes showing similarity with phage proteins (green bars); differentially expressed genes in 4) 0.1 MPa vs. 28 MPa microarray experiment (green and red bars represent genes up-regulated respectively at 0.1 MPa and 28 MPa); 5) 45 MPa vs. 28 MPa (green and red bars represent genes up-regulated respectively at 28 MPa and 45 MPa); 6) 4°C vs. 16°C (green and red bars represent genes up-regulated respectively at 16°C and 4°C); 7) genomic regions absent (red bars) or duplicate (blue bars) in 3 TCK strain (compared to SS9 strain) and 8) in DSJ4 strain (compared to SS9); 9) GC content variation; 10) 10 kbp windows genomic signature compared to total genomic signature; 11) absolute expression level at 28 MPa obtained from microarray experiments.
In order to define if the regions absent in the 3TCK-DSJ4 strains could be considered to have been acquired by horizontal gene transfer we performed three different analyses: GC content variation (Figure 2, 9th circle), tetranucleotide composition (genomic signature) (Figure 2, 10th circle) [7,8] and codon bias relative to the average gene versus S3 percentage (G+C content of codon site 3) (Figure 3 and Additional file 1) [9].
Figure 3 Evidence for lateral gene transfer in SS9. Each P. profundum gene of ≥ 200 codons is represented in this graph by a point with co-ordinates corresponding to its codon bias relative to average gene and G+C frequency of codon site three. Genes having G+C frequency lower than 0.28 and codon bias higher than 0.35 correspond to the left horn. Genes having G+C frequency higher than 0.46 and codon bias higher than 0.35 correspond to the right horn. Genes absent in 3TCK and/or DSJ4 genomes (highlighted in black) are more frequent in horn regions (64% in left horn, 66% in right horn) respect to the middle region (10% of ORFs). The low number of strains analyzed leads to an underestimation of the laterally transferred regions in SS9 and this probably accounts for those genes localized in left and right horn which are present in all three P. profundum strains analyzed.
Taken together these three different analyses were able to identify a large number of potentially laterally transferred regions. For example, a region named Chr1.11 (named for its chromosome location and clockwise order of position, Figure 2) has an altered tetranucleotide composition but its GC content is similar to the surrounding regions and it has a "normal" codon bias relative to the average gene versus S3 percentage (data not shown). Conversely, region Chr2.1 is only characterized by a slight GC content variation. The results obtained from these analyses are discussed in-depth below.
A BLASTP similarity search of SS9 proteins identified various phage-related proteins, mostly encoded in three regions named Chr1.8, Chr2.3 and Chr2.5. Microarray data obtained by comparing the SS9, 3TCK and DSJ4 genomes (Figure 2) confirmed that these genomic portions are absent in both the 3TCK and DSJ4. These regions present characteristics typical of a genomic island (GI): (1) GC content anomalies, (2) altered codon bias, (3) insertion at the 3'-end of a tRNA gene (tRNA-N) and (4) presence of a gene encoding an integrase at one end (Table 1) [10].
Table 1 General characteristics of putative horizontally transferred elements found in the genome of P. profundum SS9 strain.
Element Absent in strain Start-end position Size (kbp) GC content percent (chr1_42%; chr2_41.2%) tRNA Integrase-like genes at the end Expression level Relevant resident genes
Plasmid 3TCK DSJ4 1–80,033 80 44% NO NO moderate (4,458)
Chr1.1-LF 3TCK 13,500–51,360 38 45% NO NO low (1,079) LF-lateral flagella coding region; there are two genes up-regulated at 0.1 MPa at both ends (PBPRA0013- PBPRA0050).
Chr1.2 3TCK DSJ4 479,200–490,850 11.6 44% NO NO low (2,623) Xylose transport and metabolism; one gene up-regulated at 4°C (PBPRA0466).
Chr1.3 3TCK 717,590–751,055 33.4 41% NO NO moderate (3,879)
Chr1.4 3TCK DSJ4 957,470–983,750 26.3 40% YES YES (PBPRA0877) moderate (3,744)
Chr1.5 3TCK DSJ4 1,081,250–1,093,850 12.6 41% NO NO low (1,094)
Chr1.6 DSJ4 1,262,100–1,281,040 21.7 41% NO NO low (1,111) One up-regulated gene at 0.1 MPa near 5'-end (PBPRA1136).
Chr1.7 DSJ4 1,436,830–1,453,450 16.6 42% NO NO low (2,432) ORFs involved in amino acid metabolism and transport.
Chr1.8 3TCK DSJ4 1,482,130–1,528,980 46.8 40% NO YES (PBPRA1336) moderate (3,453) 12 phage-related proteins and genes involved in tryptophan transport and metabolism.
Chr1.9 3TCK DSJ4 (smaller) 1,803,270–1,847,340; 1,803,270–1,837,500 44.1 3TCK; 34.2 DSJ4 40% 41% NO NO moderate (4,586) One gene up-regulated at 28 MPa (vs. 0.1 MPa) (PBPRA1598) and numerous transposases.
Chr1.10 3TCK-DSJ4 1,887,600–1,921,860 34.2 42.9% YES NO low (2,509) Genes involved in fatty acid synthesis and tryptophan biosynthesis.
Chr1.11 3TCK-DSJ4 2,063,315–2,097,000 33.7 40% YES NO moderate (3,276) One gene (PBPRA1810) up-regulated at 16°C.
Chr1.12 DSJ4 2,399,560–2,410,040 10.5 DSJ4 40% YES NO moderate (4,179) Three genes differentially expressed (PBPRA2084, PBPRA2086, PBPRA2087).
Chr1.13 3TCK (smaller) DSJ4 2,640,200–2,650,840 10.6 3TCK 17.7 DSJ4; 41% YES NO moderate (12,480-7,436) Gene cluster involved in tricarboxylic acid fermentation and transport; some genes were up-regulated at 16°C and/or 0.1 MPa.
Chr1.14 DSJ4 2,892,510–2,901,270 8.8 45.9% NO NO high (11,847) Genes involved in pilus assembly, three genes up-regulated at 28 MPa (vs. 0.1 MPa) and/or down-regulated at 45 MPa (PBPRA2498, PBPRA2499, PBPRA2505).
Chr1.15 3TCK-DSJ4 3,104,770–3,110,020; 3,145,100-3,145,095 5.2+ 29.3 29%–37% YES NO moderate (9,965) flm genes; two genes down-regulated at 4°C (PBPRA2692, PBPRA2701).
Chr1.16 3TCK-DSJ4 3,220,530–3,230,610 10 43.9% NO NO moderate (4,394) Genes coding for a phosphotransferase system cellobiose-specific; one gene up-regulated at 4°C (PBPRA2779).
Chr1.17 3TCK-DSJ4 3,706,515–3,712,920 6.4 45.3% NO NO moderate (6,448)
Chr2.1 3TCK-DSJ4 159,600–184,460 24.9 43.4% NO NO low (2,659) Genes involved in pentose and glucuronate interconversions and PTS system.
Chr2.2 3TCK-DSJ4 342,190–349,500 7.3 43.2% NO NO low (1,741) Very small region, no orthologous in other Vibrio species
Chr2.3 3TCK-DSJ4 637,160–679,550 42.4 41.1% NO YES (PBPRB0551) low (2,485) 12 phage-related proteins; similarities with Gifsy -1, Gifsy -2 prophages protein.
Chr2.4 DSJ4 1,179,200–1,191,215 12 42.8% NO NO low (1,191) Phosphotransferase system (PTS) N-acetylgalactosamine-specific genes.
Chr2.5 3TCK (smaller) DSJ4 1,419,870–1,427,600; 1,419,870–1,439,670 19.8 46.2% NO YES (PBPRB1271) low (1,737) 12 phage-related proteins.
Chr2.6 3TCK-DSJ4 1,661,040–1,685,300 24.3 39.2% YES NO moderate (3,201) Four transposases in the central part; genes involved in pentose and glucuronate interconversions; genes involved in C4-dicarboxylate transport system
Chr2.7-PI 3TCK DSJ4 1,745,750–1,814,900 69.1 44.8% NO YES (PBPRB1675) low (2,060) This region could be derived from a plasmid integration.
Chr2.8 3TCK-DSJ4 1,872,240–1,891,940 19.7 40.4% NO NO moderate (4,347) Genes involved in C4-dicarboxylate transport system and in pentose and glucuronate interconversion pathway.
Chr2.9 DSJ4 1,901,330–1,928,060 26.7 41.9% YES? NO moderate (4,074) Highly discontinuous region.
Chr2.10 3TCK-DSJ4 2,013,000–2,087,900; 1,975,200–2,095,800 ~108 3TCK; ~145 DSJ4 41.2%–42.7% This region is probably duplicated in the test strains (DSJ4 and 3TCK.
Low, moderate and high expression values are calculated on the basis of the microarray clones fluorescence intensity and correspond respectively to values lower than 3,000, comprise between 3,000 and 10,000 and higher than 10,000 arbitrary fluorescence units. Negative controls clones have a mean fluorescence value of 377 arbitrary fluorescence units. As an example, clones localized in the ATP synthase operon have a mean fluorescence value of 36,100 arbitrary fluorescence units and clones localized in polar flagellum cluster have a mean fluorescence value of 10,200. For region Chr2.10 mean fluorescence value are not reported because it is very large and a mean fluorescence value has little significance.
Chr1.8 (46.8 kbp) has an altered tetranucleotide composition and presents twelve phage-related proteins, one of these (PBPRA1336) encodes a putative integrase protein. Nevertheless this region lacks the other two characteristics of a typical GI: GC content anomalies and the presence of a tRNA gene at one end. This region also contains two genes, encoding tryptophanase (TnaA) (PBPRA1344) and a hypothetical tryptophan-specific transport protein (PBPRA1345), involved in tryptophan transport and metabolism, and 27 hypothetical or conserved hypothetical proteins.
The region Chr2.3 is located on chr2, spans approximately 42.4 kbp, has no GC content anomalies (41.1%) but has a slightly altered tetranucleotide composition. Chr2.3 contains twelve ORFs that have similarity with phage proteins, one of these being a hypothetical integrase (PBPRB0551). Moreover it encodes a putative NAD(P)H oxidoreductase (PBPRB0548) and a putative TrkA family protein (glutathione-regulated potassium efflux system protein) (PBPRB0550). Furthermore PBPRB0559 gene has similarities with enterohemolysin 1, a gene also present in the Gifsy-1 prophage, and PBPRB0560 has similarities with exodeoxyribonuclease of the Gifsy-2 prophage of Salmonella typhimurium LT2 [11].
Chr2.5 region appears to be completely absent in DSJ4 whereas only the first part is absent in 3TCK. Chr2.5 contains a hypothetical integrase gene (PBPRB1271), twelve phage related proteins and various hypothetical proteins. The higher GC content (46.2%) of Chr2.5 suggests that it has been acquired more recently than Chr1.8 and Chr2.3.
A large part of the genes located in the Chr1.8, Chr2.3 and Chr2.5 regions clearly lacks orthologous genes in others bacteria [12] such as V. cholerae, V. vulnificus (strains CMCP6 and YJ016), V. parahaemolyticus, V. fisheri, E. coli, and B. subtilis. The high number of hypothetical proteins encoded in these regions suggests that these loci could have been acquired from bacteria still unknown. Consistent with its recent acquisition Chr2.5 presents an altered codon bias and a large percentage of its ORFs are located on the right horn of a graph of the codon bias versus the GC content frequency in third position (Additional file 1).
A large 69.1 kbp element (Chr2.7-PI for Plasmid Integration) is present only in strain SS9 and seems to be the result of plasmid integration into the chromosome. This region has a high GC content and altered codon bias and genomic signature. Various bacterial conjugation factors (TrbCDEJBFGI and TraFLGIKL) are present in this element. These genes are typically found in widespread conjugationally transmitted plasmids [13]. This element carries a large number of genes, but of particular interest is a multidrug efflux system (PBPRB1635, PBPRB1637, PBPRB1638).
SS9 also contains a 80 kbp plasmid (named "plasmid" in Table 1) that shows similarity with V. vulnificus plasmid YJ016 at least in the region spanning the genes related to conjugation. This plasmid is absent in both DSJ4 and 3TCK and presents characteristics of horizontally transferred DNA, in fact various ORFs belonging to this element are localized to the right horn of Figure 3 (see also Additional file 1).
PCR examination for the presence of this plasmid in various laboratory derivatives of P. profundum SS9 has revealed that it can be lost: strain TW30 [14] is a toxR- derivative of DB110 [15] which lacks the plasmid (Additional file 2), yet TW30 exhibits no pressure or temperature growth defects. This raises the question of the function of the genes located on the plasmid which must be playing some role in environmental adaptation for the plasmid to be retained [16].
One of the most interesting results obtained from the annotation process was the finding that in the SS9 genome there are two flagellar clusters. One of these (tentatively identified as the polar flagella cluster, PF) is located on chr1 between 993676 and 1046789 bp and the second cluster (tentatively identified as the lateral flagella cluster, Chr1.1-LF) is localized on chr1 near the origin between 13496 and 49254 bp. Similarity searches revealed that the PF region contains most of the genes involved in polar flagellum assembly, with a gene organization typical of a Vibrionaceae polar flagellar cluster [17]. Chr1.1-LF contains all the genes involved in lateral flagellar synthesis and most of the genes localized in this region have similarity with the flagellar cluster of V. parahaemolyticus that codes for lateral flagella [18]. This region is absent only in 3TCK strain (Figure 2).
Finally, the data on variable regions was compared with the transcriptome results to discern if any of the horizontally acquired genes might be indicated to perform a role in pressure or temperature adaptation. The absolute fluorescence value from microarray analysis indicates that the expression levels of most of these genes was quite low. Indeed, only three regions in chr1 (Chr1.13, Chr1.14, Chr1.15) and two in chr2 (Chr2.8, Chr2.9) had fluorescence values higher than the mean fluorescence level of the entire chromosome (Table 1 and Figure 2, 11th circle).
Region Chr1.12 is absent in DSJ4 (and in 3TCK this region is smaller) and has three genes, PBPRA2087 (hypothetical protein), PBPRA2084 (putatively evolved beta-D-galactosidase, alpha subunit) and PBPRA2086 (putative oxidoreductase), differentially expressed in pressure experiments.
Region Chr1.13 is absent in both strains and contains a large gene cluster involved in the tricarboxylic acid fermentation and in the cleavage of citrate to oxaloacetate and acetate. This region contains six genes up-regulated at 16°C and/or 0.1 MPa (PBPRA2289, PBPRA2292, PBPRA2295, PBPRA2298, PBPRA2300, PBPRA2303) (Additional file 3).
Region Chr1.14 is absent in DSJ4 and contains genes involved in pilus assembly, some of these are up-regulated at 28 MPa (vs. 0.1 MPa) and down-regulated at 45 MPa (PBPRA2498, PBPRA2499, PBPRA2505).
Finally, region Chr1.15 is lacking in both of the comparison strains and contains genes having a high expression level at 28 MPa, some of which are differentially expressed at 28 MPa and 4°C (PBPRA2692, PBPRA2701, PBPRA2710). This region contains flm genes that in other bacteria are involved in LPS O-Ag biosynthesis and flagellar filament assembly [19]. Interestingly changes in LPS O-antigen structure have been observed in Yersinia pestis KM218 grown at low temperatures [20]. This element has very low GC content (29%–37%), altered genomic signature and most part of its genes are localized to the left horn of the graph reported in supporting online material (Table 1, Figure 3 and Additional file 1), thus supporting the idea that it could have been laterally acquired.
It is curious that some of the variable regions differentially expressed in pressure experiments are lacking or are highly divergent in both 3TCK and DSJ4. This indicates that although genes located in these regions could be involved in the high pressure response of SS9, they are not essential to it and other P. profundum strains will achieve piezophily with different strategies. Moreover, genes differentially expressed at 28 MPa or 45 MPa, but present in both DSJ4 and 3TCK, could be beneficial but not sufficient for high pressure adaptation.
Considering only pressure regulated genes belonging to the group absent in 3TCK and/or DSJ4, 29 genes are absent in both strains but only 9 genes are absent in 3TCK strain alone (Additional file 3 and Additional file 4). These data were obtained using a profile search with JExpress software [21]. Of these 9 genes, 6 are up-regulated at 28 MPa (vs. 0.1 MPa) and/or 45 MPa (PBPRB0026 hypothetical sensor protein TorS; PBPRA0776 hypothetical protein; PBPRA1912 hypothetical protein; PBPRA2251 hypothetical ABC transporter, periplasmic solute-binding protein, family 5; PBPRA2252 hypothetical ABC transporter, permease protein; PBPRA2573 putative long-chain fatty acid transport protein).
In the SS9 genome we found two genes for TorS proteins (PBPRA1232 and PBPRB0026). Only one of these (named PBPRB0026) is differentially expressed at 28 MPa and this gene is also absent in 3TCK strain. TorS is able to regulate various genes in response to trimethylamine N-oxide (TMAO) [22], in particular it regulates TMAO reductase (PBPRA1467) that is also up-regulated at 28 MPa. It is conceivable that trimethylamine reduction increases the pH of the cytoplasm and, for this reason, other genes identified with microarray experiments, such as tryptophanase (PBPRB0382 and PBPRA2532) increase their expression in order to counter this alcalinization. Alternatively, since no TMAO was added to the SS9 cultures used for the microarray experiments, the second TorS could also be responding to an as yet undiscovered signal.
In addition to these six genes, others could perform an important role in high pressure adaptation. We expect that a high number of genes and proteins are regulated at the post-transcriptional level and have an important role in high pressure adaptation but these studies are beyond the scope of this paper. Moreover protein structural adaptations, that were not considered in this analysis, could also have great importance for SS9 piezophily.
Transcriptome analysis of P. profundum strain SS9 under different pressure and temperature conditions
In a previous paper a preliminary analysis was presented for genes differentially expressed at 28 MPa (the optimum growth pressure for SS9) versus 0.1 MPa, highlighting a stress response at low pressure, an heavy involvement of membrane transporters in pressure adaptation and an increased expression at 28 MPa of genes involved in the Stickland reaction and TMAO reduction. Nevertheless, additional transcriptional responses remain to be elucidated including the response to low temperatures and the effect of supraoptimal pressures (45 MPa) on SS9. To better elucidate these points expression profiling was performed at different temperatures (4°C vs. 16°C) and different pressures (28 MPa vs. 45 MPa). To clarify the role of specific biological processes on temperature and pressure adaptation, we also performed a Gene Ontology analysis [23] on differentially expressed genes using specific software such as GoMiner [24] and FatiGO [25].
Analysis of co-regulated genes in pressure and temperature experiments
The rationale for temperature experiments was the comprehension of which genes are co-regulated during pressure and temperature changes. It is known that low temperature and high pressure have similar effects on some biological structures (for example membranes) [5]. We found 36 (43 considering also those that are co-regulated between 45 MPa and temperature) out of 319 genes that share a similar expression pattern between temperature (16°C vs. 4°C) and pressure (0.1 MPa vs. 28 MPa) experiments, a number higher that expected only by chance. In fact, the number of up-regulated genes at 28 MPa and at 4°C is respectively 101 and 75 over 4752 (the total number of genes covered by our array). If the two conditions are independent, the expected number of genes that are up-regulated both at high pressure and low temperature could be estimated as (101/4752)*(75/4752)*4752 = 1.6. Similarly we could estimate the number of down-regulated genes at 28 MPa and 4°C as (108/4752)*(85/4752)*4752 = 1.9. So the total number of genes expected to be up- or down-regulated in both the conditions considered (3.5) is approximately 1/10th of the differentially expressed genes observed. Therefore microarray experiments corroborate the hypothesis that high pressure and low temperature have a overlapping effects on gene expression.
The SS9 genome sequence reveals the presence of two iron transporters: one (PBPRB0373-PBPRB0375) that is up-regulated at 28 MPa (vs. 0.1 MPa) and at 4°C, while the other one (PBPRA3182, PBPRA3183, PBPRA3184) is up-regulated at 0.1 MPa (vs. 0.1 MPa) and at 16°C. Iron accumulation in organisms that live in the ocean environment is difficult [26] and the evolution of two alternative transporters could be important in order to survive under different physical conditions.
The presence of different isoforms of the same transporter that work at different pressure and temperature conditions is not limited to iron transporters. ORFs PBPRA0098-PBPRA0101 code for a hypothetical oligopeptide transporter and are up-regulated at 0.1 MPa, while ORFs PBPRA2251-PBPRA2254 code for a different oligopeptide transporter that is up-regulated at 28 MPa (vs. 0.1 MPa) and 45 MPa (vs. 28 MPa). Other oligopeptide transporters seem to have the same behaviour such as those codified by PBPRA0521-PBPRA0525 and PBPRA2934-PBPR2938, the first being up-regulated at 28 MPa (vs. 0.1 MPa) and 4°C, while the second is weakly up-regulated at 0.1 MPa.
There is an entire region (Table 1; region Chr1.13) containing a large number of genes that are up-regulated both at 0.1 MPa and at 16°C. This region was also identified in the above genome comparisons because it is absent in both 3TCK and DSJ4 strains.
Analysis of co-regulated genes in 28 MPa and 45 MPa experiments
The expression profile of SS9 at 28 MPa (the optimal growth pressure) was also compared with that at 45 MPa.
Interestingly the 45 MPa vs. 28 MPa expression profile comparison revealed only 68 differentially expressed genes (33 up-regulated and 35 down-regulated), in contrast to the high number of differentially expressed genes between 28 MPa and 0.1 MPa (101 up-regulated and 108 down-regulated). Of these 68 differentially expressed genes, only 31 were specific for very high pressure adaptation, the remaining 37 also being expressed under other environmental conditions tested. This result indicates that SS9 undergoes a heavy reorganization in gene expression between atmospheric pressure and 28 MPa, while this is not seen moving from 28 MPa to 45 MPa.
A Gene Ontology search with GoMiner software [24] indicated that among the 31 genes specific for very high pressure adaptation there is an enrichment of genes involved in arginine metabolism (GO: 0006525), catabolism (GO: 0006527) (Additional file 5) and transport (PBPRA2073-PBPRA2076). Experiments at 45 MPa were also useful in identifying genes whose expression follows the direction of pressure variation, being up-regulated or down-regulated both at 28 MPa (compared to 0.1 MPa) and at 45 MPa (compared to 28 MPa). Twenty one genes matched this expression profile (Additional file 3).
One of these genes encodes a putative delta-9 fatty acid desaturase (PBPRB0742). High pressure increases the rigidity of membranes [27,28], and for this reason genes such as the putative delta-9 fatty acid desaturase presumably were up-regulated in order to increase the membrane unsaturation and thus membrane fluidity. Despite the fact that much is known about membrane modification in response to pressure variation in SS9 [29,30], these experiments reveal the possible involvement of a previously unrecognized gene in fatty acid unsaturation. This is particularly noteworthy because fatty acid unsaturation is critical to high pressure growth of SS9.
A search performed using GoMiner software [24] on differentially expressed genes obtained in the 28 MPa vs. 0.1 MPa experiments and in the 28 MPa vs. 45 MPa experiments indicated that transport is one of the main biological processes involved (Additional file 5). Similar result was obtained using FatiGO software [25] (data not shown).
Moreover six of the ORFs that are up-regulated both at 28 MPa (vs. 0.1 MPa) and 45 MPa (vs. 28 MPa) are involved in transport processes (GO:0006810) (PBPRB1789, PBPRB1788, PBPRA2251, PBPRA1366, PBPRA0555, PBPRA1297). As described in the previously published 28 MPa versus 0.1 MPa experiments [4], transport is strongly influenced by pressure, probably due to the effect of pressure on membrane modification and because of the pressure influence on the activation volume ΔV# (the difference between the transition state volume and the initial volume in the system at equilibrium) of the transport process.
Discussion
In the microbial world genetic material can be transferred between species by several mechanisms involving conjugative plasmids, phages, phage-like elements or transposable elements [31]. These elements allow exchange among bacteria of a flexible gene pool encoding additional functions that usually are not essential for bacterial growth, but which provide advantages under particular conditions. Despite the strong pressure on bacteria to maintain a small genome size by deleting the more expendable sequences from their genomes, advantages provided by some regions allow the maintenance of a flexible gene pool [32,33]. Recently [34] it was demonstrated that in the case of Xylella fastidiosa a large part of its flexible gene pool seems to be important in order to explain the broad host range of this phytopatogen. This despite the lack of expression of these genes under the culture conditions examined. Using clones selected from the SS9 genome sequencing project and from a specific genomic library with short inserts, a microarray covering large part of the SS9 genomic sequence (78.0% on chr1, 69.7% on chr2 and 79.3% on plasmid) was prepared and used to identify variable regions in three P. profundum strains. The lower coverage of chr2 was due to the higher frequency of repeated regions present on this chromosome, identified using Phred/Phrap software during finishing step. Clones were selected which lacked these regions in order to avoid cross hybridization.
This type of analysis has previously been applied to other bacterial species [34-37] but this is the first report regarding a microbial species containing members adapted to a high pressure environment. In fact, using genomic DNA derived from deep-sea (SS9 and DSJ4) and coastal (3TCK) isolates we were able to identify 28 regions that are absent or highly divergent in DSJ4 and 3TCK strains plus a large number of small regions scattered over the entire SS9 genome.
In silico analysis performed on the SS9 genome sequence reveals that some of these regions (for example Chr1.15, Chr2.5, Chr2.7-PI) show differences in GC content, codon usage and genomic signature (see also Table 1). There are also regions, such as Chr1 1634244–1651614 bp (containing the omega-3 polyunsaturated fatty acid synthase, pfa, operon), that appear as alien DNA from these analyses, but which are present in all three strains analyzed. While pfa mutants are not pressure-sensitive in laboratory culture [29], it is likely that the pfa operon is involved in high pressure adaptation under some deep-sea conditions because omega-3 polyunsaturated fatty acids (PUFAs) do increase in abundance in SS9 membranes with increasing pressure [29,30] and because such fatty acids are known to modify membrane fluidity in response to hydrostatic pressure and temperature.
These results indicate that bioinformatics and genomic microarray analysis can be merged in order to obtain a more comprehensive picture of the flexible gene pool in bacteria. Moreover, some of the variable regions identified by microarray analysis but not by bioinformatic analysis could derive from genomic reduction and not from lateral gene transfer.
On the other hand, when the two separate analyses are in accordance with each other, lateral gene transfer is the most plausible explanation. This is the case for the 80 kbp plasmid. However, the adaptive value of this plasmid to SS9 remains a mystery. It contains no obvious essential genes and can be lost during laboratory cultivation without a detectable phenotypic change.
A region that is particularly interesting is Chr1.1-LF, consisting mainly of a gene cluster present in SS9 and DSJ4 strains and coding for a second flagellar motility system. It is known that bacterial motility in the sea is a commonly expressed phenotype [38]. This character is important in enhancing bacteria-organic-matter coupling. Moreover, high speed swimming in bacteria may reduce the ability of protozoa to graze on them [39]. Some bacteria, such as Vibrio parahaemolyticus, posses dual flagellar systems that operate under different circumstances: a polar flagellum allows motility in liquid environments (i.e. swimming), while multiple lateral flagella allow translocation over surfaces or in viscous media [40,18]. The role of the second motility system in SS9 and DSJ4 motility is currently being investigated.
The three genomic islands Chr1.8, Chr2.3 and Chr2.5 are likely to be prophages in the P. profundum SS9 genome. It has been suggested that prophages might carry genes beneficial for survival of the host in a selective environment [41]. In this respect the presence, in these regions, of metabolic genes such as NAD(P)H oxidoreductase (PBPRB0548), a putative TrkA family protein (glutathione-regulated potassium efflux system protein) (PBPRB0550) and tryptophanase (PBPRA1344) might be meaningful. The latter gene has two other paralogues in the SS9 genome one of which (PBPRA2532) is pressure regulated suggesting some role in deep-sea adaptation.
Recently [4] we have shown that when moving from 28 MPa to 0.1 MPa, SS9 undergoes various modifications in metabolic processes. At 28 MPa up-regulation of genes involved in the Stickland reaction (an amino acid fermentation process typical of anaerobic bacteria such as Clostridiales and Spirochetales) and of TMAO reduction occurs.
Another interesting metabolic modification involves genes of the citrate fermentation pathway located on region Chr1.13. In Leuconostoc paramesenteroides the citMCDEFGRP operon, involved in citrate utilization, is located on a plasmid [42]. In SS9 these genes are down-regulated both at 28 MPa and at 4°C, moreover this region is absent both in DSJ4 and 3TCK genomes. The reason for the altered expression of this pathway in SS9 is not clear but pressure could favour some metabolic processes only on the basis of chemical-physical parameters. Some of these genes are absent in 3TCK and DSJ4 and this could reduce the growth rate of these strains at high pressure. In fact reactions accompanied by large volume variation are greatly influenced by pressure, but even if the value of ΔV (the difference between the final and initial volume in entire system at equilibrium, reaction volume) or ΔV# (apparent volume change of activation or activation volume) is known, it is still difficult to predict how elevated pressure will affect metabolic pathways in living organisms [43].
As previously reported [4], low pressure induces various stress responses reflected by the up-regulation of chaperones (PBPRA0698, PBPRA1023, PBPRA1484, PBPRA3387) and DNA repair enzymes (PBPRA3513, PBPRB0986, PBPRA0694). This stress response is not present at 45 MPa despite being supraoptimal for SS9 growth. Perhaps low pressure affects protein folding because these SS9 proteins are adapted to high pressure and this effect is not evident in high pressure experiments (at least up to 45 MPa). Another curious result stemming from the 45 MPa transcriptome experiment is the apparent reduction in arginine biosynthesis and transport at high pressure relative to that at 28 MPa. It could be that this large amino acid is selected against within much of the protein pool present at this high pressure, in contrast to the thermophilic and piezophilic Archaea Pyrococcus abyssi which has a higher arginine content in its proteins respect to the non-piezophilic bacterium Pyrococcus furiosus [44]. Alternatively, other genes could govern arginine utilization at high pressure.
Transport seems to be the cellular process most influenced by pressure, at least considering the number of regulated genes belonging to this category. This influence could be due to volume changes associated with the transport process [45]. Also temperature variation influences the transport process, which could be due to alterations of transporters efficiency induced by membrane fluidity modifications. It is interesting to note that some transporters are present in two or more copies in the SS9 genome and therefore could work at specific pressures and temperatures.
Likewise, SS9 may choose among different metabolic processes (amino acid reduction, TMAO reduction, citrate fermentation pathway, etc.) as a function of pressure in order to optimize energy gain. Other mechanisms previously described [4], such as the influence of pressure on enzymes involved in complex polysaccharides utilization (PBPRA0480, PBPRA2198, PBPRB1016), support this hypothesis. The ability of SS9 to choose between different transporters or metabolic strategies could also be related to the fact that SS9 is not an obligate and narrow spectrum piezophile, but is able to grow over a large range of pressures.
Conclusion
Our findings on genome organization and transcriptional activity of different P. profundum strains depict a high level of genetic diversity, where variable regions influence extremely important processes such as motility and energy production. But, due to the complexity of deep-sea environment, characterized by a peculiar combination of chemical-physical parameters and nutrient resources, it is difficult to assign a role to all the variable regions present only in the pressure adapted strains.
Expression studies on P. profundum SS9 performed at different pressure and temperature conditions reveal a complex adaptation network involving a great number of membrane transporters, metabolic processes, and amino acid biosynthesis and membrane modification enzymes. Some of these genes are located on variable regions.
The pressure-regulated genes unique to piezophilic P. profundum strains are likely to be fruitful targets of future genetic investigations into genes which facilitate growth and survival under deep-sea conditions.
Methods
Origin of the isolates and growth conditions
Photobacterium profundum SS9 and Photobacterium profundum DSJ4 have been previously described [2,3]. Photobacterium profundum strain 3TCK was isolated by the University of California San Diego Environmental Microbiology Laboratory class during a cruise on board the RV Robert Gordon Sproul on April 17, 1999. It was obtained from a superficial sediment sample within San Diego Bay following dilution plating onto a modified K Medium [46] and incubation at 23°C for several weeks. One of the resulting colonies subsequently analyzed by 16S rRNA sequence analysis was determined to be a new strain of P. profundum. The 3TCK 16S rDNA was amplified and sequenced using general eubacterial primers [47] on a MegaBACE 1000 automated sequencer (Amersham Biosciences, Piscataway, NJ) according to manufacturer's instructions.
Maintaining bacterial cultures in balanced growth at high pressure, when the cultures are separated from the investigator by ~2 cm of stainless steel, is extremely challenging. Because of these constraints it was not possible for us to serially culture cells under anaerobic, high pressure conditions (at least not without regularly shocking cultures with decompression for several minutes), and thus only one round of culturing was used. Overnight log phase atmospheric pressure cultures of P. profundum SS9 cells were grown in Difco Marine 2216 Broth supplemented with 20 mM glucose and 100 mM HEPES, pH 7.5. These cells were diluted 250-fold into the same medium, transferred to polyethylene bulbs, sealed with no air space and placed inside pressure vessels. Cultures were then rocked gently back and forth in a large refrigerated water bath shaker at either atmospheric pressure, at 28 MPa or at 45 MPa. Pressure was applied using a hand operated Haskel pump and a quick fit connector to the pressure vessel. With this system pressurization requires approximately 1 min and decompression ~1 second.
Under these culture conditions the cells quickly used up most of the available dissolved oxygen, as reflected by resazurin dye reduction analyses. After about 20–30 hours the cultures were harvested. In all cases cells were obtained from early log phase cultures (optical density at A600 nm of 0.1–0.21). The cells were immediately pelleted and RNA was immediately extracted and stored in 75% ethanol at -80°C.
For genomic experiments approximately 1 liter of bacterial cells was harvested by centrifugation for 15 min at 5,000 × g and the pellet was resuspended in 5 ml buffer A (50 mM Tris, 50 mM EDTA, pH 8.0). The suspension was incubated overnight at -20°C. Then 500 μl of buffer B (250 mM Tris, pH 8.0, 10 mg/ml lysozime) were added to the frozen suspension, which was thawed at room temperature and subsequently incubated on ice for 45 min. At this point 1 ml of buffer C (0.5% SDS, 50 mM Tris, 400 mM EDTA, pH 7.5, 1 mg/ml Proteinase K) was added and the suspension was incubated at 50°C for 60 min. Additional 750 μl of buffer C were added followed by another 30 min incubation at 50°C. The genomic DNA was extracted twice with 5 ml. of phenol:chloroform:isoamyl alcohol (24:24:1), and precipitated with 0.8 volumes of isopropanol. The DNA pellet was recovered by spooling on a glass rod, and rehydrated in 4 ml of buffer D (50 mM Tris, 1 mM EDTA, 200 μg/ml RNAse A, pH 8.0) and incubated overnight at 4°C. The solution was extracted once with an equal volume of chloroform, then precipitated with 0.8 volumes of isopropanol. The DNA pellet was recovered by centrifugation, washed once with 70% ethanol and stored dry at -20°C.
For growth curves, the strains (SS9, DSJ4, 3TCK) were grown at 9°C in polyethylene bulbs at different pressures (0.1, 15, 30, 45 and 75 MPa). At time intervals, one bulb was removed and its optical density at 600 nm was recorded. The growth rate was calculated from the log-portion of the curve obtained.
Detection of the 80 kb plasmid in the different P. profundum strains was done using primers
SS9PLAS1F (5'-ACAAGAGGCAGCAAAAAGACTAAC-3'),
SS9PLAS2R (5'-TGCCGCACAGGTAATGATAGGATG-3'),
SS9PLAS3F (TCAGTGCATCGCTAGGGTTAGACT-3'),
SS9PLAS4R (AAAGCATTATGAAAAATTGGTAGA).
The PCR cycle for amplification was 94°C for 5 min, followed by 30 cycles of 94°C for 30 s, 52°C for 30 s, 72°C for 1 min, and a final extension at 72°C for 7 min.
DNA microarray preparation
Most of the microarray clones (2174) identifying a single ORF were selected from a small-insert genomic library, representing 1264 individual ORFs. In order to obtain a general overview of the SS9 genome we selected also 2227 non overlapping clones from a large-insert library, representing other 3488 ORFs. Clones were selected from 384 well plates using Biomek 2000 workstation and were inoculated in 96 well plates with 100 μl of LB/Ampicillin (50 μg/ml). Clones were grown overnight and then 1 μl was used for PCR amplification in 96 well plates. PCR amplification mix (one sample volume) (H2O mQ AF 47.616 μl; buffer 10 × 6 μl; MgCl2 25 mM 3.6 μl; dNTPs 10 mM 1.32 μl; primer -21M13 100 μM 1.32 μl; primer M13REV 100 μM 1.32 μl; Taq polimerase 0.2 μl; total volume 59 μl). The PCR cycle for amplification was 96°C for 5 min, followed by 35 cycles of 96°C for 25 s, 56°C for 30 s, 72°C for 2 min and a final extension step at 72°C for 10 min.
PCR products were purified using ethanol precipitation. PCR reactions were transferred to 96-well U-bottom tissue culture plates (Costar #3790) and 1/10 vol. 3 M sodium acetate (pH 5.2) and 2.5 volumes ethanol were added to the PCR. Plates were mixed by inversion and stored at -20°C overnight. Plates were centrifuged in Sorvall RC-3B at 3500 rpm for 1 h (RCF = 3565 g) and supernatant was poured off from plates. PCR products were washed with 100 μl of ice-cold 70% ethanol and centrifuged again for 30 min. Pellets were dried in speed-vac for 10 min. Before use, PCR products were resuspended in 15 μl 3 × SSC overnight using Titramax 101 plate shaker (Heidolph). For expression microarray experiments amplicons were arrayed onto MICROMAX™ Glass Slides (PerkinElmer Life Sciences, Inc.) using "MicroGrid II" spotter (Biorobotics, Genomic Solutions®) equipped with 16 pins. Microarrays have a spotted area of 18 mm × 18 mm and are composed by 16 subarrays (4 metacolumns and 4 metarows) containing three replicates of each spot. Spots have a diameter of approximately 80 μm and were spaced 150 μm. For genomic microarray experiments amplicons were arrayed onto MICROMAX™ Glass Slides (PerkinElmer Life Sciences, Inc.) using "GenPack Array21" spotter equipped with 32 pins. Microarrays have a spotted area of 36 mm × 18 mm, are composed by 32 subarrays (4 metacolumns and 8 metarows) containing two replicates of each spot. Spots have a diameter of approximately 130 μm and were spaced 250 μm.
Nucleic acid labeling and hybridization conditions
Expression experiments
For microarray analysis at different pressure, bacterial cultures were grown as explained above at two different conditions (0.1 MPa/16°C and 45 MPa/16°C) and were queried against a common reference culture (28 MPa/16°C). For temperature microarray experiments bacterial cultures were grown at 0.1 MPa/4°C and were queried against a culture grown at 0.1 MPa/16°C. For each growth condition total RNA was extracted from three independent cultures using trizol (Gibco) and RNeasy columns (Qiagen). Genomic DNA was removed using DNAsi (Ambion). For microarray expression experiments 2 μl of random hexamer primers (3 mg/ml) were added to 20 μg of total RNA, volume was brought to 18.5 μl with RNase-free water and were labeled with Cy3 and Cy5 using an aminoallyl indirect labeling method according to the TIGR protocols with some minor modifications [48]. Solution was mixed well and incubated at 70°C for 10 min. After denaturation RNA was snap-freezed in dry ice/ethanol bath for 30 s, then it was centrifuged briefly at 10,000 rpm and maintained at room temperature. 6 μl of 5 × First Strand buffer, 3 μl of 0.1 M DTT, 0.6 μl of 10 mM dATP, dCTP and dGTP, 0.9 μl 10 mM dTTP, 0.6 μl 10 mM dUTP-AA, and 2 μl of SuperScript II RT (200U/ μl) (Invitrogen) were added to the solution that was mixed gently and incubated at 42°C for 3 h. To hydrolyze RNA 3 μl of NaOH 1 M and 0.6 μl of 500 mM EDTA were added to the mixture that was incubated at 65°C for 15 min in a water bath. 3 μl of 1 M HCl and 8.5 μl of 2 M HEPES (SIGMA) was added in order to neutralize pH. Removal of unincorporated aa-dUTP and free amines was made using Microcon YM-30 Cleanup method (Millipore). Sample volume was reduced to 9 μl in a speed vac and 1/10th volume of carbonate buffer (Na2CO3 1 M, pH 9) was added to the cDNA and this solution was used to resuspend the Cy- NHS-ester dye. Reaction was incubated for 1 h in the dark at room temperature. Labelled cDNA was purified using the GenEluteTM PCR Clean-up kit (SIGMA). Analysis of labelled cDNA was made using a 50 μl quartz MicroCuvette (Beckman) to analyze the entire undiluted sample in a spectrophotometer. After analysis the two differentially labeled probes (Cy3 vs. Cy5) were combined and precipitated with NH4Ac and ethanol according to standard protocols. DNA was recovered by centrifugation, and the pellet was washed twice with 70% EtOH. The pellet was briefly air dried, then resuspended in hybridization buffer containing 50% formamide, 5 × SSC, 0.1% SDS, 100 ng/μl SS-DNA, 5 × Denhardt's solution. Labeled DNA was heated to 95°C for 2 min and chilled on ice before use in hybridization. Hybridization was performed over night in hybridization chamber (GeneMachines) at 42°C using coverslip. After hybridization microarrays were washed two times for 5 min with 1 × SSC, 0.1% SDS pre-heated at 42°C, two times for 5 min with 0.2 × SSC, 0.1% SDS at room temperature, two times for 5 min with 0.2 × SSC at room temperature and finally two times for 5 min with SSC0.1 × at room temperature.
Genomic experiments
For a 50 μl reaction, 10 μg of genomic DNA (gDNA) was combined with 4 μg of random hexamer primers and heated to 95°C for 5 min. After denaturation gDNA was snap-freezed in dry ice/ethanol bath for 30 s, was centrifuged briefly at 10,000 rpm and mantained at room temperature. 1.5 μl of 10 mM dATP, dCTP and dGTP, 0.9 μl 10 mM dTTP, 0.6 μl 10 mM dUTP-AA, 3 μl of Eco Pol buffer and 15 U of the Klenow fragment of Escherichia coli polymerase (NEB) were added to the reaction. The reaction was placed at 37°C for 2 h. Labeled DNA was then purified and labeled as described in the previous section. After precipitation DNA was recovered by centrifugation, and the pellet was washed with 70% EtOH and re-centrifuged.
Hybridization and post-hybridization washes were performed as described in the previous section "expression experiments".
Image acquisition and analysis
Arrays were scanned with a ScanArray® Lite scanner (PerkinElmer Life Sciences, Inc.). Hybridization signals were quantified using QuantArray software (PerkinElmer Life Sciences, Inc.). Data representing weak signals (median pixel intensity lower than 400 in both channels) were removed. Signal intensities were normalized using MIDAS V2.15 [49] and used to generate relative hybridization ratios (query/reference).
Expression experiments
normalization was performed using loc fit normalization (LOWESS) normalizing each single microarray block separately. The ratios from a maximum of nine data points (triplicate spots, hybridizations performed in triplicate) were analyzed with SAM software [50]. When less than six over nine data points existed, the clone was treated as data missing and was excluded from SAM analysis. Only clones having log2 ratio higher than 0.7 or lower than -0.7 and having d (the T-statistic value) higher than +2.3 or lower than -2.3 were considered. Differentially expressed clones spanning a single ORF were used for the identification of differentially expressed genes and the other clones were used in order to confirm or reject data obtained. All the fluorescence intensity data were used after normalization for the analysis of absolute gene expression on the two chromosomes.
Genomic experiments
analysis was performed as described for expression analysis but only clones having log2 ratio higher than 1 or lower than -1 and having d (the T-statistic value) higher than +2.7 or lower than -2.7 were considered.
GC content, codon bias and genomic signature analysis of SS9 genome
GC content reported in Figure 2 is calculated as the G+C frequency considering 5 kb windows with 0.1 kb shift.
Codon bias and GC frequency in third position (S3%) in SS9 genes longer than 200 codons was calculated using a home-made perl script based on formula described in [9].
Genomic signature was calculated as described by Karlin S. [7] considering tetranucleotide frequency calculated as the odds ratio ρXYZK = fXYZK/fXfYfZfK where fXYZK is the frequency of the tetranucleotide XYZK in the sequence under study. For double-stranded DNA sequences, symmetrized version ρ*XYZK is computed from frequencies of the sequence concatenated with its inverted complementary sequence. ρ*XYZK was calculated for the whole genome (c*XYZK) and for 5 kb windows (r*XYZK) (with a 0.5 kb step). The delta distance, calculated as δ* = 1/256 Σ |r*XYZK-c*XYZK|, was plotted in Figure 2 (10th circle).
Authors' contributions
SC conceived of the study, performed the microarray experiments and analysis, and drafted the manuscript. AV participated in the design and coordination of the study, revised the manuscript and participated in the interpretation of the microarray results. NV participated in the design of the study, performed the bioinformatic analysis (GC content, codon usage, genomic signature, phage identification and UCSC genome browser implementation). FML participated in the design of the study, carried out the P. profundum coltures, growth curves, phylogenetic studies, RNA and gDNA extraction and revised the manuscript. MD and FS participated in microarray production and in the interpretation of the microarray results. AC participated in Gene Ontology annotation. GM participated in microarray experiments. GB participated in the interpretation of the microarray results. GV and DHB participated in the design and in the coordination of the study, in the interpretation of the microarray results and revised the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Codon bias relative to average gene versus third position GC content in eight variable regions of the Photobacterium profundum SS9 genome. In these graphs are represented only ORFs longer than 200 codons. Eight variable regions of the SS9 genome are considered, one for each graph. In red are highlighted ORFs located in regions that were found absent in 3TCK/DSJ4 genomes using comparative genomic hybridization experiments. A large number of ORFs are positioned in left and right horn of the graphs, a strong indication that they belong to laterally transferred regions.
Click here for file
Additional File 2
Detection of the 80 kbp plasmid in two different P. profundum strains. Comparison of strain TW30 (lanes A, B) with parental strain DB110 (lanes C, D) with plasmid specific primers. Lane E: no DNA control for PCR. Lane F: 2-log ladder marker (New England Biolabs).
Click here for file
Additional File 3
List of differentially expressed genes obtained in microarray experiments. This table reports only the ORFs identified univocally by the microarray clones and only those having log2 ratio ≥|0.7| and "d score" obtained from SAM analysis = |2.3| (indicated with "1" in columns 7–12). Table columns report respectively: (1, 2, 3) ORFs differentially expressed in the three conditions under analysis, (4, 5) clones that were found absent in 3TCK and DSJ4 strains, (6) ORFs that were found absent in 3TCK strain ("3TCK"), in DSJ4 strain ("DSJ4") or in both strains ("3TCK-DSJ4"), (7) ORFs up-regulated at 28 MPa, (8) ORFs down-regulated at 28 MPa, (9) ORFs up-regulated at 4°C, (10) ORFs down-regulated at 4°C, (11) ORFs up-regulated at 45 MPa, (12) ORFs down-regulated at 45 MPa, (13) ORF annotation, (14) spotted microarray clones identifying each ORF, (15, 16), (17, 18), (19, 20) ("log2 ratio" and "d value") mean value of expression differences calculated as log2 [(Σ fluorescence value at 28 MPa)/(Σ fluorescence value at 0.1 MPa)] and the "d" statistic value obtained from SAM analysis. Each ORF is identified by one or more clones and not all ORFs identified from each clone are reported in the table (for a complete picture see the UCSC genome browser [12]). Row data obtained from microarray experiments were submitted to ArrayExpress database at EBI with accession numbers E-MEXP-210, E-MEXP-348, E-MEXP-374, E-MEXP-375, E-MEXP-376.
Click here for file
Additional File 4
List of ORFs absent only in 3TCK strain. Data obtained from genomic comparison experiments between 3TCK (the pressure sensitive strain) and SS9 strain. In this table are reported ORFs that were found absent only in 3TCK and not in DSJ4 strain. In order to obtain a more reliable result, clones obtained from SAM analysis of 3TCK strain were filtered considering only those having log2 ratio higher than +1 and d statistic value higher than +1.8. Using self-written PERL scripts, from the "3TCK clones" set, were subtracted the clones that were found absent also in DSJ4. In DSJ4 strain we used less stringent criteria and we considered all clones having log2 ratio higher than 0.5. Clones absent both in 3TCK and DSJ4 were discarded. Clones absent only in 3TCK were associated with ORFs position and the list of ORFs obtained were manually verified. Table columns show: (1) the locus name, (2) the ORF description, (3) the TrEMBL code, (4) the microarray clone/s overlapped to the ORFs, (5) the log2 ratio obtain from comparison with SS9 genome (reference), (6) the "d" statistic value obtained from SAM analysis and (7) the result obtained in microarray expression experiments. In the first column large groups of ORFs that are adjacent in the SS9 genome are highlighted using bold character.
Click here for file
Additional File 5
List of Gene Ontology categories of differentially expressed genes. In this table differentially expressed genes reported in Additional file 3 are categorized in Gene Ontology classes using GoMiner software. Only highly significant GO classes are reported. Table columns show: (1) the Gene Ontology ID, (2) the total number of P. profundum SS9 genes belonging to each class, (3, 4, 5) the "p statistic value" of down-regulated, up-regulated and differentially expressed genes, (6) the Gene Ontology term, (7) the "UniProtKB/TrEMBL" primary accession number, (8) the ordered locus name, (9) the name of differentially expressed genes belonging to each Gene Ontology category and (10) their "up-" or "down-regulation".
Click here for file
Acknowledgements
We are grateful to the Italian MIUR (grant FIRB/RBAU012RN8/RBNE01F5WT_007) and Fondazione CARIPARO for financial support. D.H.B. and F.M.L. are grateful to the National Science Foundation (NSF/MCB 02–37059) for financial support. We thank Bradley M. Tebo and Chris A. Francis for help in the isolation of Photobacterium profundum 3TCK and C. Romualdi for statistical analysis suggestions.
==== Refs
Bartlett DH Pressure effects on in vivo microbial processes Biochem Biophys Acta 2002 1595 367 381 11983409
DeLong EF Adaptations of deep-sea bacteria to the abyssal environment PhD thesis 1986 University of California
Nogi Y Masui N Kato C Photobacterium profundum sp. nov., a new, moderately barophilic bacterial species isolated from a deep-sea sediment Extremophiles 1998 2 1 7 9676237 10.1007/s007920050036
Vezzi A Campanaro S D'Angelo M Simonato F Vitulo N Lauro FM Cestaro A Malacrida G Simionati B Cannata N Romualdi C Bartlett DH Valle G Life at depth:Photobacterium profundum genome sequence and expression analysis Science 2005 307 1459 1461 15746425 10.1126/science.1103341
Royer CA Application of pressure to biochemical equilibria: the other thermodynamic variable Methods Enzymol 1995 259 357 377 8538462
Heidelberg JF Eisen JA Nelson WC Clayton RA Gwinn ML Dodson RJ Haft DH Hickey EK Peterson JD Umayam L Gill SR Nelson KE Read TD Tettelin H Richardson D Ermolaeva MD Vamathevan J Bass S Qin H Dragoi I Sellers P McDonald L Utterback T Fleishmann RD Nierman WC White O Salzberg SL Smith HO Colwell RR Mekalanos JJ Venter JC Fraser CM DNA sequence of both chromosomes of the cholera pathogen Vibrio cholerae Nature 2000 406 477 482 10952301 10.1038/35020000
Karlin S Burge C Dinucleotide relative abundance extremes: a genomic signature Trends Genet 1995 11 283 290 7482779 10.1016/S0168-9525(00)89076-9
Dufraigne C Fertil B Lespinats S Giron A Deschavanne P Detection and characterization of horizontal transfers in prokaryotes using genomic signature Nucleic Acids Res 2005 33 e6 15653627 10.1093/nar/gni004
Karlin S Mrazek J Campbell AM Codon usages in different gene classes of the Escherichia coli genome Mol Microbiol 1998 29 1341 1355 9781873 10.1046/j.1365-2958.1998.01008.x
Karlin S Detecting anomalous gene clusters and pathogenicity islands in diverse bacterial genomes Trends Microbiol 2001 9 335 343 11435108 10.1016/S0966-842X(01)02079-0
Centre de Génétique Moléculaire (CGM)
Photobacterium profundum genome project
Motallebi-Veshareh M Balzer D Lanka E Jagura-Burdzy G Thomas CM Conjugative transfer functions of broad-host-range plasmid RK2 are co-regulated with vegetative replication Mol Microbiol 1992 6 907 920 1376390
Welch TJ Bartlett DH Identification of a regulatory protein required for pressure-responsive gene expression in the deep-sea bacterium Photobacterium species strain SS9 Molecular Microbiology 1998 27 977 985 9535087 10.1046/j.1365-2958.1998.00742.x
Chi E Bartlett DH Use of a reporter gene to follow high-pressure signal transduction in the deep-sea bacterium Photobacterium sp. strain SS9 Journal of Bacteriology 1993 175 7533 7540 8244922
Kurland CG Canback B Berg OG Horizontal gene transfer: a critical view Proc Natl Acad Sci U S A 2003 100 9658 9662 12902542 10.1073/pnas.1632870100
Kim YK McCarter LL Analysis of the polar flagellar gene system of Vibrio parahaemolyticus J Bacteriol 2000 182 3693 3704 10850984 10.1128/JB.182.13.3693-3704.2000
Stewart BJ McCarter LL Lateral flagellar gene system of Vibrio parahaemolyticus J Bacteriol 2003 185 4508 4518 12867460 10.1128/JB.185.15.4508-4518.2003
Gryllos I Shaw JG Gavin R Merino S Tomas JM Role of flm locus in mesophilic Aeromonas species adherence Infect Immun 2001 69 65 74 11119490 10.1128/IAI.69.1.65-74.2001
Knirel YA Lindner B Vinogradov E Shaikhutdinova RZ Senchenkova SN Kocharova NA Holst O Pier GB Anisimov AP Cold temperature-induced modifications to the composition and structure of the lipopolysaccharide of Yersinia pestis Carbohydr Res
Dysvik B Jonassen I J-Express: exploring gene expression data using Java Bioinformatics 2001 17 369 370 11301307 10.1093/bioinformatics/17.4.369
Bordi C Theraulaz L Mejean V Jourlin-Castelli C Anticipating an alkaline stress through the Tor phosphorelay system in Escherichia coli Mol Microbiol 2003 48 211 223 12657056 10.1046/j.1365-2958.2003.03428.x
Ashburner M Ball CA Blake JA Botstein D Butler H Cherry JM Davis AP Dolinski K Dwight SS Eppig JT Harris MA Hill DP Issel-Tarver L Kasarskis A Lewis S Matese JC Richardson JE Ringwald M Rubin GM Sherlock G Gene ontology: tool for the unification of biology. The Gene Ontology Consortium Nat Genet 2000 25 25 29 10802651 10.1038/75556
Zeeberg BR Feng W Wang G Wang MD Fojo AT Sunshine M Narasimhan S Kane DW Reinhold WC Lababidi S Bussey KJ Riss J Barrett JC Weinstein JN GoMiner: A Resource for Biological Interpretation of Genomic and Proteomic Data Genome Biology 2003 4 R28 12702209 10.1186/gb-2003-4-4-r28
Al-Shahrour F Díaz-Uriarte R Dopazo J FatiGO: a web tool for finding significant associations of Gene Ontology terms with groups of genes Bioinformatics 2004 20 578 580 14990455 10.1093/bioinformatics/btg455
Church MJ Hutchins DA Ducklow HW Limitation of bacterial growth by dissolved organic matter and iron in the Southern ocean Appl Environ Microbiol 2000 66 455 466 10653704 10.1128/AEM.66.2.455-466.2000
Macdonald AG Cossins AR The theory of homeoviscous adaptation of membranes applied to deep-sea animals Symp Soc Exp Biol 1985 39 301 322 3938881
Hazel JR Williams EE The role of alterations in membrane lipid composition in enabling physiological adaptation of organisms to their physical environment Prog Lipid Res 1990 29 167 227 2131463 10.1016/0163-7827(90)90002-3
Allen EE Facciotti D Bartlett DH Monounsaturated but not polyunsaturated fatty acids are required for growth of the deep-sea bacterium Photobacterium profundum SS9 at high pressure and low temperature Appl Environ Microbiol 1999 65 1710 1720 10103272
Allen EE Bartlett DH Structure and regulation of the omega-3 polyunsaturated fatty acid synthase genes from the deep-sea bacterium Photobacterium profundum strain SS9 Microbiology 2002 148 1903 1913 12055309
Ochman H Lawrence JG Groisman EA Lateral gene transfer and the nature of bacterial innovation Nature 2000 405 299 304 10830951 10.1038/35012500
Andersson SGE Kurland CG Reductive evolution of resident genomes Trends Microbiol 1998 6 263 268 9717214 10.1016/S0966-842X(98)01312-2
Andersson JO Andersson SGE Insights into the evolutionary process of genome degradation Curr Opin Genet Dev 1999 9 664 671 10607609 10.1016/S0959-437X(99)00024-6
Nunes LR Rosato YB Muto NH Yanai GM da Silva VS Leite DB Microarray analyses of Xylella fastidiosa provide evidence of coordinated transcription control of laterally transferred elements Genome Res 2003 13 570 578 12670998 10.1101/gr.930803
Dorrell N Mangan JA Laing KG Hinds J Linton D Al-Ghusein H Barrell BG Parkhill J Stoker NG Karlyshev AV Butcher PD Wren BW Whole genome comparison of Campylobacter jejuni human isolates using a low-cost microarray reveals extensive genetic diversity Genome Res 2001 11 1706 1715 11591647 10.1101/gr.185801
Pearson BM Pin C Wright J I'Anson K Humphrey T Wells JM Comparative genome analysis of Campylobacter jejuni using whole genome DNA microarrays FEBS Lett 2003 554 224 230 14596944 10.1016/S0014-5793(03)01164-5
Hinchliffe SJ Isherwood KE Stabler RA Prentice MB Rakin A Nichols RA Oyston PC Hinds J Titball RW Wren BW Application of DNA microarrays to study the evolutionary genomics of Yersinia pestis and Yersinia pseudotuberculosis Genome Res 2003 13 2018 2029 12952873 10.1101/gr.1507303
Grossart HP Riemann L Azam F Bacterial motility in the sea and its ecological implications Aquat Microb Ecol 2001 25 247 258
Matz C Jürgens K The role of bacterial phenotypic traits in bacteria-flagellate interactions: which is the most effective mechanism 9th International Symposium on Microbial Ecology: Amsterdam 2001 157 26–31 August 2001;
Shinoda S Okamoto K Formation and function of Vibrio parahaemolyticus lateral flagella J Bacteriol 1977 129 1266 1271 845115
Lindell D Sullivan MB Johnson ZI Tolonen AC Rohwer F Chisholm SW Transfer of photosynthesis genes to and from Prochlorococcus viruses Proc Natl Acad Sci U S A 2004 101 11013 11018 15256601 10.1073/pnas.0401526101
Martin M Corrales MA de Mendoza D Lopez P Magni C Cloning and molecular characterization of the citrate utilization citMCDEFGRP cluster of Leuconostoc paramesenteroides FEMS Microbiol Lett 1999 174 231 238 10339813 10.1016/S0378-1097(99)00145-7
Abe F Kato C Horikoshi K Pressure-regulated metabolism in microorganisms Trends Microbiol 1999 7 447 453 10542424 10.1016/S0966-842X(99)01608-X
Di Giulio M A comparison of proteins from Pyrococcus furiosus and Pyrococcus abyssi: barophily in the physicochemical properties of amino acids and in the genetic code Gene 2005 346 1 6 15716096 10.1016/j.gene.2004.10.008
Abe F Horikoshi K Tryptophan permease gene TAT2 confers high-pressure growth in Saccharomyces cerevisiae Mol Cell Biol 2000 20 8093 8102 11027279 10.1128/MCB.20.21.8093-8102.2000
Van Waasbergen LG Hoch JA Tebo BM Genetic analysis of the marine manganese-oxidizing Bacillus sp. strain SG-1: Protoplast transformation, Tn917 mutagenesis, and identification of chromosomal loci involved in manganese oxidation J Bacteriol 1993 175 7594 7603 8244928
DeLong EF Archaea in coastal marine environments Proc Natl Acad Sci U S A 1992 89 5685 5689 1608980
TIGR Microarray Protocols and Standard Operating Procedures
Dudoit S Gentleman RC Quackenbush J Open source software for the analysis of microarray data Biotechniques 2003 Suppl 45 51 12664684
Tusher VG Tibshirani R Chu G Significance analysis of microarrays applied to the ionizing radiation response Proc Natl Acad Sci U S A 2001 98 5116 5121 11309499 10.1073/pnas.091062498
|
16162277
|
PMC1239915
|
CC BY
|
2021-01-04 16:32:48
|
no
|
BMC Genomics. 2005 Sep 14; 6:122
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-122
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-1261616474710.1186/1471-2164-6-126Methodology ArticleFish and chips: Various methodologies demonstrate utility of a 16,006-gene salmonid microarray von Schalburg Kristian R [email protected] Matthew L [email protected] Glenn A [email protected] Gordon D [email protected] A Ross [email protected] Colleen C [email protected] William S [email protected] Ben F [email protected] Centre for Biomedical Research, University of Victoria, Victoria, British Columbia, V8W 3N5, Canada2 Great Lakes WATER Institute, University of Wisconsin-Milwaukee, Milwaukee, WI, 53204, USA3 The Prostate Centre at Vancouver General Hospital, Gene Array Facility, Vancouver, British Columbia, V6H 3Z6, Canada4 Department of Molecular Biology and Biochemistry, Simon Fraser University, Burnaby, British Columbia, V5A 1S6, Canada2005 15 9 2005 6 126 126 14 6 2005 15 9 2005 Copyright © 2005 von Schalburg et al; licensee BioMed Central Ltd.2005von Schalburg et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We have developed and fabricated a salmonid microarray containing cDNAs representing 16,006 genes. The genes spotted on the array have been stringently selected from Atlantic salmon and rainbow trout expressed sequence tag (EST) databases. The EST databases presently contain over 300,000 sequences from over 175 salmonid cDNA libraries derived from a wide variety of tissues and different developmental stages. In order to evaluate the utility of the microarray, a number of hybridization techniques and screening methods have been developed and tested.
Results
We have analyzed and evaluated the utility of a microarray containing 16,006 (16K) salmonid cDNAs in a variety of potential experimental settings. We quantified the amount of transcriptome binding that occurred in cross-species, organ complexity and intraspecific variation hybridization studies. We also developed a methodology to rapidly identify and confirm the contents of a bacterial artificial chromosome (BAC) library containing Atlantic salmon genomic DNA.
Conclusion
We validate and demonstrate the usefulness of the 16K microarray over a wide range of teleosts, even for transcriptome targets from species distantly related to salmonids. We show the potential of the use of the microarray in a variety of experimental settings through hybridization studies that examine the binding of targets derived from different organs and tissues. Intraspecific variation in transcriptome expression is evaluated and discussed. Finally, BAC hybridizations are demonstrated as a rapid and accurate means to identify gene content.
==== Body
Background
Atlantic salmon are part of the Salmonidae family which comprise all salmon, trout, whitefish, grayling, and charr. A tremendous amount of basic biology is already known about salmonids from studies carried out on their physiology, population dynamics, behavioural ecology and phylogenetics [1]. Salmon also provide an excellent model system in which to study fundamental genetic mechanisms of growth, development, reproduction and response to infection and disease. For example, salmonids serve as prominent models for studies involving environmental toxicology [2], carcinogenesis [3], comparative immunology [4], the molecular genetics and physiology of the stress response [5], olfaction [6], vision [7], osmoregulation [8], growth [9] and gametogenesis [10].
Answers to fundamental scientific questions can also be gained from the study of salmonid genomes. The ancestor of all extant salmonids underwent a whole genome duplication and after a series of subsequent genetic events, salmon are now considered to be pseudo-tetraploid. How a genome reorganizes itself to cope with a duplicated genome and the importance of gene duplications for evolution and adaptation are long standing issues that remain unresolved. Questions regarding the origins of genomes have direct implication for our understanding of the roles of gene families, duplication and deletion of segments of genomes, and the mutational process in human health and disease. They also provide a foundation for understanding the genome of Atlantic salmon to benefit conservation and enhancement of wild stocks, aquaculture and environmental assessments. Genomic resources enable us to address fundamental scientific questions concerning the evolution of salmonid genomes, and the expression of genes and proteins in a wide variety of natural and altered environments and conditions.
Toward these goals, more than 175 cDNA libraries have been constructed from a wide variety of tissues and different developmental stages and more than 300,000 salmonid cDNA sequence reads have been combined from a consortium comprising groups from Canada (Ben Koop et al. and the Genomics Research on Atlantic Salmon Project (GRASP); Susan Douglas et al. and the Institute for Marine Biosciences, NRC); France (Yann Guiguen et al. and INRA-SCRIBE); Norway (Bjorn Hoyheim et al. and the Norwegian School of Veterinary Science (NSVS)) and the U.S.A. (Caird Rexroad III and the USDA/ARS National Center for Cool and Cold Water Aquaculture). These sequences were assembled into over 40,000 unique contigs. A preliminary microarray of 3,557 cDNAs was constructed and assessed on its' ability to provide new data in the study of cellular and tissue responses to pollutants, diseases and stress, as well as for reproduction and development [11-15]. On the basis of these results, a larger array of 16,006 genes has been constructed and initial results have shown sensitivity of gene expression patterns to disease challenge, and to small environmental and physiological changes [16].
Results and discussion
Library construction (directional cloning by 5'EcoRI, 3'XhoI in pBluescript II XR, Stratagene; or TOPO TA cloning of suppression subtractive hybridization PCR products, Invitrogen and Clontech) and subsequent EST sequencing (using M13 forward primer) were designed to generate 3'-end sequences to enable us to distinguish between potential paralogs arising from the recent salmonid genome duplication. We have determined from a weighted average measurement comparing four different directionally-cloned library types (such as non-normalized versus normalized libraries) that approximately 9% of inserts are in the reverse orientation and therefore yield 5' sequence with the M13 forward primer [11]. The GRASP 3'-end reads were used as a framework on which to build the contigs from additional data provided by the NRC, INRA, USDA/ARS and the NSVS. Part of the evaluation process for selecting genes for the microarray required criteria that would guard against chimeras. Simply put, this meant that each gene choice had to be part of a contig with multiple distinct clones covering each region, or that it was sufficiently similar to another sequence across its whole length that it was unlikely to be chimeric. We did select for immune-specific and reproduction-relevant genes for the microarray, but the preponderance of ESTs on the 16K chip were randomly picked based on EST cluster quality and uniqueness and therefore represent a wide variety of different classes of genes.
Application of a 16K cDNA microarray to different species
To explore the validity of using the 16K microarray with other fish species, the 13,421 Atlantic salmon (AS) and 2,576 rainbow trout (RT) cDNA features were interrogated with labeled liver targets from four members of the order Salmoniformes (AS, RT, chinook salmon and lake whitefish) and one member of the order Osmeriformes (rainbow smelt) (Table 1). The average percentage binding of AS, RT, chinook salmon, lake whitefish (LW) and rainbow smelt liver targets to the 16K chip was 54.0%, 63.3%, 51.0%, 50.6% and 30.1%, respectively. The average percentage of targets bound to AS and RT features for each species are also shown (Table 1).
Table 1 Determination of features bound by labeled cDNAs from different species on the 16K salmonid microarray.
Average % bound to all featuresa Average % bound to S. salar featuresa Average % bound to O. mykiss featuresa
S. salar (n = 8) 54.0 ± 7.8 52.6 ± 7.9 59.5 ± 7.5
O. mykiss (n = 4) 63.3 ± 9.1 60.7 ± 9.6 74.4 ± 6.6
O. tschawytscha (n = 4) 51.0 ± 4.3 48.3 ± 4.4 62.9 ± 4.8
C. clupeaformis (n = 2) 50.6 ± 2.1 48.9 ± 2.4 57.8 ± 0.9
O. mordax (n = 2) 30.1 ± 3.5 28.8 ± 3.2 35.1 ± 5.4
aAverage % bound ± standard deviation
Our study indicates that there are no significant differences in the percent of targets that bound to the 16K microarray for the four salmonids examined (AS, RT, chinook and LW). There is a similar hybridization performance for all salmonids. However, RT targets do consistently show higher overall binding to the microarray; the reason for this efficiency is not yet clear.
The hybridization performance of the rainbow smelt targets were roughly one-half those of the salmonid cDNAs. Of the species contributing targets to our heterologous hybridization experiment, the osmerid targets were the most phylogenetically removed from the salmonid features. Indeed, a recent mitogenomic study places the Osmeroidei in a separate clade from the Salmoniformes [17]. These two clades are separated by at least 200 MY with the Salmonidae having undergone at least one genome duplication event since their divergence [18,19]. Other factors such as genome gene content (ie., numbers of paralogs) and genome size are likely to be factors affecting the overall degree of hybridization [11].
Application of a 16K cDNA microarray to different tissues
Different tissues and organs exhibit differences in transcriptome complexity, depending on their cellular heterogeneity and differentiated specializations. The mRNAs of a typical somatic cell are divided into three classes based on their sequence complexity and diversity [20]. The most prevalent class consists of only a few mRNA species that comprise the abundant transcripts present in a cell. Often these transcripts are dedicated to cellular functions common to all tissues, but they usually represent genes that specify an organs' unique function. The high complexity class of mRNAs includes thousands (perhaps millions) of different mRNA species, each represented by fewer than 15 copies per cell [20].
However, it should be noted that some subsets of genes that have been thought to be unique to one organ have been found to be expressed in others. This has been demonstrated for transcripts in the brain-gonad axis, and is probably not exclusive to these organs. For example, mammalian pheromone/odorant receptors and specific piscine hormones and receptors of the brain are also expressed in the gonad [12,21,22]. To date, the biological functions of these transcripts in the gonad have not been determined, raising intriguing questions regarding multiplicity of functions for complex transcripts, even in diploid vertebrates such as mammals.
To determine the differences in the transcriptome complexity of seven different AS tissues and organs, the 13,421 AS and 2,576 RT cDNA features were hybridized with labeled targets from midgut, brain, spleen, muscle, ovary, kidney and testis (Table 2). The average percentage binding of midgut, brain, spleen, muscle, ovary, kidney and testis targets to the 16K chip was 64.4%, 54.7%, 54.6%, 52.8%, 51.0%, 49.7% and 30.2%, respectively. In general, about 45% of the salmonid microarray features were not bound by targets from the various AS tissues and organs.
Table 2 Determination of features bound by labeled cDNAs from different tissues on the 16K salmonid microarray.
Tissue Type Average % bound
Brain 54.7 ± 12.2
Kidney 49.7 ± 10.5
Midgut 64.4 ± 5.1
Spleen 54.6 ± 0.8
Testis 30.2
Ovary 51.0
Muscle 52.8 ± 9.7
Liver 54.0 ± 7.8
Application of a 16K cDNA microarray to the same tissue from cohorts
To determine the amount of gene expression variability that exists between individuals of a single species, we compared the transcriptomes of livers from three fish with identical histories. We compared the average percent of variation (or scatter) in expression of liver transcripts between cohorts 1 and 2 (liverpairs 1/2), cohorts 1 and 3 (liverpairs 1/3) and cohorts 2 and 3 (liverpairs 2/3). Two separate experiments of six hybridizations each were conducted with each liverpairing having one dye-flip.
Examining each individual array in the intraspecies study showed that the overall mean scatter was 12.6% (Table 3). When the liverpair arrays and their respective dye-flips were combined and averaged, the overall mean scatter was reduced to 9.7%. This indicates that systematic unequal dye incorporation exists resulting in high scatter values. This dye bias has been well-documented by other researchers [23-25] and illustrates the importance of incorporating dye swap pairs when performing microarray hybridizations whenever possible. The overall mean scatter was further reduced to 5.2% when the analysis included technical dye swap replicates between respective liverpairs (Table 3). This demonstrates that increasing the number of technical replicates in a microarray experiment is an important factor to consider for reducing random scatter. It is encouraging that the overall scatter between individuals from the same broodstock was quite low. Thus technical and biological variability across arrays and individuals can be significantly reduced by the investigator if the appropriate experimental design is employed.
Table 3 Determination of variation in liver transcriptome expression between three cohorts.
Mean scatter for individual arrays Mean scatter for dye swap pairs Mean scatter for technical replicates
12.6% 9.7% 5.2%
Application of a 16K cDNA microarray to analyze BAC contents
To assess the use of the 16K array as a screening tool to identify the genes present in a salmonid BAC, the 13,421 AS and 2,576 RT features were interrogated with nebulized and labeled fragments from a single BAC whose sequence has been determined (Table 4). Analysis of our initial BAC hybridizations revealed that a high proportion of transposon-like sequences and long and short interspersed nuclear elements were binding to the array. It is known that many different repeat elements derived from once-mobile transposable segments comprise large portions of the Atlantic salmon genome [26-29]. In an effort to improve the specificity of target binding to the microarray for BAC hybridization, we employed a Cot-1 DNA protocol to reduce the binding of these repetitive elements (Table 4). The addition of Cot-1 DNA increased the number of expected genes identified and the number of hits for the expected genes by displacing many of the repeat family and transposon associated elements.
Table 4 Analysis of gene content in BAC hybridizations.
BAC Name Expected Gene Numbera Hybridized BAC (no Cot) Hybridized BAC (with Cot) Repetitive Elements (no Cot)b Repetitive Elements (with Cot)b Transposon Associated Sequences (no Cot)c Transposon Associated Sequences (with Cot)c
92I04 8 5 8 5 4 9 7
aBAC 92I04 was previously characterized. The assembled BAC sequence was BLASTED against the microarray gene identification list to determine the expected gene number. Repetitive elements, transposon-associated sequences and unknown ESTs were not included in the total. Only the top 50 hits were examined.
bNumber of identified microsatellite and repeat family elements.
cNumber of identified transposon, transposase and reverse-transcriptase associated sequences.
Although Cot-1 DNA did improve the ability to identify genes for the BAC we examined, Cot-1 DNA alone is not enough to block the complications that arise from repetitive elements in whole genome hybridizations. In preliminary comparative genomic hybridization studies we have found that even with Cot-1 DNA included in the hybridizations, the repetitive DNA segments found in salmonid genomes interfere with the interpretation of the data. Most investigators are not interested in these repetitive segments, but rather in the genes that are interspersed between them. Moreover, we have found that often these repetitive elements lead to false positives. Using other methods, such as including repeat-element amplified products with Cot DNA, as well as higher stringency washes, might improve binding specificities. We are currently working on various strategies to maximize blocking of this repeat element 'noise'.
Conclusion
We validate and demonstrate the usefulness of the 16K microarray over a wide range of teleosts, even for transcriptome targets distantly removed from salmonids phylogenetically. We show the potential of the use of the microarray in a variety of experimental settings through hybridization studies that examine the binding of targets derived from different organs and tissues. Intraspecific variation in transcriptome expression is evaluated and discussed. Finally, BAC hybridizations are demonstrated as a rapid and accurate means to identify gene content. We expect that this array will serve as an important resource for genetic, physiological, ecological and many other fields of salmonid study.
Methods
Gene selection
cDNA library construction, recombinant plasmid preparation and extraction, sequencing, sequence analysis and contig assembly for the GRASP have been described previously in detail [11-13]. Selection criteria for unique Atlantic salmon (AS) and rainbow trout (RT) cDNAs for inclusion on the 16K microarray were as follows: ESTs (cDNA fragments) were assembled into contiguous sequences (contigs) by PHRAP [30] under stringent assembly parameters (minimum overlap score:100; repeat stringency: 0.99). Contig consensus sequences and singleton sequences were aligned with non-redundant GenBank nucleotide and amino acid sequence databases using BLASTN and BLASTX, respectively [31,32]. Threshold for a significant BLAST hit was set at E = 1e-15.
It was determined that a contig must contain at least one "usable" sequence, where "usable" was a)- the sequence must be 3' (with high probability; containing polyA signal or having been sequenced with an oligo-dT primer or being at the 3'-end of a contig, with orientation determined by a strong hit against a protein in GenBank's non-redundant protein database), b)- be a sequence stretch containing more than 400 bp, and c)- the sequence must be at least 95% similar to the consensus of the contig.
It was also determined that if a contig was a singleton or singleton-equivalent (where all sequences were from the same plate or library thus not providing sufficient evidence for non-chimera status), then the contig selection was reinforced either by a)- a significant BLAST hit, E<1e-15 (BLASTN or BLASTX), or b)- it having 94% (or more) identity with a homolog (either paralog or ortholog) covering at least 400 nucleotides. If the contig was a non-singleton, it was determined that it must be a)- one "block" (having no regions in the interior of the contig covered by only one sequence, to decrease probability of chimeras), and b)- of high enough overall quality (with an overall score > 95% positions without conflicts, weighted by number of sequences which support the consensus) and c)- have few leading and trailing singleton positions (no more than 25%), since such positions make it a de facto singleton.
Approximately 3,500 additional sequences were selected with the following criteria: a)- no chosen contig could have 94% or more identity with another chosen contig, and b)- tentative consensus sequences (TC) identified by TIGR [33] could be included. By these criteria, approximately 1000 clones were picked indiscriminately from both normalized AS and RT cDNA libraries, 800 clones were selected from suppression subtracted hybridization libraries and 700 sequences were added from requests of potential array users. Additionally, 949 non-overlapping sequences (856 AS, 93 RT) from clones included in the preliminary 3,557-gene chip (plus one T cell receptor beta) were selected. Finally, approximately 500 immune-specific genes were also chosen to bring the total number of genes represented on the chip to 16,006. In the 16,006 cDNA features there are 13,421 AS, 2,576 RT, 4 chinook salmon, 3 rainbow smelt and 2 LW representatives.
Gene identification
EST contigs were built using cDNAs on the array as reference and all ESTs currently in the GRASP database. Subsequent to microarray fabrication, the consensus sequences were screened for repeats using a custom salmonid repeat database with RepeatMasker. Masked consensus sequences were compared to GenBank databases. Using the stringent selection threshold above, the current percentage of the 16K features that are known and unknown genes is 55.8% and 44.2%, respectively. Analysis at less stringent thresholds is ongoing to identify all genes on the microarray.
Microarray fabrication
Clones were robotically rearrayed from daughter glycerol stock 384-well plates into 96-well plates pre-filled with 7% glycerol in LB + ampicillin, incubated overnight at 37°C, and checked for uniform optical density. Plasmid inserts were PCR amplified in a Tetrad PTC-200 thermocycler (MJ Research) using 1 ul overnight culture, 0.2 mM M13/pUC forward primer (5'-CCCAGTCACGACGTTGTAAAACG-3'), 0.2 mM M13/pUC reverse primer (5'-AGCGGATAACAATTTCACACAGG-3'), 2 mM MgCl2, 10 mM Tris-HCl, 50 mM KCl, 250 mM dNTPs, 1U AmpliTaq (Perkin Elmer), and nuclease-free H2O (Gibco) to 100 ul. PCR conditions were: 2 min for 95°C; then 35 cycles of 95°C for 30 sec, 60°C for 45 sec, 72°C for 3 min; followed by 72°C for 7 min. Five ul of each PCR product were run on a 1% agarose gel to assess yield and quality. PCR products were robotically cleaned (Qiagen) and consolidated into 384-well plates, lyophilized by speed-Vac, and resuspended in 20 ul 3X SSC. Each purified PCR product concentration was determined and diluted to give a final concentration of 400 ng/uL.
All cDNAs were printed as single spots on EZ Rays aminosilane slides (Matrix/Apogent Discoveries) with the Biorobotics Microgrid II microarray printer (Genomic Solutions). Microspot™ 10K quill pins (Biorobotics) in a 48 pin tool were used to deposit approximately 0.5 nl (0.2 ng cDNA) per spot onto the slide. The slides were crosslinked in a UV Stratalinker 2400 (Stratagene) at 300 mJ. The resulting microarrays have a 4-by-12 metagrid layout with 19 X19 spot subgrid, each spot having an approximate diameter and pitch of 100 um and 0.20 mm, respectively. A 280 bp GFP (green fluorescent protein) cDNA was amplified from a GFP clone (Clontech) using the primers (5'-GAAACATTCTTGGACACAAATTGG-3') and (5'-GCAGCTGTTACAAACTCAAGAAGG-3') and printed in each subgrid corner to assist in gridding.
Six exogenous genome (Arabidopsis) cDNAs were amplified from the following clones kindly provided by The Arabidopsis Information Resource: rubisco activase [GenBank:T41667], protochlorophyllide reductase precursor [GenBank:R30630], chlorophyll a/b-binding protein CP29 [GenBank:N65746], PSII oxygen-evolving complex protein 2 [GenBank:H36167], tonoplast intrinsic protein root-specific RB7 [GenBank:AA067532] and ferredoxin (2Fe_2S) precursor [GenBank:W43249]. The Arabidopsis cDNAs were spotted in quadruplicate on each microarray and used for thresholding (determining number of transcripts present). Also, a ubiquitin normalizer serially diluted (50 pg, 5 pg, 500 fg, 50 fg, 5 fg, 0.05 fg and 0.005 fg) was applied to the array. Spot morphology was assessed by visual inspection, SYBR® Green 1 (Molecular Probes) staining or hybridization with labeled non-specific probe. To check clone tracking, 47 high quality sequences were obtained from randomly-selected wells of the cleaned, consolidated 384-well plates used for microarray printing. Each tracked clone had BLAST identifiers matching gene IDs predicted from the re-array spreadsheet, indicating highly accurate clone tracking throughout the process of microarray fabrication.
Animals
Various tissues (brain, kidney, midgut, spleen, ovary, testis, muscle) were sampled from two three-year-old AS (S. salar) adults (Pacific Biological Station, Nanaimo, B.C.). Livers were obtained from several 2.5 year-old AS (McConnell strain) and chinook salmon (O. tshawytscha) subadults (Fisheries and Oceans Canada, West Vancouver, B.C.). RT (O. mykiss) tissues (Spring Valley Strain) were obtained from Mountain Trout Sales (Sooke, B.C.). LW (C. clupeaformis) livers were obtained from three-year-old animals (Laboratoire Bernatchez, Université Laval, Quebec) and rainbow smelt (Osmerus mordax) livers were obtained from adult smelt (NRC Institute for Marine Biosciences and Memorial University of Newfoundland). Each institution that provided tissue, raised and treated the fish in compliance with ethics committee or government body guidelines.
Tissue and RNA extraction
Fish were exsanguinated for several minutes. The tissues were removed and flash frozen in liquid nitrogen and stored at -80°C until RNA extraction. Flash frozen tissues were ground using baked (220°C, 5 h) mortars and pestles under liquid N2, then total RNA was extracted in TRIzol reagent (Invitrogen). RNAs obtained from these preparations were used for generating labeled targets for microarray hybridizations.
Microarray hybridizations
The microarray experiments were designed to comply with MIAME guidelines [34]. To minimize technical variability, all targets were synthesized in one round and each hybridization experiment was conducted simultaneously on slides from a single batch where possible. Each hybridization experiment included dye-flips to compensate for cyanine fluor effects. Total RNA samples were quantified and quality-checked by spectrophotometer and agarose gel, respectively.
All hybridization experiments were performed using the SuperScript Indirect cDNA Labeling System kit and instructions (Invitrogen). Briefly, 5.0 ug total RNA was reverse transcribed using an anchored oligo d(T)20 primer in cDNA synthesis reactions that incorporated aminoallyl- and aminohexyl-modified nucleotides. The modified cDNAs were then labeled with fluorescent Cy5 or Cy3 dye in reactions with the amino-functional groups in coupling buffer.
BAC DNA preparation
Previously sequenced Atlantic salmon BAC 92I04 obtained from the Children's Hospital Oakland Research Institute Atlantic Salmon BAC library (CHORI – 214) was isolated and purified. A total of 30 ug BAC DNA was added to shearing buffer containing 10 mM Tris-HCl pH 7.5, 1 mM EDTA and 20% glycerol. The DNA was sheared into fragments to a concentrated mass of 1500 bp by nebulization in an Invitrogen nebulizer (Cat# 45-0071) at 30 psi of N2 and concentrated by ethanol precipitation.
A total of 5 ug of nebulized BAC DNA was combined with 7.5 ug of pd(N)6 random hexamers (Amersham Biosciences), heated to 100°C for 5 minutes and then cooled on ice for 5 minutes. BAC fragment probes were then generated using Klenow Fragment DNA Polymerase (exo-) (New England Biolabs) in the presence of amino-modified nucleotides (Invitrogen) and labeled with fluorescent Cy3 dye in coupling buffer (see above). Before hybridizations, labeled BAC with Cot-1 salmon DNA was heated to 100°C for 15 minutes, placed on ice for 5 minutes, then warmed to 37°C; labeled BAC without Cot-1 salmon DNA was heated to 80°C for 10 minutes and then cooled to 65°C, before application of treated BAC to microarrays (see below).
Microarray preparation
All microarrays were prepared for hybridization by washing 2 X 5 min in 0.1% SDS, washing 5 X 1 min in MilliQ H2O, immersing 3 min in 95°C MilliQ H2O, and drying by centrifugation (5 min 2000 rpm in 50 ml conical tube). All slides were prehybridized in 5 X SSC, 0.1% SDS, 0.5% BSA for 1.5 h at 49°C. Arrays were briefly washed 2 X 20 sec in MilliQ H2O, then dried by centrifugation. Labeled DNAs were hybridized to prewarmed microarrays in a formamide based buffer (25% formamide, 4X SSC, 0.5% SDS, 2X Denhardt's solution) 16 h at 49°C. The arrays were washed 1 X 10 min in 49°C (2X SSC, 0.1% SDS), and then 2 X 5 min in (2X SSC, 0.1% SDS), 2 X 5 min in 1X SSC and 2 X 5 min in 0.1X SSC at room temperature, then dried by centrifugation.
Microarray analyses
Fluorescent images of hybridized arrays were acquired immediately at 10 um resolution using ScanArray Express (PerkinElmer). The Cy3 and Cy5 cyanine fluors were excited at 543 nm and 633 nm, respectively, at the same laser power (90%), with adjusted photomultiplier tube settings between slides to balance the Cy5 and Cy3 channels. Fluorescent intensity data was extracted from TIFF images using Imagene 5.5 software (Biodiscovery). Quality statistics were compiled in Excel from raw Imagene fluorescence intensity report files. Features were sorted (16,006 salmonid spots each representing different cDNAs; 24 Arabidopsis spots representing 6 different cDNAs) and median signal values and mean numbers of salmonid features passing threshold were determined for Cy3 and Cy5 data separately.
For cross-species and tissue-on-tissue experiments, the hybridization performance of labeled targets to salmonid features was assessed as a percentage of features bound from the numbers of AS and RT features passing a hybridization signal threshold, defined as two standard deviations above Arabidopsis signal mean. No transformations or normalizations were performed on these data. Only features deemed present by Imagene 5.6.1 (excluding marginal and absent values) were used for analyses. We also analyzed some of these data at two standard deviations above empty spot mean signal intensity and found that this was a less stringent method of thresholding (data not shown).
Intraspecific liver and BAC hybridization data analysis (background correction, Lowess normalization, and fold change gene list formation) was performed in GeneSpring 6.1 (Silicon Genetics). All scanned microarray TIFF images, extracted ImaGene grid files, the gene identification file and ImaGene quantified data files are available on-line as supplemental data [35]. The data is deposited in NCBI's GEO repository under PLATFORM GPL 2716 [36].
Authors' contributions
KRVS: co-creator of cDNA libraries; participated in coordination of project; performed organ complexity and intraspecific hybridization experiments; drafted the manuscript.
MLR: constructed and participated in the characterization of high-complexity cDNA libraries, and participated in the development of selection criteria for genes included on the 16006-gene microarray.
GAC: development and optimization of hybridization protocols; performed various species and BAC labeling hybridization experiments; data analysis for manuscript.
GDB: implemented the sequence processing pipeline (from screening and trimming reads to BLAST-identifying assembled contigs); contributed to the set of criteria for establishing that a read was not a chimera, and assisted in other computational aspects of the project.
ARG: contributed to gene selection and identification protocols and general informatics.
CCN: participated in the design and manufacture of the microarray.
WSD and BFK conceived of the study and supervised its design and coordination.
All authors read and approved the final manuscript.
Acknowledgements
This research was supported by Genome Canada, Genome BC, and the Province of BC, and additionally by the Natural Sciences and Engineering Research Council of Canada (BFK, WSD). We thank Jeff Zeznik, Bob Shukin and Bruce Dangerfield for microarray processing and printing (Array Facility, Prostate Centre, Vancouver General Hospital). We also would like to thank Robert Devlin, Dionne Sakhrani and Nicole Hofs (Fisheries and Oceans Canada, West Vancouver, B.C., CA) for chinook salmon and AS tissues; Simon Jones and Kim Taylor (Pacific Biological Station, Nanaimo, B.C., CA) for AS tissues; Jack and Kevin Nickolichuk (Mountain Trout Sales, Sooke, B.C., CA) for RT tissues; Robert Saint-Laurent (Laboratoire Bernatchez, University of Laval, Quebec, CA) for LW livers and Clayton Williams, Vanya Ewart (NRC Institute for Marine Biosciences, Nova Scotia, CA) and Connie Short (Ocean Sciences Center, Memorial University of Newfoundland, Newfoundland, CA) for rainbow smelt tissues.
==== Refs
Thorgaard GH Bailey GS Williams D Buhler DR Kaattari SL Ristow SS Hansen JD Winton JR Bartholomew JL Nagler JJ Walsh PJ Vijayan MM Devlin RH Hardy RW Overturf KE Young WP Robison BD Rexroad C Palti Y Status and opportunities for genomics research with rainbow trout Comp Biochem Physiol B Biochem Mol Biol 2002 133 609 646 12470823 10.1016/S1096-4959(02)00167-7
Katchamart S Miranda CL Henderson MC Pereira CB Buhler DR Effect of xenoestrogen exposure on the expression of cytochrome P450 isoforms in rainbow trout liver Environ Toxicol Chem 2002 21 2445 2451 12389925 10.1897/1551-5028(2002)021<2445:EOXEOT>2.0.CO;2
William DE Bailey GS Reddy A Hendricks JD Oganesian A Orner GA Pereira CB Swenberg JA The rainbow trout (Oncorhynchus mykiss) tumor model:recent applications in low-dose exposures to tumor initiators and promoters Toxicol Pathol 2003 31 58 61 12597433
Shum BP Guethlein L Flodin LR Adkison MA Hedrick RP Nehring RB Stet RJM Secombes C Parham P Modes of salmonid MHC class I and II evolution differ from the primate paradigm J Immunol 2001 166 3297 3308 11207285
Basu N Todgham AE Ackerman PA Bibeau MR Nakano K Schulte PM Iwama GK Heat shock protein genes and their functional significance in fish Gene 2002 295 173 183 12354651 10.1016/S0378-1119(02)00687-X
Zhang C Brown SB Hara TJ Biochemical and physiological evidence that bile acids produced and released by lake char (Salvelinus namaycush) function as chemical signals J Comp Physiol [B] 2001 171 161 171 11302533
Faillace MP Julian D Korenbrot JI Mitotic activation of proliferative cells in the inner nuclear layer of the mature fish retina: regulatory signals and molecular markers J Comp Neurol 2002 451 127 141 12209832 10.1002/cne.10333
Tipsmarck CK Madsen SS Seidelin M Christensen AS Cutler CP Cramb G Dynamics of Na(+), K(+), 2Cl(-) cotransporter and Na(+), K(+)-ATPase expression in the branchial epithelium of brown trout (Salmo trutta) and Atlantic salmon (Salmo salar) J Exp Zool 2002 293 106 118 12115907 10.1002/jez.10118
Devlin RH Biagi CA Yesaki TY Smailus DE Byatt JC Growth of domesticated transgenic fish Nature 2001 409 781 782 11236982 10.1038/35057314
Madigou T Uzbekova S Lareyre JJ Kah O Two messenger RNA isoforms of the gonadotrophin-releasing hormone receptor, generated by alternative splicing and/or promoter usage, are differentially expressed in rainbow trout gonads during gametogenesis Mol Reprod Dev 2002 63 151 160 12203824 10.1002/mrd.90006
Rise ML von Schalburg KR Brown GD Mawer MA Devlin RH Kuipers N Busby M Beetz-Sargent M Alberto R Gibbs AR Hunt P Shukin R Zeznik JA Nelson C Jones SR Smailus DE Jones SJ Schein JE Marra MA Butterfield YS Stott JM Ng SH Davidson WS Koop BF Development and application of a salmonid EST database and cDNA microarray: data mining and interspecific hybridization characteristics Genome Res 2004 14 478 490 14962987 10.1101/gr.1687304
von Schalburg KR Rise ML Brown GD Davidson WS Koop BF A comprehensive survey of the genes involved in maturation and development of the rainbow trout ovary Biol Reprod 2005 72 687 699 15496514 10.1095/biolreprod.104.034967
Rise ML Jones SR Brown GD von Schalburg KR Davidson WS Koop BF Microarray analyses identify molecular biomarkers of Atlantic salmon macrophage and hematopoietic kidney response to Piscirickettsia salmonis infection Physiol Genomics 2004 20 21 35 15454580 10.1152/physiolgenomics.00036.2004
Abstract
Abstract
Abstract
Ishiguro NB Miya M Nishida M Basal euteleostean relationships: a mitogenomic perspective on the phylogenetic reality of the " Protacanthopterygii" Mol Phylogenet Evol 2003 27 476 488 12742752 10.1016/S1055-7903(02)00418-9
Allendorf FW Thorgaard GH Turner BJ Tetraploidy and the evolution of salmonid fishes Evolutionary Genetics of Fishes 1984 New York: Plenum Press 1 53
Ohno S Wolf U Atkin NB Evolution from fish to mammals by gene duplication Hereditas 1968 59 169 187 5662632
Gilbert SF Control of development by RNA processing Developmental Biology 1988 second Sinauer Associates, Inc. Sunderland, Massachusetts 441 464
Kawasawa Y McKenzie LM Hill DP Bono H RIKEN GER Group and GSL Members Yanagisawa M G protein-coupled receptor genes in the FANTOM2 database Genome Res 2003 13 1466 1477 12819145 10.1101/gr.1087603
Kumar RS Ijiri S Kight K Swanson P Dittman A Alok D Zohar Y Trant JM Cloning and functional expression of a thyrotropin receptor from the gonads of a vertebrate (bony fish): potential thyroid-independent role for thyrotropin in reproduction Mol Cell Endocrinol 2000 167 1 9 11000515 10.1016/S0303-7207(00)00304-X
Yang YH Dudoit S Luu P Lin DM Peng V Ngai J Speed TP Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation Nucleic Acids Res 2002 30 e15 11842121 10.1093/nar/30.4.e15
Jin W Riley RM Wolfinger RD White KP Passador-Gurgel G Gibson G The contribution of sex, genotype and age to transcriptional variance in Drosophila melanogaster Nat Genet 2001 29 389 395 11726925 10.1038/ng766
Tseng GC Oh MK Rohlin L Liao JC Wong WH Issues in cDNA microarray analysis: quality filtering, channel normalization, models of variations and assessment of gene effects Nucleic Acids Res 2001 29 2549 2557 11410663 10.1093/nar/29.12.2549
Furano AV Duvernell DD Boissinot S L1 (LINE-1) retrotransposon diversity differs dramatically between mammals and fish Trends in Genetics 2004 20 9 14 14698614 10.1016/j.tig.2003.11.006
Murata S Takasaki N Saitoh M Okada N Determination of the phylogenetic relationships among Pacific salmonids by using short interspersed elements (SINEs) as temporal landmarks of evolution Proc Natl Acad Sci USA 1993 90 6995 6999 8346208
Goodier JL Davidson WS Characterization of a repetitive element detected by NheI in the genomes of Salmo species Genome 1994 37 639 645 7958824
Goodier JL Davidson WS Tc1 transposon-like sequences are widely distributed in salmonids J Mol Biol 1994 241 26 34 8051704 10.1006/jmbi.1994.1470
University of Washington Genome Centre
Altschul SF Madden TL Schäffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
The Institute for Genomic Research
Brazma A Hingamp P Quackenbush J Sherlock G Spellman P Stoeckert C Aach J Ansorge W Ball CA Causton HC Gaasterland T Glenisson P Holstege FC Kim IF Markowitz V Matese JC Parkinson H Robinson A Sarkans U Schulze-Kremer S Stewart J Taylor R Vilo J Vingron M Minimum information about a microarray experiment (MIAME)-toward standards for microarray data Nat Genet 2001 29 365 371 11726920 10.1038/ng1201-365
Genomics Research on Atlantic Salmon Project
National Center for Biotechnology Information Gene Expression Omnibus
|
16164747
|
PMC1239916
|
CC BY
|
2021-01-04 16:32:48
|
no
|
BMC Genomics. 2005 Sep 15; 6:126
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-126
|
oa_comm
|
==== Front
BMC Med Res MethodolBMC Medical Research Methodology1471-2288BioMed Central London 1471-2288-5-281615939710.1186/1471-2288-5-28DebateCausal inference based on counterfactuals Höfler M [email protected] Clinical Psychology and Epidemiology, Max Planck Institute of Psychiatry, Munich, Germany2005 13 9 2005 5 28 28 18 5 2005 13 9 2005 Copyright © 2005 Höfler; licensee BioMed Central Ltd.2005Höfler; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The counterfactual or potential outcome model has become increasingly standard for causal inference in epidemiological and medical studies.
Discussion
This paper provides an overview on the counterfactual and related approaches. A variety of conceptual as well as practical issues when estimating causal effects are reviewed. These include causal interactions, imperfect experiments, adjustment for confounding, time-varying exposures, competing risks and the probability of causation. It is argued that the counterfactual model of causal effects captures the main aspects of causality in health sciences and relates to many statistical procedures.
Summary
Counterfactuals are the basis of causal inference in medicine and epidemiology. Nevertheless, the estimation of counterfactual differences pose several difficulties, primarily in observational studies. These problems, however, reflect fundamental barriers only when learning from observations, and this does not invalidate the counterfactual concept.
==== Body
Background
Almost every empirical research question is causal. Scientists conducting studies in medicine and epidemiology investigate questions like "Which factors cause a certain disease?" or "How does a certain therapy affect the duration and course of disease?" Clearly, not every association is temporarily directed, and not every temporarily directed association involves a causal component but might be due to measurement error, shared prior factors or other bias only. The only sine qua non condition for a causal effect in an individual is the precedence of the factor to its effect, and 100% evidence for causality is impossible. This insight dates back at least to the 18th century Scottish philosopher David Hume [[1]; 2 chap. 1]. The question is how much evidence for a causal effect one can collect in practice and what statistical models can contribute to such evidence.
The history of causal thinking, especially in philosophy, is a history of controversies and misunderstandings. For a detailed description of these controversies, see [[1]; 2, chap. 1; [3,4]]. In this article, I argue that the counterfactual model of causal effects captures most aspects of causality in health sciences. A variety of conceptual as well as practical issues in estimating counterfactual causal effects are discussed.
The article is organized as follows: In the first two sections of the Discussion part, the counterfactual model of causal effects is defined, and some general aspects on statistical inference are discussed. The next chapters provide an overview on causal interactions and causal inference in randomised and nonrandomised studies. In the last two sections, several special topics and related approaches for assessing causal effects are reviewed.
Discussion
1. The counterfactual model of causal effects
Statistics cannot contribute to causal inference unless the factor of interest X and the outcome Y are measurable quantities [3]. The temporal direction can be assessed with substantial knowledge (e.g. gender may effect diet but not vice versa) but substantial knowledge might be uncertain or even wrong. Alternatively, it can be established through the study design. Here, the causal order is ideally guaranteed by a condition in an experiment that has been manipulated before an outcome is measured [5]. If an experiment is not feasible, it is preferable to infer the temporal direction from a prospective design (e.g. a reported traumatic event at the baseline assessment as a potential risk factor for incident depression during the follow-up period) instead of collecting information on the temporal direction retrospectively in a cross-sectional study [[1]; 6 chap. 1]. Generally, in non-experimental studies, measurement error can occur not only in both X and Y but also in the assessment of their temporal direction.
To define a causal effect in an individual i, let us assume that we want to assess the effect of an index treatment or exposure level t (e.g. intake of a specific drug) as compared to another treatment or exposure level c (e.g. no treatment) on an outcome Yi. The outcome can be binary or quantitative (e.g. the amount of segregation of a hormone or a psychological score). According to Greenland and Brumback [7], we basically assume in counterfactual inference that
(a) at the fixed time point of assignment, the individual i could have been assigned to both treatment levels (Xi = t or Xi = c) and
(b) the outcome Yi exists under both Xi = t (denoted by Yi,t) and Xi = c (denoted by Yi,c).
Counterfactuals and potential outcomes
Obviously, the outcome can be observed only (or more precisely, at most) under one, and not under both conditions. If individual i is assigned to treatment level t, then Yi,c is unobservable; likewise, if individual i is assigned to treatment level c, then Yi,t is unobservable. The treatment that individual i actually does not receive is called counterfactual treatment. Likewise, the outcome under this treatment is referred to as counterfactual or potential outcome. The term potential outcome reflects the perspective before the treatment assignment and is more widespread in statistics (e.g. [8]). In contrast, the term counterfactual outcome denotes the perspective after the allocation; it originated in philosophy and has caught on in epidemiology (e.g. [2]). Throughout this paper, I shall use the term counterfactual.
A meaningful counterfactual constitutes a principally possible condition for individual i at the fixed time of assignment. For example, having a certain gynaecological disease instead of not having it would be an odd counterfactual condition for men. As a consequence, "influences" of intrinsic variables like sex, race, age or genotype cannot be examined with counterfactual causality in most contexts [9]. Whether "effects" of such variables should be labelled causal is controversial [7]; see [10,11] for conflicting opinions. If the discussion on causal effects, however, is restricted to those variables that might, at least in principle, be manipulated, this controversy is no longer relevant. Other factors are hardly subject to empirical research and do not serve for intervention.
In general, counterfactuals are quite natural, and, although sometimes claimed [12], there is nothing "esoteric" or "metaphysical" about them. Counterfactual reflections seem to play a vital role in creativity when human beings deal with "what would have happened if" questions [13]. In quantum physics, they have even measurable consequences [14].
Definition of causal effect
There is a causal effect of treatment level t versus treatment level c in individual i at the time where treatment is assigned if the outcomes differs under both conditions [e.g. [15]]:
Yi,t ≠ Yi,c.
The magnitude of the effect can be defined in various ways: for instance, as the difference in the outcome between the two treatment levels:
Yi,t - Yi,c.
If the outcome is strictly positive, one may also use the ratio. The choice of a measure, however, affects the interpretability of a summary of individual effects as the population average effect, and the interpretability of heterogeneity of individual effect magnitudes as causal interaction (see sections 2 and 3).
To imagine a causal effect in a binary outcome suppose that an individual i had a particular disease. After having received a certain treatment (Xi = t), the person no longer has any symptoms of the disease (Yi,t = 0). The question is whether the treatment was the cause of the remission of the disease – in comparison to another treatment level (e.g. Xi = c: "no treatment"). Within the counterfactual conception, this question is equivalent to the one whether the disease would have persisted if the comparison treatment level c had been assigned to the same individual i at the same time, that is, whether Yi,c = 1. According to Maldonado and Greenland [16], this definition of a counterfactual causal effect on a binary outcome dates back to the 18th century when the Scottish philosopher David Hume wrote:
"We may define a cause to be an object followed by another ... where, if the first object had not been, the second never had existed."
Counterfactual causality was the central idea that stimulated invention of randomised experiments by Ronald A. Fisher and statistical inference on them by Fisher around 1920 and, later, by Jerzey Neyman and Egon Pearson in a somewhat different way [3,17]. Much later, in 1974, Rubin [18] has firstly applied the counterfactual model to statistical inference in observational studies.
Choosing the reference treatment
The first difficulty in assessing counterfactual causal effects is to choose the reference condition when comparing one treatment level t with another treatment level c, that is, the substantive meaning of "treatment c". This does not yet constitute a real problem, because researchers should know against what alternative condition the effect of the index treatment is to be evaluated. For instance, in drug treatment trials, the effect of a drug treatment is often examined against that of a placebo treatment (placebo-controlled trial), because an effect resulting from the patient's impression of being treated is not the relevant kind of effect in most cases. On the other hand, if a drug has already been shown to have a positive effect, treatment with this drug may serve in comparing the efficacy of a new drug (drug-controlled trial). Thus, in drug-controlled trials a different effect is estimated than in placebo-controlled trials.
Multiple causal factors and causal mechanisms
In the counterfactual model, a causal factor is a necessary factor without which the outcome (e.g. treatment success) would not have occurred. As the condition is not required to be sufficient for the outcome, multiple causal factors are allowed. This is in line with the fact that the etiology of most physical diseases and almost all mental disorders (e.g. [19]) is multi-causal, resulting from a complex interplay between genetical and environmental factors. Furthermore, a causal effect does not have to be a direct effect. This is desirable because an intervention like drug prescription by a doctor (if the patient complies) often causes an outcome by triggering a whole cascade of consecutive events (of biological, biochemical, mental or social origin), which, in turn, affect the outcome (directly or indirectly). In the causal graph shown in Figure 1, there is no direct effect of X on Y, but X causes Y by affecting Z, which, in turn, influences Y.
Figure 1 Causal graph for an indirect effect of X on Y via Z.
Investigating a causal effect does not require knowing its mechanism. The ability to explain an association, however, often supports the conclusion that it has a causal component (especially if the explanation is given before a researcher looks at the data). The mechanism of an effect is closely related to the terms of effect-modifying and mediating variables. An effect-modifier (or moderator) is neither affected by X nor by Y – but is associated with a "different effect of X on Y" (see section 3); a mediator is affected by X, and, in turn, has an effect on Y.
2. Statistical inference on counterfactual effects
As already mentioned, one can evaluate a fixed individual i at a fixed time only under one condition (Xi = c or Xi = t). Usually, no objective criteria exist to assess with a single observation whether an outcome, such as treatment success (Yi,t = 1) has been caused by the received treatment or by other factors. One exception is ballistic evidence for a bullet stemming from a particular gun and found in a killed person [20] (but here, evidence is still uncertain because the person could have died of sudden coronary failure at the moment the bullet was fired, but this possibility can be checked by autopsy). In the absence of such criteria, one can only estimate average causal effects. This requires several observations, involving different individuals or different time points or both. Many observations are also required for statistically stable conclusions.
Average causal effects
The aim is to estimate the average causal effect, that is, the average of the individual causal effects in the target population. The target population includes all the individuals on whom inference is to be made, whereas the population the sample is actually taken from is the source population [[2]; chap. 20]. Ideally, the source population equals the target population, and the individuals are randomly sampled from that population. If the sample is taken from another than the target population, selection bias will arise if the average causal effect in the source population differs from that in the target population. Moreover, the existence and magnitude of different biases (e.g. confounding [21], see below) depend on the choice of target population, and information on biases stemming from populations other than the target population might not apply.
To be interpreted as an estimate of the population average effect, the difference between the arithmetic mean in X = t versus X = c (summary over all individuals in the respective treatment group) has to equal the arithmetic mean of the differences at the level of the individuals. Linear differences can always be interpreted in this way [22], whereas for multiplicative measures like the mean ratio and the risk ratio the geometric mean has to be used instead. The population average interpretation of the summary odds ratio, however, becomes increasingly false with an increasing number of individuals at high risk for the outcome (under one or both conditions) [22].
The following discussion is restricted to the more frequent case of a sample consisting of different individuals rather than of different time points (or both).
Stable-unit-treatment assumption
Before treatment assignment, there are two random variables for each individual i in the population: the outcome under treatment c (Yi,c) and the outcome under treatment t (Yi,t). Although the theory can be extended accordingly [23], I shall now assume for simplicity that the outcomes of individual i are independent of the outcomes of other individuals and their received treatment levels. This is referred to as the stable-unit-treatment-assumption [23]. Note that this might be a quite restrictive assumption: it does not hold for contagious diseases as outcome. Influenca is such a disease in which the immunisation of certain individuals may affect the others (called "herd effect", e.g. [24]). After treatment assignment and the observation of the outcome, a sample of n individuals contains (at most) one realisation of the outcome for each individual i where the outcome corresponds either to treatment level t or c. Therefore, from the statistical point of view, the estimation of causal effects can be regarded as a particular problem of missing values (e.g. [17]).
Exchangeability
Suppose the average causal effect is defined as the difference in means in the target population between both conditions X = t and X = c. Then the simplest way to estimate it is with the difference between the two sample means (denoted by and , resp.). If individuals with X = c and X = t are "exchangeable", average causal effects can be directly estimated as without bias due to assignment (bias might exist anyway due to other causes such as measurement or selection). Exchangeable means that two conditions have to be fulfilled [21,25]:
a) The distribution of the unobserved outcome Yt under actual treatment c is the same as that of the observed outcome Yt under actual treatment t; that is, under counterfactual treatment with t, the individuals actually treated with c would behave like those actually treated with t; individuals having received treatment t are substitutes for individuals having received treatment c with respect to Yt.
b) The distribution of the unobserved outcome Yc under actual treatment t is the same as that of the observed outcome Yc under actual treatment c; that is, under counterfactual treatment with c, the individuals actually treated with t would behave like those actually treated with c; individuals having received treatment c are substitutes for those who have received treatment t with respect to Yc.
Note that, if individuals actually having received treatment c and t, respectively, correspond to different populations and inference is to be made solely on one of these two populations, then only either assumption a) or b) is required. For instance, if inference is to be made only on the population having received treatment c, condition a) is sufficient.
In the section on causal inference, I will provide an outline on how exchangeability relates to different study designs and what statistical methods can contribute to approach unbiased estimation of causal effects if the optimal design (a perfect randomised experiment) is not feasible.
3. Heterogeneity in causal effects
An important issue is the assessment of differences in causal effects between individuals. Clearly, a necessary condition for a factor Z to be a modifier of the effect of X on Y is that Z precedes the outcome Y. If such a potential effect-modifier Z is associated with X, the parameter that describes the modification of the effect of X on Y is not identified without making further assumptions. Effect-modifiers are typically assessed with interaction terms in regression models.
Choice of the effect measure
Whether and, if yes, to what extent the degree of an effect differs according to the values of Z depends, however, on the choice of the model and the associated index of effect magnitude. As mentioned above, some effect measures (e.g. the odds ratio) usually serve only to quantify the magnitude of a causal effect supposed to be constant between the individuals.
Moreover, the risk difference is the only measure for which effect heterogeneity is logically linked with causal co-action in terms of counterfactual effects. To explain this, it is necessary to define the causal synergy of two binary factors, Xi and Zi (coded as 0 or 1), on a binary outcome Yi in an individual i (at fixed time).
Clearly, if Xi and Zi do not act together in causing the event Yi = 1, then
(a) if Yi = 1 is caused by Xi only,
Yi = 1 if (Xi = 1 and Zi = 0) or
(Xi = 1 and Zi = 1)
and Yi = 0 in all other cases. Thus, Yi = 1 occurs in all cases where Xi = 1 and in no other cases.
(b) if Yi = 1 is caused by Zi only,
Yi = 1 if (Xi = 0 and Zi = 1) or
(Xi = 1 and Zi = 1)
and Yi = 0 in all other cases. Thus, Yi = 1 occurs in all cases where Zi = 1 and in no other cases.
Therefore, causal synergy means that 1) Yi = 1 if either one or both factors are present and 2) Yi = 0 if neither factor is present. Now, one is often interested in superadditive risk differences, where the joint effect of X = 1 and Z = 1 is higher than the sum of the effects of (X = 1 and Z = 0) and (X = 0 and Z = 1) as compared to the risk for Y = 1 under (X = 0 and Z = 0), that is,
P(Y = 1 | X = 1, Z = 1) > P(Y = 1 | X = 1, Z = 0) + P(Y = 1 | X = 0, Z = 1) - P(Y = 1 | X = 0, Z = 0).
If superadditivity is present, one can show that there must be causal synergy between X and Z on Y, at least for some individuals [[2], chap. 18; [26,27]]. This relation does not apply in the opposite direction: If there is causal synergy among some individuals there may be no superadditivity. Thus, one can demonstrate rather a causal interaction than its non-existence. Note that other logical relations do not exist and the risk difference is the only measure for which such a logical link exists [[2]; chap. 18; [26,27]]. Also, other measures like correlations, standardised mean differences or the fraction of explained variability do not serve to quantify the degree of causal effects because they mix up the herefore solely relevant mean difference with parameters of exposure and outcome variability [28].
Another crucial point for the choice of effect index is whether the interaction terms in regression models corresponds with so-called mechanism-based (e.g. biological) interactions [29]. For instance, if the dose of intake of a particular drug is known to influence the release of a certain hormone linearly, then the interaction term of another factor with drug intake in a linear model corresponds to the presence of a biological interaction.
Deterministic versus probabilistic causality
A fundamental question relating to heterogeneity in causal effects is the distinction between deterministic and probabilistic causality [[2], chap. 1; [30], chap. 1]. The functional-deterministic understanding of causality is based on the Laplacian conception of natural phenomena, which are assumed to follow universally valid natural laws. Here, in the absence of measurement error and other biases, the observable heterogeneity in Y – given X and the other observed covariates – would be attributed solely to unobserved factors. If we knew the causal mechanism completely (how complicated it may be) and the values of all the causal factors, the outcome Y would be exactly determined. Note that I have implicitly used this assumption in the previous discussions.
Within the probabilistic understanding of causality, individual variation exists within the outcome Y, which can not be explained by unconsidered factors. This variation might be called real randomness and can be found in quantum physics [14]. It is possible to incorporate real randomness into counterfactual models because one can specify a probability distribution for a potential outcome of a fixed individual at a fixed time [[7] and references therein]. In real situations, however, the distinction between deterministic and probabilistic causality does not play a major role in systems that are complex enough for substantial residual heterogeneity in the modelled effect to be expected. Here, the effect is practically probabilistic. Such a situation is rather the rule than the exception in medical and behavioural sciences.
On the other hand, after incorporating major effect-modifiers into a model, the effect of X on Y should be sufficiently homogeneous to allow for uniform interventions in the subpopulations defined by the values of the effect-modifiers. As a consequence of the existence of effect-modifiers, a variation in their distribution across different populations implies that one would expect to estimate different effects if the modifiers were not considered in a model. Thus, differences in estimates of effects do not imply that different causal mechanisms act; instead, they might be solely due to different distributions of hidden effect-modifiers [[2], chap. 18; [16]]. Interactions with intrinsic variables; that is, individuals' immutable properties like sex, race and birth date are often regarded as an indication of a narrow scope of a model [31]. On the other hand and as mentioned above, nonmanipulable properties are hardly subject to counterfactual arguments.
4. Causal inference in randomised and non-randomised studies
Randomised experiments
As already mentioned, if the individuals are exchangeable between the treatments and there are no other biases, causal effects can be directly estimated, most simply with the difference in the mean of Y between X = c and X = t. A stronger assumption than exchangeability is related to the propensity score. The propensity score is the probability of individual i being assigned to treatment t – at the time when group assignment to X = c or X = t takes place, denoted with PSi = P (Xi = t). The assumption that the propensity score is equal among the individuals with X = c or X = t is stronger than the assumption of exchangeability because the determinants of the propensity score do not necessarily affect the outcome Y.
In a simple randomised experiment, PSi is equal for all individuals. For example, in an experiment with balanced groups, the individuals are assigned to each treatment with a probability of 50%: PSi = 1/2 for all i. More sophisticated designs incorporate a covariate or, more generally, a vector of covariates Zinto the group assignment (block designs). Provided that such covariates are also factors of the outcome, considering them often yields increased statistical precision in the estimate of the causal effect. In randomised experiments, the propensity score is a known function g of the realisations zof Z; that is, PSi = g(zi) and the joint distribution of Yc and Yt is conditionally independent of X given Z, a property called strong ignorability [8]. Now, one can show that X and Zare conditionally independent given the propensity score; that is, the propensity score PS summarises all information contained in Zabout the group assignment [8]. As a consequence, the mean effect of X on Y can be approximatively estimated without bias due to assignment if the entities are matched pairwise according to the propensity score, if they are weighted proportionally to the inverse propensity score, or if the propensity score is adjusted for in a suitable regression model [8]. From a Bayesian perspective, the estimates of the propensity score are posterior probabilities to predict the allocation to exposure (X = t) under Z = z [32]. The problem with the propensity score is that it is sufficient to control for but not minimally sufficient (it may include unnecessary information due to covariates related to Y but not to X).
Imperfect experiments
In the discussion above I have implicitly assumed that treatment and control protocols were followed exactly; in that sense, the experiments were supposed to be perfect. In many studies, however, the actual treatment and control conditions do not equal the intended protocols, at least, not for some individuals or measurement points (imperfect or broken experiments). For instance, in the pharmacotherapy of depression with antidepressants, one often faces the problem that many individuals in the antidepressant treatment group (X = t) stop drug intake as, in the beginning, they experience only adverse effects [33]. According to Imbens and Rubin [34], imperfect experiments constitute the bridge between experiments with ideal compliance and observational studies.
Instrumental variables
If one ignores the fact that the treatment conditions were not exactly followed, one estimates the effect of the intended, not of the actual treatment. This is referred to as intent-to-treat analysis. Alternatively, one can estimate the effect of the treatment among those who complied. This can be done with approaches based on instrumental variables. Roughly speaking, an instrumental variable I is a variable that is associated with the actual treatment or exposure X but that related to the outcome Y only through its association with X. Maybe the most important example for an instrumental variable is the intended treatment. The basic idea of such approaches is that one can – under certain conditions that vary with the specific problem – compute the X - Y association or bounds of it from the I - X and the I - Y association [35,36]. These methods are useful when the observed X - Y association is more confounded than the I - X and the I - Y associations. Another situation where instrumental variable methods apply is when not X but only a surrogate I of it can be directly observed. The association between I and X then has to be known or estimable, and differences between I and X have to be independent of other variables [35,36].
Observational studies
Not every interesting factor can be translated into equivalent lab settings or can be manipulated. Factors like social support or peer relationships are difficult to observe outside their natural environment. Other conditions should not be assigned to human beings for ethical reasons (e.g. smoking). In such cases, there is no way but to conduct an observational study. In observational studies, the group assignment is neither manipulated nor randomised. The group status X is a random variable subject to measurement error, and the individuals assign themselves to X = c or X = t, for example, by deciding to smoke or not to smoke.
The propensity score then typically depends on a variety of variables (denoted as vector Z). Often, not all of such factors are observable or even known. Researchers conducting epidemiological and nonrandomised clinical studies should aim at collecting data on the major determinants of X to allow for an adequate control of confounding. Note that variables associated with X but not with Y can often be ignored. However, they can serve to reduce the variance in normally distributed outcomes, but adjusting for them sometimes yields unnecessarily high variances in outcomes with other distributions [8].
In many practical situations, one should assume substantial residual bias due to unobserved determinants of the exposure X, which, in turn, affect Y. Such kind of bias is referred to as confounding. A confounder is a variable that is associated with both X and Y and that precedes X; and adjusting for it reduces the overall bias in the estimation of the causal effect of X on Y [[2], chap. 15]. In practice, however, it is not determinable whether a certain variable is a confounder because this depends on all (other) confounders and biases together. If Zl is a candidate for a confounder, the difference between the means under X = t and X = c adjusted for Zl might be biased more strongly than the unadjusted mean difference. This can happen, for instance, if other, more important factors of group assignment are distributed more unequally across X = c und X = t after stratification than they were before stratification on Zl [[7], and the references therein].
Pearl [[30], chap. 3; [37]] has discovered formal criteria within the framework of graphical models (the "backdoor" and "frontdoor" criterion, resp.) that indicate which set of covariates is sufficient to be controlled for. Applying these criteria, however, requires assumptions on the causal system that causes X and Y. Some of the variables that cause X and Y, in turn, are often unobserved or even unknown.
Methods to adjust for unobserved confounding and other biases
There are various approaches to address unobserved confounding, bias due to measurement error, selection, and other biases. The first method is sensitivity analysis, which examines what impact one or several supposed scenarios of bias would have had on the results at hand. The results depend on the presumed values of bias parameters like misclassification probabilities, the distribution of a confounder, and the magnitude of it's effects on X and Y. For a general model for sensitivity analyses, see [38]. Rosenbaum [39] has proposed a general framework to assess how sensitive a particular study design is against assignment bias. The problem with sensitivity analysis is that only the range of expected results under different specified values for the unknown bias parameters is revealed [40].
This drawback is solved with Monte Carlo sensitivity analysis. Here, distributions are assigned to the unknown bias parameters, which reflect a researcher's knowledge or assumptions about their true values. Bias-corrected point and interval estimates can then be calculated. The results from these methods have approximatively a Bayesian interpretation if additional uncertainty is added (as would be the case if one drew random numbers from the posterior distribution of the unknown effect), the estimator of the causal effect is approximatively efficient, and the data provide no information on the bias parameters [[40] and references therein].
(Monte Carlo) sensitivity analyses and Bayesian methods outperform conventional analyses, which often yield overconfident and biased results because they are based on wrong point priors at zero (e.g. misclassification probabilities) at the parameters determining bias [40]. This is true as long as the assumptions made are not fundamentally wrong (e.g. bias downward instead of upward, [41]). In conventional analyses, the farther the left boundary is from the null, the more room there is for bias and extra-variation. Moreover, a statistically significant difference does not imply that the association found is strong enough to be of a clinical or policy concern; the absence of a statistically significant association often does not even rule out a strong relation (e.g. [[2], chap. 12; [42]]). Hence, it is essential to quantify the degree of association also in perfect randomised experiments and to report an interval estimate.
5. Some more special issues
Time-varying exposures
In many applications, the exposure level X is not a constant condition but a sequence of treatment levels (generalised or g-treatment) that varies within individuals over time. For instance, Robins et al. [43] have investigated the effect of prophylaxis therapy for pneumocystis carinii pneumonia (PCP, an opportunistic infection in AIDS patients) on survival times among AIDS patients in an uncontrolled study. In medical studies, the exposure level often varies over time, for example, because physical complications require a change in treatment or because individuals deposit drug intake because of adverse effects.
The problem with time-varying systems is that they are subject to feedback mechanisms: The causes at fixed time q might not only be affected by causes of the outcome occurring before time q (confounding), but they may also impact later time-dependent causes [44,45]. For instance, the outcome at time q-1 may be a mediator for the outcome at time q, but a confounder of the exposure at time q. As in the above example statistical inference for causal effects of time-dependent exposures is often based on survival time as outcome and, therefore, on survival models. The associated methods easily become complicated because one often has to take several issues into account. These include measured and unmeasured confounder adjustment, feedback mechanisms and censoring (not all individuals are observed throughout the whole investigation time). Then they still share all the limitations of conventional methods in observational studies (bias due to measurement, selection etc., [45]).
Details of statistical models are rather technical and thus beyond the scope of this paper. Briefly, Robins [44] has derived a general recursive g-computation algorithm, from which he has derived non-parametrical tests. These tests, however, turned out to be impractical for inference more sophisticated than simple null hypothesis testing (e.g. [45]). Later, more flexible semiparametric models (called g-estimation) of survival outcomes were developed (e.g. [43]). These models make assumptions merely on the form of the difference between the treatment levels rather than on the shape of the outcome distribution within the same treatment (and covariate) level. An alternative approach is provided by so-called "marginal structural models" and "inverse-probability of treatment-weighted estimators". In the case of censoring, these methods are less complex than g-estimation at the cost of requiring stronger assumptions here [45]. However, they often allow for improved confounder adjustment [46]. Gill and Robins [47] have developed extensions of g-estimation for continuous time.
Competing risks
Suppose that one is interested in the health burden attributable to a variable that is actually an outcome and not a treatment action in the earlier sense. Let me borrow an example from Greenland [48]: Suppose one is interested in how the number of years lived after the age of 50 (T) is affected by whether smokers died of cancer (Y = 1) or not (Y = 0). Assume that a certain individual i was a male lifetime heavy smoker and died from lung cancer at the age of 54 (Ti = 4 | Yi = 1). Now the estimation of Ti under Yi = 0 is unclear because it depends on how Yi = 0 was caused, how death from lung cancer was prevented. If death by lung cancer had been prevented through convincing the individual not to smoke at all in his entire lifetime, then the risk of other causes of death (e.g. coronary heart disease, diabetes, or other kinds of cancer) would be lower as well. In this case, Ti under Yi = 0 might be considerably higher than 4 years. On the other hand, if Y = 0 was caused by chemotherapy, the risks of other diseases, named competing risks [[48] and references therein] would not have been reduced. Hence, the outcome Ti under Yi = 0 might not have been much higher here than under Yi = 1. The expected increase in years lived would thus be much smaller if lung cancer was prevented by chemotherapy than it would be if lung cancer was prevented by lifetime absence of smoking.
To conclude, there is no single intervention in this case that would be independent of an individual's history prior to exposure. The evaluation of the effect of removing Yi = 1 depends on the mode of removal in a multivariate framework. Therefore, effects of policies should be evaluated in terms of actions that cause outcome removal rather than in terms of outcome removal per se [48].
The probability of causation
A common problem is how to determine the probability that an event in an individual has been caused by a certain exposure, that is, the probability of causation (PC). Courts define causation as an exposure without which the outcome event would a) not have happened at all or b) have happened later. Such a cause is named contributory cause [49]. The empirical basis for an estimate of the probability of causation in an individual is a sample of exposed individuals. This sample should be similar to the individual under investigation with respect to the history of exposure and (other) risk factors of disease. Then, one can estimate the rate fraction (RF – often called "attributable fraction"), the excess incidence rate due to exposure – relative to the incidence rate if exposed, given by
where IRX = 0 and IRX = 1 denote the incidence rates in the target population under exposure and under non-exposure, respectively [[2]; chap. 3]. The etiological fraction (EF) is defined as the fraction of exposed individuals with the disease for which the exposure was a contributory cause of the disease [[2]; chap. 3]. Now, the probability of causation in the individual equals the etiological fraction (PC = EF) if the individual was randomly drawn from the target population [49]. A common fallacy, however, is to confuse the rate fraction RF with the probability of causation PC (in the sense of a contributory cause). To illustrate this mistake algebraically, one can express the etiological fraction as
where:
- C1 is the number of individuals in the population in which exposure has caused an accelerated onset of disease (i.e., under non-exposure, the disease would have occurred anyway but later);
- C2 denotes the number of individuals in whom exposure has caused all-or-none disease (i.e., without exposure, these persons would not have contracted the disease at all); and
- CT is the total number of persons exposed to the disease (including also those individuals who have not been affected by the exposure, [49]).
Now, one can show that, if the probability of the exposure having an effect in the exposed is low, the rate fraction RF approximately equals A2/AT [49] – a quantity known as the excess rate [[2], chap. 4]. Thus in this case, the equation PC = RF approximatively holds only if A1 is small as compared to A2. This means that the effect is required to have an all-or-none effect in the vast majority of exposed and diseased individuals. Otherwise, the probability of causation is underestimated proportionally to the ratio A1/A2. A fundamental problem with the estimation of PC is the estimation of A1 – the number of exposed and diseased persons who would have developed the disease later under non-exposure. This estimation would require some biological model (which seems to be rarely available) for the progress of the disease [49]. Robins and Greenland [50] have provided upper and lower limits for the probability of causation that are consistent with the data. Pearl [51] showed under which conditions the probabilities that a factor was a necessary or a sufficient cause, respectively, can be estimated from the data.
6. Related approaches to causal inference
The sufficient-component-cause model
Rothman [52] has proposed a model of causal effects that is similar to but finer than the counterfactual model — the sufficient-component-cause model. Entities in this model are not individuals but mechanisms of causation. A mechanism is defined as a combination of factors that are jointly sufficient to induce a binary outcome event, Y = 1. Each of possibly many of such mechanisms has to be minimally sufficient: The omission of one factor would change Y from 1 to 0; that is, the outcome event would no longer be present. For instance, following an example by Rothman [52], it is not sufficient to drink contaminated water to get cholera; other factors are required as well. If, in this example, drinking contaminated water is part of each mechanism that leads to cholera, this constitutes a necessary factor for cholera. For a fixed individual at a fixed time, often several mechanisms are in line with the same counterfactual effect [[2], chap. 18; [7]]. Therefore, the sufficient-component-cause model is important rather for conceptional than for inferential considerations. Rothman's [52] intention was to build a bridge between metaphysical reflections and epidemiological studies.
Structural equation models
Especially in the fields of psychology, social sciences and economics, structural equation models (SEMs) with latent variables are frequently used for causal modelling. These models consist of (a) parameters for the relations among the latent variables, (b) parameters for the relations among latent and observed variables and (c) distributional parameters for the error terms within the equations. Pearl [30] has shown that certain nonparametric SEMs are logically equivalent to counterfactual models and has demonstrated how they can be regarded as a "language" for interventions in a system. Furthermore, these models are useful to structure and reduce variance, for example, to reduce measurement error if several items on a questionnaire are assumed to represent a common dimension.
There are, however, several practical problems with the use of SEMs. First, in an under-determined system of equations, several assumptions are necessary to identify the parameters (i.e. to make the estimates unique). In psychological applications, the assumptions tend to be justified only partially [53] and models with alternative assumptions are often not considered [54]. The results, on the other hand, may be very sensitive against these assumptions [55], and currently, there is no way to model uncertainty in these assumptions. Besides, the coefficients from these models are sometimes not interpretable as measures of conditional dependencies (i.e. regression coefficients), for instance, if there are loops in a model [56]. Finally, the meaning of the latent variables remains sometimes obscure, and — in economic applications — results from certain structural equation models have been found to fail to recur in experiments [57].
It is therefore recommended that one should be extremely careful in the application of SEMs. For more sophisticated discussions of the relations among structural equation models, graphical models, the corresponding causal diagrams and counterfactual causality; see [7,30,31,58] and the papers cited therein.
The controversy on counterfactual causality raised by Dawid's article [12]
Dawid [12] has argued that counterfactuals were something metaphysical because causal inference based on counterfactuals would depend on unobservable assumptions. In his own formulation of the counterfactual model, Dawid assumed that a causal effect in an individual was composed of the average effect of treatment t versus c, an individual effect and an interaction term treatment*individual. Different assumptions about the unidentified individual parameters would yield different conclusions about the variance of the counterfactual effect. Such assumptions involved the joint distribution of Yc and Yt for fixed individuals.
Together with Dawid's paper in the Journal of the American Statistical Association, not less than seven commentaries as well as Dawid's rejoinder [59] were published. Cox [60] reproached Dawid for posing too general a question and for going much too far with his conclusions: The proof of a causal effect would not require knowing its mechanism. Shafer [61], on the other hand, regretted that David had been too mild in condemning counterfactuals. Casella and Schwarz [62] mentioned that every scientific investigation had to aggregate over different individuals. Pearl [63] and Cox [60] argued that, in contrast to Dawid's claims, several aspects of counterfactual causality were at least indirectly testable. Wasserman [64] pointed out that, as in every other kind of statistical models, the identifiability of parameters would be essential in causal models but that counterfactuals provided a quite useful conception. Robins and Greenland [65] brought up the point that Dawid had largely neglected observational studies and imperfect experiments. Probabilistic causal inference (of which Dawid is an advocate) in observational studies would inevitably require counterfactuals. Otherwise, causal effects may not be identified without again making unidentified assumptions. Rubin [66] considered the modelling of the joint distribution of Yc and Yt as not always necessary.
Dawid [12] rejects the counterfactual concept seemingly because, on it's own, it is not powerful enough to solve the fundamental problems of causal inference (e.g. in a fixed individual at a fixed time one can observe the outcome only under one condition). Depending on the question and the design, there are indeed often unidentified parameters. I argue that the fact that the concept does not solve all problems does not mean that it is wrong; in that sense, denying the usefulness of counterfactuals is as if a doctor never prescribed a drug that may not remedy all his patients, but several of them. Counterfactual causal thinking is based on imagining the consequences of changing the value of a single factor in a comprehensive causal system. What would the world look like after changing the value of one variable (in one or several individuals) is what some philosophers of science call the possible worlds concept of causality [15] (see also [[30], chap. 7] for a formal definition). Our imagination of possible worlds, however, always depends on substantive knowledge required to formulate a causal system that might have produced the data that one has observed. This, though, does not mean that we should not ask for properties of possible worlds because the decisions we aim to conduct (e.g. which interventions to make) depend on these unknown properties.
Summary
1. The counterfactual concept is the basis of causal thinking in epidemiology and related fields. It provides the framework for many statistical procedures intended to estimate causal effects and demonstrates the limitations of observational data [10].
2. Counterfactual causality has also stimulated the invention of new statistical methods such as g-estimation.
3. The intuitive conception makes the counterfactual approach also quite useful for teaching purposes [65]. This can be exemplified by illustrating the difference among study designs. For instance, the benefit of longitudinal over cross-sectional studies is easily demonstrated when the aim is to study how several variables act together over time when causing an outcome.
4. Counterfactual considerations should replace vague conceptions of "real" versus "spurious" association, which occasionally can still be read. In this context, the Yule-Simpson paradox is often mentioned. This paradox indicates that an association can have a different sign (positive or negative association, resp.) in each of two different subpopulations than it has in the entire population. However, if the temporal direction of the variables is added to this paradox and there is no bias and random error, the paradox is resolved: It is then determinable which association is real and which is spurious in a causal sense.
5. Causal effects have been treated like a stepchild for a long time, maybe because many researchers shared the opinion that causality would lie outside what could be scientifically assessed or mathematically formalised. Pearl [30,37] was the first to formulate the difference between changes in variables induced by external intervention in a system and changes due to variation in other variables in the system.
Competing interests
The author(s) declare that they have no competing interests.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
I wish to thank Sander Greenland for very helpful comments on a former version of the manuscript and Evelyn Alvarenga for language editing.
==== Refs
Susser M What is a cause and how do we know one? A grammar for pragmatic epidemiology Am J Epidemiol 1991 7 635 648 2018019
Rothman KJ Greenland S Modern Epidemiology 1998 2 Philadelphia: Lippincott Williams & Wilkins
Holland PW Statistics and causal inference J Amer Stat Ass 1986 81 945 962
Kaufman JS Poole C Looking back on "causal thinking in the health sciences" Annu Rev Public Health 2000 21 101 119 10884948 10.1146/annurev.publhealth.21.1.101
Greenland S Morgenstern H What is directionality? J Clin Epidemiol 1989 42 821 824 2778461 10.1016/0895-4356(89)90092-9
Diggle PJ Heagerty P Liang KY Zeger SL Analysis of longitudinal data 2002 2 Oxford: Oxford University Press
Greenland S Brumback B An overview of relations among causal modelling methods Int J Epidemiol 2002 31 1030 1037 12435780 10.1093/ije/31.5.1030
Rosenbaum PR Rubin DB The central role of the propensity score in observational studies for causal inference Biometrika 1983 70 41 55
Holland PW The false linking of race and causality: lessons from standardized testing Race and Society 2001 4 219 233 10.1016/S1090-9524(03)00011-1
Heckman J The scientific model of causality Sociol Methodol to appear
Susser E Schwartz S Are social causes so different from all other causes? A comment on Sander Greenland Emerging Themes Epidemiol 2005 2 4 10.1186/1742-7622-2-4
Dawid AP Causal thinking without counterfactuals J Amer Stat Ass 2000 95 407 424
Hofstadter DR Metamagical themas: questing for the essence of mind and pattern 1985 New York: Basic Books
Penrose R Shadows of the mind 1994 Oxford: Oxford University Press
Hernan MA A definition of causal effect for epidemiological research J Epidemiol Community Health 2004 58 265 271 15026432 10.1136/jech.2002.006361
Maldonado G Greenland S Estimating causal effects Int J Epidemiol 2002 31 422 29 11980807 10.1093/ije/31.2.422
Little RJA Rubin DB Causal effects in clinical and epidemiological studies via potential outcomes: Concepts and analytical approaches Annu Rev Public Health 2000 21 121 145 10884949 10.1146/annurev.publhealth.21.1.121
Rubin DB Estimating causal effects of treatments in randomised and nonrandomised studies J Educ Psychol 1974 66 688 701
Rutter M The interplay of nature, nurture, and developmental influences Arch Gen Psychiat 2002 59 996 1000 12418932 10.1001/archpsyc.59.11.996
Thali MJ Kneubuehl BP Dirnhofer R Zollinger U Body models in forensic ballistics: reconstruction of a gunshot injury to the chest by bullet fragmentation after shooting through a finger Forensic Sci Int 2001 123 54 57 11731197 10.1016/S0379-0738(01)00519-9
Greenland S Robins JM Identifiability, exchangeability, and epidemiological confounding Int J Epidemiol 1986 15 413 419 3771081
Greenland S Interpreting and choice of effect measures in epidemiological effect measures Amer J Epidemiol 1987 110 693 698 555587
Halloran ME Struchiner CJ Causal inference in infectious diseases Epidemiology 1995 6 142 151 7742400
John TJ Samuel R Herd immunity and herd effect: new insights and definitions Eur J Epidemiol 2000 16 601 606 11078115 10.1023/A:1007626510002
Greenland S Robins JM Pearl J Confounding and collapsibility in causal inference Stat Sci 1999 14 29 46 10.1214/ss/1009211805
Greenland S Poole C Invariants and non-invariants in the concept of interdependent effects Scand J Work Environ Health 1988 14 125 129 3387960
Greenland S Basic problems in interaction assessment Envir Health Perspect 1993 101 59 66
Greenland S Schlessemann JJ Criqui MH The fallacy of employing standardized regression coefficients and correlations as measures of effect Amer J Epidemiol 1986 123 203 208 3946370
Greenland S Elementary models for biological interaction J Hazard Mater 1985 10 449 454 10.1016/0304-3894(85)87015-1
Pearl J Causality – Models, reasoning and inference 2002 Cambridge: Cambridge University Press
Cox DR Wermuth N Multivariate Dependencies Models, Analyses and Interpretation 1996 London: Chapman and Hall
Rubin DB Bayesian-inference for causal effects: the role of randomization Ann Stat 1978 6 34 58
McElroy SL Keck PE Friedman LM Minimizing and managing antidepressant side-effects J Clin Psychiat 1995 56 49 55
Imbens GW Rubin DB Bayesian inference for causal effects in randomised experiments with noncompliance Ann Stat 1997 25 305 327 10.1214/aos/1034276631
Greenland S An introduction to instrumentary variables for epidemiologists Int J Epidemiol 2000 29 722 729 10922351 10.1093/ije/29.4.722
Winship C Morgan S The estimation of causal effects from observational data Ann Rev Sociol 1999 25 659 706 10.1146/annurev.soc.25.1.659
Pearl J Causal diagrams for empirical research Biometrika 1995 82 669 710
Rosenbaum PR Design sensitivity in observational studies Biometrika 2004 91 153 164 10.1093/biomet/91.1.153
Rosenbaum PR Observational studies 2002 2 New York: Springer
Greenland S Multiple bias modelling J Roy Stat Soc A 2005 168 267 306 10.1111/j.1467-985X.2004.00349.x
Greenland S The impact of prior distributions for uncontrolled confounding and response bias: a case study of the relation of wire codes and magnetic fields to childhood leukemia J Amer Stat Ass 2003 98 47 54 10.1198/01621450338861905
Höfler M Statistik in der Epidemiologie psychischer Störungen 2004 Berlin: Springer
Robins JM Blevins D Ritter G Wulfsohn M G-estimation of the effect of prophylaxis therapy for pneumocystis carinii pneumonia on the survival of AIDS patients Epidemiol 1992 3 319 336
Robins JM A graphical approach to the identification and estimation of causal parameters in mortality studies with sustained exposure periods J Chron Diseases 1987 40 139s 161s 3667861
Greenland S Gelman A, Meng X-L An overview of methods for causal inference from observational studies Applied Bayesian methods and causal inference from incomplete data perspectives 2004 Chap 1 New York: Wiley
Robins JM Hernan MA Brumback B Marginal structural models and causal inference in epidemiology Epidemiology 2000 11 550 560 10955408 10.1097/00001648-200009000-00011
Gill RD Robins JM Causal inference for complex longitudinal data: the continuous case Ann Stat 2001 6 1785 1811
Greenland S Murray CJL, Salomon JA, Mathers CD, Lopez AD Causality theory for policy uses of epidemiological measures Summary measures of Population Health 2002 Cambridge, MA: Harvard University Press/WHO 291 302
Greenland S Relation of probability of causation to relative risk and doubling dose: a methological error that has become a social problem Amer J Public Health 1999 89 1166 1169 10432900
Robins J Greenland S The probability of causation under a stochastic model for individual risk Biometrics 1989 45 1125 1138 2611320
Pearl J Probabilities of causation: three counterfactual interpretations and their identifications Synthese 1999 121 93 149 10.1023/A:1005233831499
Rothman KJ Causes Amer J Epidemiol 1976 104 587 592 998606
McDonald RP Ringo Ho MH Principles and practice in reporting structural equation analyses Psychol Meth 2002 7 64 82 10.1037//1082-989X.7.1.64
McCallum RC Austin JT Applications of structural equation modeling in psychological research Annu Rev Psychol 2000 51 201 226 10751970 10.1146/annurev.psych.51.1.201
Stefanski LA Raftery AE, Tanner MA, Wells MT Measurement error models Statistics in the 21st century 2001 London: Chapman & Hall
Cox DR Wermuth N Linear dependencies represented by chain graphs Stat Sci 1993 8 204 218
Lalonde RJ Evaluating the econometric evaluations of training programs with experimental data Amer Econ Rev 1986 76 604 620
Greenland S Pearl J Robins JM Causal diagrams for epidemiological research Epidemiol 1999 10 37 48 10.1097/00001648-199901000-00005
Dawid AP Rejoinder: Causal thinking without counterfactuals J Amer Stat Ass 2000 95 444 448
Cox DR Comment: Causal thinking without counterfactuals J Amer Stat Ass 2000 95 424 425
Shafer G Comment: Causal thinking without counterfactuals J Amer Stat Ass 2000 95 438 442
Casella G Schwartz SP Comment: Causal thinking without counterfactuals J Amer Stat Ass 2000 95 425 427
Pearl J Comment: Causal thinking without counterfactuals J Amer Stat Ass 2000 95 428 431
Wasserman L Comment: Causal thinking without counterfactuals J Amer Stat Ass 2000 95 442 443
Robins JM Greenland S Comment: Causal thinking without counterfactuals J Amer Stat Ass 2000 450 431 435
Rubin DB Comment: Causal thinking without counterfactuals J Amer Stat Ass 2000 95 435 437
|
16159397
|
PMC1239917
|
CC BY
|
2021-01-04 16:32:52
|
no
|
BMC Med Res Methodol. 2005 Sep 13; 5:28
|
utf-8
|
BMC Med Res Methodol
| 2,005 |
10.1186/1471-2288-5-28
|
oa_comm
|
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-581615689410.1186/1471-2202-6-58Research ArticleThe neurotrophin receptor p75NTR mediates early anti-inflammatory effects of estrogen in the forebrain of young adult rats Nordell Vanessa L [email protected] Danielle K [email protected] Shameena [email protected] Farida [email protected] Department of Human Anatomy and Medical Neurobiology, Texas A&M University System Health Science Center College of Medicine, College Station, TX, 77843 USA2005 12 9 2005 6 58 58 29 3 2005 12 9 2005 Copyright © 2005 Nordell et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Estrogen suppresses microglial activation and extravasation of circulating monocytes in young animals, supporting an anti-inflammatory role for this hormone. However, the mechanisms underlying estrogen's anti-inflammatory effects, especially in vivo, are not well understood. The present study tests the hypothesis that anti-inflammatory effects of estrogen are mediated by the pan-neurotrophin receptor p75NTR. Previously, we reported that estrogen attenuated local increases of interleukin(IL)-1β in the NMDA-lesioned olfactory bulb, while further increasing NGF expression.
Results
The present studies show that this lesion enhances expression of the neurotrophin receptor p75NTR at the lesion site, and p75NTR expression is further enhanced by estrogen treatment to lesioned animals. Specifically, estrogen stimulates p75NTR expression in cells of microvessels adjacent to the lesion site. To determine the role of this receptor in mediating estrogen's anti-inflammatory effects, a p75NTR neutralizing antibody was administered at the same time the lesion was created (by stereotaxic injections of NMDA) and specific markers of the inflammatory cascade were measured. Olfactory bulb injections of NMDA+vehicle (preimmune serum) increased IL-1β and activated the signaling molecule c-jun terminal kinase (JNK)-2 at 6 h. At 24 h, the lesion significantly increased matrix metalloproteinase (MMP)-9 and prostaglandin (PG)E2, a COX-2 mediated metabolite of arachadonic acid. All of these markers were significantly attenuated by estrogen in a time-dependent manner. However, estrogen's effects on all these markers were abolished in animals that received anti-p75NTR.
Conclusion
These data support the hypothesis that estrogen's anti-inflammatory effects may be, in part, mediated by this neurotrophin receptor. In view of the novel estrogen-dependent expression of p75NTR in cells associated with microvessels, these data also suggest that the blood brain barrier is a critical locus of estrogen's neuro-immune effects.
==== Body
Background
Estrogen replacement in adult female rats is neuroprotective for a variety of experimentally induced neural injury models, such as ischemia [1-3], toxins [4-6], and forebrain transections [7]. Estrogen is an important modulator of the neural inflammatory response, principally via its actions on microglia, the brain-resident immune cell. In primary cultures of microglia or in glial cell lines, estrogen pretreatment attenuates lipopolysaccharide (LPS)-induced superoxide release, phagocytic activity [8] and inducible nitric oxide synthase (iNOS) [8-11]. In vivo, estrogen treatment prevents the activation of microglia and recruitment of monocytes following injections of LPS [6], and attenuates expression of the inflammatory cytokine interleukin (IL)-1β following excitotoxic lesions [5].
The mechanisms by which estrogen exerts an anti-inflammatory effect, especially in vivo, remain poorly understood. In this report, we test the hypothesis that estrogen's anti-inflammatory effects are mediated via the neurotrophin family of growth factors and receptors. Members of this family include nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF) and neurotrophins (NT)-3 and 4/5. Although first identified by their potent effects on neuronal survival and differentiation, NGF and BDNF are widely distributed and have a broad range of effects. BDNF, for example, is widely synthesized in the vascular system [12-15] and NGF is synthesized in immune cells (for review see [16]). Local NGF expression increases dramatically following a variety of injurious stimuli [17-19] and may prevent toxic effects of other inflammatory cytokines [20].
P75NTR, the pan neurotrophin receptor which binds NGF and all the known neurotrophins, also increases following neuronal injury [21,22]. This increase may be associated with glia [23] and endothelial cells [24]. Several studies have shown that NGF, when bound to p75NTR, initiates apoptosis in several cell types (reviewed in [25]) and recent evidence suggests that pro-NGF may preferentially bind p75NTR and initiate apoptosis [26,27]. P75NTR-NGF interactions have also been implicated in suppressing the immune response by inhibiting the induction of major histocompatibility complex (MHC) class II proteins [28] and by suppressing the transmigration of circulating immune cells to the brain [24,29].
Although classified as a neurotrophin receptor, p75NTR may also be considered a member of the tumor necrosis factor receptor (TNFR) family, due to the presence of specific domains called TRAF-interacting motifs or TIMs [30]. TRAF's are TNFR-associated factors, and recruitment of these factors by TIM-containing members of the TNFR family results in the activation of multiple signaling pathways such as NFkB, JNK, ERK and PI-3K, which lead to immune and inflammatory responses as well as cell survival and differentiation (for review see [31]). P75NTR has been shown to bind recruitment molecules such as the interleukin-receptor associated kinase (IRAK) [32], which are typically recruited by IL ligand-receptor complexes to activate NFkB and to initiate a pro-inflammatory cascade. Although the downstream consequences of p75NTR-IRAK interactions are not well known, one possibility is that p75NTR competes with IL-1β for this molecule, and may thus attenuate the inflammatory cascade initiated by the interleukins.
The present study was designed to determine the mechanism of estrogen's anti-inflammatory effect in lesioned forebrain of young adult females. Specifically, we examined whether estrogen interacted with the pan-neurotrophin receptor p75NTR to suppress the inflammatory cascade. Following brain injury, there is a rapid synthesis and release of inflammatory cytokines such as IL-1β. These cytokines activate signaling intermediaries such as NFkB and JNK to initiate the transcription of downstream inflammatory genes. We tested the effects of bulb lesions on key portions of this cascade (IL-1β expression, JNK activation, PGE2 and MMP-9 expression) in animals that were given either estrogen or control pellets. The role of p75NTR was evaluated by using a neutralizing antibody to this receptor, injected into the lesion site. In the present study, we used NMDA-induced olfactory bulb lesions, which we have previously shown results in an inflammatory response as indicated by microglial activation and increases in IL-1β [5]. Here we report that olfactory bulb lesions to ovariectomized females increase p75NTR expression, and receptor expression is further enhanced in females that received estrogen treatment. Specifically, estrogen treatment stimulates p75NTR expression in cells associated with microvessels. Antibody experiments reveal that while estrogen treatment suppresses the inflammatory cascade resulting from bulb lesions, blocking the p75NTR receptor completely abolishes estrogen's effects on the early and late markers of the inflammatory cascade, but has no effect on these markers in controls (non-estrogen treated animals). These data support the hypothesis that estrogen's anti-inflammatory effects are mediated through p75NTR. In view of our recent studies that estrogen reduces extravasation of dye from circulation to brain tissue [33] and the current evidence that estrogen stimulates p75NTR expression in the wall of microvessels, we propose that early in the course of brain injury, estrogen pretreatment suppresses the inflammatory cascade by affecting the blood brain barrier.
Results
Confirmation of estrogen treatment
Plasma estradiol levels were measured from trunk blood collected at termination (approximately three weeks after pellet implantation). Ovariectomized animals replaced with an estrogen-containing pellet (E2) had average plasma estradiol levels of 51.66 ± 4.19 pg/ml, typical of levels seen at proestrus in this animal. The average weight gain as a result of estrogen treatment was -6.24 ± 2.1 g. In contrast, ovariectomized, control pellet-replaced animals (OVX) had low estradiol levels (9.1 ± 0.49 pg/ml) with a corresponding weight gain of 57.5 ± 2.76 g.
Confirmation of anti-p75NTR antibody treatment
In the present studies, the effectiveness of anti-p75NTR antibodies was tested in vivo by measuring caspase-3 activity. Caspase-3, an effector caspase that is proximal to the apoptotic cell death cascade [34], was used as a marker since p75NTR has been implicated in apoptotic cell death. Caspase-3 activity was significantly reduced (by 40%, p < 0.05) in animals that received NMDA+anti-p75NTR antibody (0.73 ± 0.06 pmol AMC liberated/min at 37°C/mg protein) as compared to animals that received NMDA+preimmune serum (1.23 ± 0.04 pmol AMC liberated/min at 37°C/mg protein).
Specificity of the actions of the p75NTR antibody was not tested directly, by the use of non-specific antibody or antibody to an unrelated protein. However, as described below, the actions of the antibody were seen only in estrogen-treated animals and not in control-treated animals. Furthermore, the actions of the antibody reversed estrogen's actions on inflammatory markers (see Figures 3, 5, 6 and 7), but did not reverse estrogen's effects on the neurotrophin NGF (see Figure 8).
P75NTR regulation following olfactory bulb lesions
P75NTR expression in olfactory bulb lysates was determined by Western blot assay (Figure 1A), and the quantitated signal, normalized to JNK, is shown in a histogram in Figure 1B. P75NTR expression increased 3-4 fold in lesioned animals as compared to sham-injected animals (F1,20: 12.2; p < 0.05). Estrogen treatment further enhanced the level of lesion-induced p75NTR (F1,20: 4.99, p < 0.05). As reported before, this is the group where expression of the inflammatory cytokine IL-1β is attenuated by estrogen [5]. The p75NTR antibody used in Western blot assays recognized a size-appropriate band in olfactory bulb tissue and PC12 cells, a prototypic p75NTR positive cell line.
P75NTR immunohistochemistry
P75NTR expression in the olfactory bulb of sham and lesioned animals was determined by immunofluorescence (Figure 2), using the same antibody that was used for Western blot analysis. A few sections were also probed with the neutralizing p75NTR antibody. Figure 2a shows the cellular architecture of the olfactory bulb in a section stained with a nuclear dye (DAPI). With either p75NTR antibody, prominent staining was noted in the fibers of the glomerular layer of lesion and sham-injected animals (Figure 2b; taken from region indicated by solid-line box in 2a), similar to that reported by others [35]. While the central portion of the bulb was poorly immunoreactive for p75NTR in the sham-lesion animals (Figure 2c), NMDA injections resulted in bright staining in fibers and cells surrounding the lesion site (Figure 2d). Although the type of cell and fibers is not known currently, the pattern of staining did not appear to be any different in the estrogen-replaced and estrogen-deprived animals. This diffuse staining pattern likely accounts for the bulk of p75 expression observed following lesions in the Western blot analysis. The only location where p75NTR expression differed in estrogen- and control-treated lesioned animals was associated with blood vessels. P75NTR-positive staining was visible in cells lining microvessels near the lesion site in estrogen-treated animals (Figure 2e and 2g). Typically, these vessels contained one or two curved nuclei, characteristic of endothelial cells (visualized by DAPI, Figure 2f and 2h), indicating that these vessels were either capillaries or postcapillary venules. A high magnification image from a section of an estrogen-treated animal photographed under both UV and fluorescein illumination, shown in Figure 2i, depicts p75NTR (in green) localized to a cell lining the vessel wall, identified by its curved nucleus (in blue). While microvessel staining was easily seen in the estrogen-treated animals, there was virtually no p75NTR staining seen in microvessels in the control pellet replaced animals (Figure 2j; nuclear stain in 2k) and in control sections from lesioned, estrogen-replaced animals where the primary antibody was not applied (Figure 2l; nuclear stain in 2m). Photomicrographs (e) through (m) were obtained from the region indicated by the hatched box shown in 2a.
Markers of inflammation
To test if p75NTR mediates the anti-inflammatory effects of estrogen, control and estrogen-pellet replaced animals were injected with NMDA+anti-p75NTR, and compared to sham lesioned controls and controls that received NMDA+preimmune serum. Animals were either terminated 6 or 24 h after the lesion, and olfactory bulb protein lysates were used to measure several proteins associated with the inflammatory cascade.
IL-1β expression
Local levels of IL-1β, measured by ELISA assay, increased dramatically at 6 h (F2,30: 51.61, p < 0.05) and 24 h (F2,30: 39.31, p < 0.05) after NMDA injections (Figure 3). At 6 h after lesion, estrogen treated groups had slightly lower levels of IL-1β as compared to placebo groups (F1,30: 5.72, p < 0.05). At 24 hours after lesion, estrogen replacement (OVX+E) clearly blunted the injury-related increase in IL-1β (F2,30: 4.99, p < 0.05). However, neutralizing p75NTR antibodies completely abolished estrogen's effects, causing IL-1β levels to be no different from those seen in the control-pellet (OVX) replaced NMDA-lesioned group. Note that anti-p75NTR treatment had no effect on the control-replaced group.
Regulation of IL-1β mRNA and ICE activity
To determine the locus of hormone action on this cytokine, IL-1β mRNA levels and the activity of the interleukin-converting enzyme (ICE or Caspase-1) was measured. As shown in Figure 4A, RT-PCR analysis indicated that the lesion substantially increased IL-1β mRNA (F1,12: 146.85; p < 0.05) at 24 h, however, estrogen pretreatment did not affect mRNA expression (F1,12: 0.01, p > 0.05). Furthermore, activity levels of caspase-1, which cleaves the pro-peptide to mature IL-1β, were not regulated by lesion or estrogen treatment (F5,30: 0.99; p > 0.05) at 24 h, (Figure 4B), suggesting that this pathway is not initially responsible for the increase in activated IL-1β.
pJNK expression
C-jun kinase activation was determined using antibodies specific for the phosphorylated form of the protein (pJNK) and normalized to total JNK protein (Figure 5). Histogram shows the ratio of pJNK2/JNK at 6 and 24 hours post injury (hpi). JNK is one of the signaling molecules activated by IL-1β when the latter is bound to its receptor. As shown in Figure 5, pJNK2, but not pJNK1, was increased by lesion at 6 h (F2,30: 19.22, p < 0.05) and remained activated at 24 h (F2,30: 19.9, p < 0.05). pJNK1, on the other hand, was constitutively active, which is typical for this isoform. JNK2 has a higher affinity for c-jun [36], and activated JNK2 translocates to the nucleus, while pJNK1 generally remains in the cytoplasmic compartment [37]. Post hoc analysis indicated that estrogen treatment (OVX+E) attenuated pJNK2 activation in lesioned animals at 24 h as compared to control replaced animals. However, concurrent injection of NMDA+antip75NTR abolished estrogen's effects on pJNK2 at 24 h, such that the relative activation of this molecule was no different from that of control-treated (OVX) animals.
Prostaglandin E2 (PGE2) levels
At 6 h after the lesion, PGE2 levels in the olfactory bulb were no different from sham controls (F2,30: 0.21, p > 0.05; Figure 6), indicating that this marker occurs later in the inflammatory cascade. At 24 hours post injury, there was a 2-fold increase in PGE2 levels in lesioned animals that were deprived of estrogen, while estrogen treatment completely attenuated this increase in PGE2. Anti-p75NTR antibodies completely eliminated estrogen's effects, causing PGE2 levels to rise similar to those seen in the control-replaced (OVX) animals. Note that as with IL-1β, anti-p75NTR treatment had no effect on PGE2 levels in the control-replaced group.
MMP-9 activity
The 96 kD metalloproteinase MMP-9 is secreted by several cell types and is activated by cytokines via the JNK/NFkB pathways. To test whether the lesion influenced expression of MMP-9, both mRNA expression of this gene and functional activation of its protein were analyzed by RT-PCR (Figure 7A) and gelatin zymography (Figure 7B), respectively. While MMP-9 mRNA expression was increased in olfactory bulb lysates 24 h after the lesion (F1,12: 200.32; p < 0.05; Figure 7A), estrogen treatment resulted in a small but signifcant reduction of MMP-9 mRNA (F1,12: 12.4; p < 0.05, main effect of hormone). Lesion-induced increases in MMP-9 mRNA were paralleled by increased MMP-9 activity at 24, but not 6 hours post injury (data not shown), as assayed by gelatin zymography. At 24 hpi, there was a significant increase in MMP-9 gelatinolytic activity as a result of the lesion (F2,30: 10.12, p < 0.05), and estrogen treatment attenuated MMP-9 activity in lesioned animals (F1,30: 21.15, p < 0.05). As with other markers of inflammation measured here, post-hoc analysis indicated that estrogen's effect on MMP-9 activity was reversed in estrogen-treated animals that received anti-p75NTR treatment (F2,30: 3.84, p < 0.05; Figure 7B).
NGF expression
NGF expression in olfactory bulb lysates was measured by an ELISA assay and normalized to total protein. NGF is synthesized by a variety of cell types in the nervous and immune system and one of its principal actions in the brain is to promote growth and regeneration. As shown in Figure 8, olfactory bulb lesions increased NGF levels when measured at 24 h post lesion (F2,30: 24.154, p < 0.05) and estrogen treatment also increased constitutive and lesioned-induced expression of this growth factor (F2,30: 9.41, p < 0.05). However, anti-p75 antibodies did not attenuate the lesion-induced increase in NGF expression in either control or estrogen-treated animals, suggesting that the anti-p75NTR specifically targets the inflammatory cascade (IL-1β, PGE2, MMP-9) rather than repair pathways.
Discussion
In previous studies, we have shown that the mechanical and excitotoxic damage caused by an NMDA injection to the olfactory bulb results in cell loss, astrocytosis and loss of cholinergic function [4], as well as microglial activation and a concomitant increase in the inflammatory cytokine IL-1β [5]. The present studies confirm our earlier observations that the inflammatory response in the forebrain of ovariectomized young adult rats is attenuated in animals that were replaced with estrogen as compared to animals that received a control (non-estrogen) pellet [5]. This report also shows that estrogen treatment further stimulates p75NTR expression, and that antibodies to p75NTR, administered concurrently with the lesion, completely abolish estrogen's anti-inflammatory effects. This study supports the hypothesis that estrogen's anti-inflammatory effects in the early portion of neural injury are mediated via the p75NTR. Since anti-p75NTR treatment had no effect on control replaced lesioned animals, these data further suggest that p75NTR expression in cells associated with the microvessel wall, which estrogen stimulates in lesioned animals only, may be a key cellular target of estrogen's anti-inflammatory effects.
The inflammatory cascade contains several protein and proteolipid mediators designed to contain the infectious agent and to phagocytize injured and dead cells. Mature IL-1β binds to its receptor and recruits the interleukin receptor associated kinase (IRAK), and this complex subsequently activates NFkB by association with TRAF-6, a member of the TNF receptor associated factor family [38,39]. Activation of signaling molecules, such as NFkB and JNK, result in the coordinate transcriptional expression of several downstream inflammatory mediators such as IL-6, IL-10, MMP-9 and cyclooxygenase-2 (COX-2), all of which contain motifs for NFkB or AP-1 in their promoter regions. The present study focused on a few key mediators, representing the early and later points of the inflammatory cascade. IL-1β, which is secreted by activated microglia/macrophages, is an early indicator of inflammation, and is known to activate JNK [40]. Both JNK and NFkB regulate MMP-9 [41] and COX-2 (the synthesizing enzyme for PGE2) [42]. In the present study, the lesion resulted in an early increase in IL-1β, JNK, and MMP-9 and a delayed increase in PGE2. All these indicators were attenuated in lesioned animals that received estrogen, and estrogen's actions were abrogated by anti-p75NTR antibodies. These data strongly support the hypothesis that p75NTR is an important mediator of estrogen's anti-inflammatory effects. While the mechanism of p75NTR anti-immune action is not yet known, one possibility is that p75NTR may interfere with IL-1β signaling pathways by competing for a limited pool of IRAK. NGF/p75NTR complexes are also capable of recruiting IRAK [32], and may therefore reduce the ability of interleukin receptor ligand complexes to activate NFkB or JNK, and consequently, hamper transcription of inflammation-related genes.
Identified as the first neurotrophin receptor [43,44], p75NTR has been implicated in both survival and apoptotic pathways [45]. While p75NTR interacts with the receptor tyrosine kinases, or trks, it has independent signaling activities mediated through the ceramide and JNK pathways as well [46]. P75NTR also has significant homology with a family of cell death molecules, such as TNFr and Fas (see [45,47] for reviews) and has been shown to promote apoptotic cell death in PC12 cells [48], retinal cells [49], oligodendrocytes [50], Schwann cells [51] and otic vesicles [52]. P75NTR expression also increases following injury, such as spinal cord injury [27,53]. More recently, studies have implicated p75NTR in the regulation of the inflammatory response, ranging from suppression of MHC molecules to preventing the transmigration of leukocytes. In hippocampal explant cultures, endogenous neurotrophins produced by neurons appear to control the antigen presenting ability of microglia, and neurotrophins, via p75NTR, suppress MHC II molecules [28]. P75NTR has also been localized to peripheral monocytes, and NGF, via p75NTR, prevents monocyte transmigration through the blood brain barrier [29]. In p75 knock-out animals, the severity of inflammation caused by a cranial nerve injury [54] or experimental allergic encephalomyelitis [24] is significantly greater as compared to wildtype controls. The p75NTR null animals had increased staining of activated microglia, and a massive recruitment of T-lymphocytes at the lesion site, indicative of increased permeability of the blood brain barrier [54]. The blood brain barrier serves to exclude peripheral cells, specific proteins and molecules from the brain, and p75NTR induction in endothelial cells following experimental allergic encephalomyelitis [24] suggests that this receptor may serve to maintain the integrity of the blood brain barrier. The present study indicates that estrogen exploits p75NTR-dependent anti-inflammatory mechanisms, and that this action may also occur at the blood brain barrier. In fact, estrogen itself is known to reduce transport across the blood brain barrier in injured [55,56] and non-injured animals [33] and reduces monocyte migration into the brain following ischemia [6].
Local levels of inflammatory cytokines such as IL-1β after traumatic neural injury are derived from two sources: initially from the activation of local microglia and eventually from the influx of circulating macrophages (and their products) that enter the brain due to progressive changes in the blood brain barrier. At 6 h after injury, estrogen treatment has a very modest effect on the IL-1β levels, and consequently, has little or no effect on JNK2 activation, which is activated by IL-1β/IL receptor complexes. At 24 h, however, estrogen markedly suppresses IL-1β levels and this is reflected in a pronounced decrease in JNK activation at this time point. This biphasic effect of estrogen on IL-1β suggests that estrogen may exert its effects not by acting on local microglia, but by reducing the pool of activated immune cells that are recruited to a central lesion site from circulatory sources, either by maintaining the integrity of the blood brain barrier, or by direct action on monocytes/macrophages. Our recent studies show that both possibilities are likely. For example, estrogen treatment suppresses cytokine production in circulating immune cells, but not microglia, when challenged ex vivo with LPS [57]. Furthermore, estrogen treatment also reduces permeability of the blood brain barrier as measured by extravasation of Evan's blue dye [33]. This latter hypothesis is also supported by the fact that estrogen fails to suppress IL-1β mRNA or the activity of caspase-1, which cleaves precursor IL-1β to its mature active form, at 24 hours post injury. This suggests that estrogen may not directly alter the availability of a local pool of IL-1β by canonical pathways but acts to gate the entry of cytokines from circulating immune cells or those produced by microvessel-associated cells.
Vascular permeability and leukocyte invasion can be increased by matrix remodelling resulting from the actions of the MMP family [58,59]. MMP-9 increases following experimentally induced stroke, and the availability of this proteinase directly contributes to the size of the infarct [60]. Several factors may control MMP-9 regulation and one report indicates that NGF associates with p75NTR to downregulate MMP-9 activity [61]. Estrogen also has been shown to reduce MMP-9 mRNA [6,62] which likely contributes to its neuroprotective actions. In the present study, estrogen-induced decreases in MMP-9 were reversed by anti-p75NTR. Although this study does not address the source of MMP-9 or its substrate, it may be the case that estrogen-induced reductions in MMP-9 provide the means for maintaining the integrity of the blood brain barrier in these animals.
The observation that estrogen treatment exacerbates p75NTR expression in lesioned animals is surprising in view of the many studies where estrogen has been shown to decrease p75NTR mRNA [63-65] and protein [66,67] expression. However, these studies were performed in unlesioned animals (without neural trauma), and may therefore be representative of neuronal p75NTR, as has been shown in basal forebrain neurons [65]. This was also the case in a more recent study where estrogen suppressed p75NTR expression in hippocampal neurons following ischemic injury in male gerbils [68]. Moreover, the same study also reported that p75 expression in the hippocampus was first seen 48 h after injury. The present study, on the other hand, describes a very early event in the injury process. Furthermore, the only region where p75NTR expression was regulated differentially was in the microvasculature of estrogen-treated, lesioned animals. This specificity may explain why anti-p75NTR treatment abolished estrogen's effects on the inflammatory cascade but had no discernable effects on the placebo-replaced animals. We hypothesize that the early increase in microvascular p75NTR may play a significantly different role in inflammation as compared to the later expression of neuronal p75NTR.
The present study shows that estrogen replacement to ovariectomized young females suppresses the inflammatory response following neural injury; however, hormonal regulation of the inflammatory response is not always benign. While estrogen reduces experimentally induced inflammation in the anterior chamber of the eye [69], lungs [70] and the tibiotarsal joint in adjuvant-induced arthritis [71,72], this hormone promotes inflammation in the prostate [73,74], and stimulates edema [75], vascular permeability and influx of macrophages in the uterus [76-79]. Even in models of neurogenic inflammation, such as the present study, estrogen treatment is not uniformly neuroprotective. For example, a similar 3 wk regimen of estrogen to reproductive senescent females does not attenuate the production of IL-1β, and actually exacerbates it [5]. Furthermore, in these senescent animals, estrogen fails to enhance p75NTR expression following a lesion (data not shown). While the reasons for estrogen's disparate actions on different tissues is not clear, the present study suggests that p75NTR may be one of the switches that predict whether estrogen will exert neuroprotective or neurotoxic effects. An important direction for this work would be to test this hypothesis in the p75 knock-out model or, preferably, a conditional p75NTR knock-out animal, which eliminates the compensatory changes that might occur as a result of developmental loss of p75.
Methods
Animals
Sprague Dawley females (~250 g, 4 months) were purchased from Harlan Laboratories (IN) and maintained in an AALAC-approved facility on a 12-h light: 12-h dark cycle with lights on at 06:00 h, with food and water available ad libitum. All procedures were in accordance with NIH and institutional guidelines governing animal welfare. Several sets of animals were prepared, each with internal controls. For convenience a chart is provided below. For the p75 expression studies (Westerns blot analysis and immunohistochemistry), we used tissue from four sets of animals that had been prepared for a previous study, and related data from these animals is published in [5]. Briefly, young adult animals were ovariectomized and replaced with either control or estrogen pellets and 3 weeks later were assigned to either sham lesion or NMDA lesion groups. Twenty four hours later, animals were sacrificed and the olfactory bulbs were dissected and later harvested for proteins. A second set was prepared as above for histological analysis, where animals were perfused following anesthetic overdose and the brains recovered for histological processing (described later).
For p75NTR antibody studies, two sets of animals were prepared: placebo or estrogen-replaced animals were assigned to one of three groups: sham lesion, NMDA+vehicle control (preimmune rabbit serum) or NMDA+anti-p75NTR antibodies (made in rabbit). One set of animals was sacrificed at 6 h after the lesion, and the second set at 24 h after the lesion. In all sets, each treatment group consisted of 5-6 animals per group (See Table 1). Specific surgical procedures are detailed below.
Animals were also used for pilot studies to (a) determine the most effective concentration of anti-p75NTR antibodies (n = 12) and (b) to ensure that there were no differences between animals injected with NMDA and NMDA+preimmune serum (the vehicle for the anti-p75NTR antibody) (n = 6).
Surgical techniques
Ovariectomies: As previously described [4,66,80,81], animals were anesthetized with ketamine (87 mg/kg)/xylazine (13 mg/kg) and bilateral ovariectomies were performed using a dorsal midline incision inferior to the palpated rib cage and kidneys. Ovaries and surrounding tissue were removed and 60-day time-release 17β estradiol pellets (1.0 mg) or control pellets (Innovative Research, FL) were inserted subcutaneously (s.c.) prior to closing the incision. These pellets have been used extensively in our work and have resulted in physiological levels of plasma estradiol for periods of 3 to 6 weeks [4,5,66].
Stereotaxic surgeries
After 3 weeks of estrogen or control treatment, all animals were anesthetized with ketamine (87 mg/kg)/xylazine (13 mg/kg) and placed in a rodent stereotaxic apparatus. Skin and cranial fascia were resected and the skull exposed. Two small craniotomies were made in all rats assigned to lesion groups, to expose the olfactory bulbs at the following coordinates: 7.6 mm anterior to bregma and 1.0 mm lateral to the sagittal suture. The tip of a Hamilton syringe needle was briefly lowered to a depth of 3.4 mm, and immediately raised by 0.2 mm to create a trough. A total of 2 μl (1 μl per side) of 50 nM NMDA was injected at a rate of 0.2 μl/30 s. For anti-p75NTR experiments, 1 μL of 50 nM NMDA + 2 μL preimmune rabbit serum (L groups) or 1 μL 50 nM NMDA + 1.5 μL preimmune rabbit serum + 0.5 μL ap75NTR (P groups) was injected in each bulb at a rate of 0.5 μL/30 s. After injections, the needle was raised slowly, craniotomies filled with gel-foam, and scalp sutured with wound clips. Surgical controls were anesthetized, restrained in the stereotaxic apparatus and their scalp and cranial fascia resected and reclipped. Other reports [82] and our previous studies [5] indicate that at 24 h post injury a saline injection is not an appropriate control since the injection itself results in injury.
The anti-p75NTR antibody used here (Chemicon, CA, AB1554) recognizes the extracellular domain of the p75NTR and has been shown to block NGF/p75NTR interactions [83-85]. This antibody is not recommended for Western blots (hence a different antibody was used for the Western blot assays), but can be used for immunohistochemistry. Both the neutralizing antibody used here and the antibody used for Westerns revealed similar staining patterns in the olfactory bulb when used for immunohistochemistry.
Animals were sacrificed at 6 h and 24 h after stereotaxic surgery by rapid decapitation and trunk blood was collected for estimation of estradiol content by radioimmunoassay (Diagnostic Systems Laboratories, TX). Olfactory bulbs were rapidly removed and stored at -80°C. Proteins were isolated from the entire olfactory bulb using previously established procedures [4,80,81] and total protein concentrations were determined using the BCA protein assay kit (Pierce, IL). In some cases, bulbs were processed for RNA extraction described below.
Western blot analysis
Equal amounts of total protein from tissue lysates were size-fractionated on a polyacrylamide gel, transferred to a nylon membrane (Hybond C-Super, Amersham, NJ) and analyzed for p75NTR and pJNK. Blots were also probed for total JNK expression as a loading control, since this protein did not vary with lesion or estrogen treatment [81]. Membranes were blocked with 5% milk in 1 × Tris Buffered Saline + 0.05% Tween (TTBS) solution for 1 h, followed by incubation with the primary antibody in milk-TTBS (1/ 1500 Ms × NGF Receptor (p75NTR; MAB365), Chemicon, CA; 1/2500 anti-active JNK, Promega, WI; 1/2000 JNK, Santa Cruz, CA) for 1 h at room temperature (p75NTR) or overnight at 4°C (pJNK, JNK). Blots were washed with TTBS (3 × 10 min) and primary antibody was detected using an HRP-conjugated secondary antibody in milk-TTBS (anti-mouse IgG 1/3000 for p75NTR, Transduction Laboratories, KY; 1/6000 anti-rabbit IgG for pJNK, Promega, WI; 1/2000 anti-rabbit IgG for JNK, Santa Cruz, CA). All incubations were performed at room temperature with gentle shaking. After the last wash step, an enzyme-catalyzed chemiluminescent reagent (Renaissance NEN, MA) was used for immunodetection. Signals were detected using X-ray film and bands quantified by densitometric analysis (Molecular Analyst, Bio-Rad, CA) and normalized to JNK expression.
Immunohistochemistry for p75NTR
Immunohistochemistry was performed on olfactory bulb sections of placebo- or estrogenreplaced ovariectomized young-adult animals that were subject to olfactory bulb lesions as outlined above and sacrificed 24 h later. Sections were prepared using the NeuroScience Associates Multibrain technology and some sections from this set were used in previous assays to detect activated microglia [5]. Briefly, animals were anesthetized with 0.6 mL pentobarbital solution (50 μg/mL pentobarbital, 10% ethanol, 40% propylene glycol) and perfused transcardially with PBS followed by 4% paraformaldehyde. Brains were removed from the skull and post-fixed in 4% paraformaldehyde (2 h). Brains were then shipped to NeuroScience Associates (Knoxville, TN), where they were treated with 20% glycerol and 2% dimethylsulfoxide to prevent freeze artifacts, and then embedded coronally in groups of 14-16 per block in a gelatin matrix. 40 μm freeze-cut sections through the entire olfactory bulbs were obtained with an AO 860 sliding microtome and collected in 4 × 6 array of containers filled with 10% phosphate-buffered formaldehyde. After 24 h, sections were rinsed and transferred into Antigen Preserve solution (50% PBS pH 7.0, 50% Ethylene glycol, 1% PVP) for storage and shipping. Prior to use in immunohistochemistry, sections were rinsed in PBS and mounted on gelatin coated glass slides. P75NTR was detected by fluorescence-labeled secondary. Sections were incubated with block solution and followed by overnight incubation with the primary antibody (1/200; mouse anti-NGFR; Chemicon, CA) diluted in PBS with 3% goat serum and 0.4% Triton. Controls were incubated with diluent only. Following washes, sections were incubated with a rat-adsorbed FITC-conjugated Fab specific secondary antibody (1/160 goat anti-mouse, Vector Labs, CA)or AlexaFluor 488 (1:2000 for goat anti-mouse, Invitrogen, CA) for 45 min at room temperature. Sections were then washed and coverslipped using a fluorescence-compatible mounting media which contained DAPI, a nuclear dye. The presence of glomerular staining in the bulb was also used as a positive control. Sections from the midpoint of the bulb (around the injection site) were examined under fluorescence illumination for FITC for the presence of p75NTR immunostaining. Five sections were examined from each animal and 5-6 animals were examined per treatment group.
Caspase-1 and Caspase-3 assays
A commercial kit was used to determine Caspase-1 (interleukin-1 converting enzyme) and Caspase-3 activity (Promega, WI) as before [81], using manufacturer's recommended procedures. Briefly, blanks, samples and negative controls were pipetted into a flat-bottom, black 96-well plate and incubated at 30°C for 30 minutes. Plates were then incubated with Caspase-1 (ICE) or CCP32 substrate (Ac-YVAD-AMC and Ac-DEVD-AMC, respectively) at 30°C for 60 minutes. A standard curve was prepared just prior to reading at 360 nm (excitation) and 460 nm (emission) in a fluorescence microplate reader (BioTek, VT). Measurement of liberated AMC, interpolated from standard curves, was normalized to protein per unit time (min).
Enzyme-linked immunosorbent assay (ELISA)
Commercial kits for IL-1β (R&D Systems, MN) and PGE2 (Cayman Chemical, MI) were used to determine cytokine and prostaglandin levels in tissue lysates, using procedures recommended by the manufacturers [5]. Briefly, standards, controls, samples and a biotinylated (IL-1β) or acetylcholinesterase-linked (PGE2) secondary antibody were pipetted into 96-well plates pre-coated with antibodies specific for rat IL-1β and PGE2, and incubated at RT for 2 h (IL-1β) or 4°C overnight (PGE2). Following washes, plates were sequentially incubated with streptavidin peroxidase (IL-1β: 2 h) or Ellman's reagent (PGE2: 60-90 min) and substrate solution for 30 min (IL-1β). Plates were read at 450 nm (IL-1β) or 405 nm (PGE2) in a microplate reader (Bio-Tek, VT). Standard curves were established from optical densities of wells containing known dilutions of standard, using KC3 software (Bio-Tek, VT) and sample measurements were interpolated from standard curves.
Reverse transcription-polymerase chain reaction
IL-1β and MMP-9 gene expression was determined by semi-quantitative reverse transcription-polymerase chain reaction (RT-PCR). RNA was extracted using the TRIZOL® Reagent RNA extraction method [86], followed by further purification with the Qiagen RNeasy Kit (Qiagen, CA). The RNA concentration was determined using the RiboGreen® RNA Quantification Kit (Molecular Probes, OR). Reverse transcription of total RNA (2 μg) was performed with the Gibco Superscript™ First Strand Synthesis System. PCR of the newly synthesized cDNA (8 ng) was performed using PCR Supermix (Invitrogen, CA; primer conc. 0.25 μM; 200 μM dNTP; 1× PCR Buffer, 1.5 mM MgC12, 20 U recombinant Taq DNA polymerase). IL-1β (5'-3': IL-1β-For TTG AAT CTA TAC CTG TCC TGT GTG; IL-1β Rev TGA CTT GGC AGA GGA CAA AGG) PCR cycles were as follows: 95°C 2 min; 30 cycles of 95°C 30 sec, 60°C 1 min, 72°C 2 min. PCR reactions for MMP-9 were performed using previously published primers [11]. PCR conditions were as follows: 95°C for 5 min followed by 40 cycles at 92°C for 1 min, 56°C for MMP-9 for 1 min, and 72°C for 1 min. Cyclophilin (5'-3': CPI-1, TGG TCA ACC CCA CCG TGT TCT TCG; CP1-2 TGC CAT CCA GCC ACT CAG TCT TGG) was used to normalize IL-1β and MMP-9 expression. The PCR cycles for cyclophilin are as follows: 95°C 2 min, 20 cycles of 95°C 30 sec, 62°C 1 min, 72°C 2 min. PCR reactions were performed on a Perkin-Elmer Thermal Cycler 480. Initial analyses were performed to ensure that IL-1β and cyclophilin were amplifying in the linear range. The PCR reaction was separated on a 1.5% agarose gel and quantified using Molecular Analyst™ (BioRad, CA). To avoid quantitation artifacts, only samples loaded on the same gel were analyzed. Hence, for each treatment group, 4 independent samples were analyzed. In every case, only a single band of the expected size was detected and bands amplified by each primer set was sequenced (Gene Technologies Laboratory, TAMU) and determined to be homologous to rat IL-1β and MMP-9 gene sequences.
Gelatin zymography
Procedures used here are a modification of our previous protocol using culture media [57]. Equal amounts of total protein were size fractionated on a 10% polyacrylamide gel containing 0.01% gelatin, along with prestained protein size markers. On some gels, a positive control was included (conditioned media from human umbilical vein endothelial cells; kind gift of G.E. Davis, TAMUS HSC). After electrophoresis, the gels were rinsed with MQ water and then incubated with renaturing buffer (2% Triton X-100 in MQ) at room temperature for 1 h with 3 buffer changes. Gels were then rinsed 3× with MQ water and incubated with developing buffer (50 mM Tris, 0.2 M NaCl, 5 mM CaCl2 0.02% Brij 35) overnight at 25°C with gentle shaking. After a brief rinse with MQ water, gels were stained with a 0.25% coomassie blue solution (50% methanol, 20% acetic acid) for 1 h and destained in 20% methanol: 10% acetic acid. Gels were dried at 50°C for 1 h, and later digitized. A standard densitometric program (Quantity One, BioRad, CA) was used to calculate the intensity of the lytic area.
Statistical analysis
Statistical analysis was performed using a statistical software package (SPSS Inc, IL), and group differences were considered significantly different at p < 0.05. Data was subject to a two-way analysis of variance (ANOVA) with hormone and lesion as independent variables. Planned post-hoc comparisons were performed for the lesion variable in those studies where there were three groups (sham, lesion+vehicle (preimmune serum) and lesion+anti-p75NTR).
Abbreviations
ERK: Extracellular-signal regulated kinase; ICE: Interleukin-1 converting enzyme; IL-1β: Interleukin-1beta; IRAK: Interleukin receptor-associated kinase; JNK: C-jun terminal kinase; LPS: Lipopolysaccharides; MMP-9: Matrix metalloproteinase-9; NFkB: Nuclear factor-kappaB; NGF: Nerve Growth factor; NMDA: N-methyl-d-aspartate; P75NTR: P75 neurotrophin receptor; PGE2: Prostaglandin E2, PI-3K: Phosphatidylinositol-3 kinase; TNFR: Tumor necrosis factor receptor; TRAF: TNFR associated factor
Authors' contributions
VLN performed almost all experiments and prepared most figures, DKL performed RT-PCR analyses and prepared the associated figures, SB participated in data analysis and interpretation, FS conceived the study, performed statistical analysis and photomicroscopy. All authors read and approved the final manuscript.
Acknowledgements
The authors wish to thank Debbie Geevarghese, Najma Ahmed and Lynne O'Kelley for technical assistance. Supported by grants from the NIH (AG 19515) and the Alzheimer's Association to FS.
Figures and Tables
Figure 1 P75NTR regulation following olfactory bulb lesions: P75NTR expression was determined by Western blot assays (A) and receptor expression was quantified and normalized to a loading control (JNK) in the histogram (B). P75NTR was increased in lesioned animals, and estrogen treatment for 3 weeks prior to the lesion further enhanced the expression of this receptor. Two representative examples are shown of each treatment condition in the Western blots. Bars represent mean ± SEM, n = 6 per group. (C) Specificity of p75NTR antibody: The antibody used in this Western assay recognizes a size appropriate band at 75 kD, which is not seen when the membrane is probed in the absence of the primary antibody. Lane 1: PC12 cell lysate, Lanes 2,3: olfactory bulb lysate. Key: OVX Sham: Ovariectomized female replaced with a control pellet, with sham injection, OVX lesion: ovariectomized female replaced with a control pellet, with NMDA injection, Est Sham: ovariectomized female replaced with an estrogen pellet, with sham injection, Est Lesion: ovariectomized female replaced with an estrogen pellet, with NMDA injection. *: p < 0.05.
Figure 2 P75NTR expression in the lesioned olfactory bulb: Olfactory bulb sections from sham and NMDA lesioned animals that were either estrogen or control-pellet replaced were probed for p75NTR expression and counter-stained with a nuclear dye (DAPI). A low magnification photomontage of olfactory bulb illuminated for DAPI is shown in (a). P75NTR is normally expressed in the glomerular layer of the olfactory bulb (b), photographed from the region indicated by the solid line box in (a). P75NTR is poorly expressed in other regions of the olfactory bulb, an example of which is shown in (c), taken from the region indicated by the wire frame in (a). However, in a comparable region of the NMDA-lesioned animals, diffuse p75NTR staining is seen in cells and fibers at the injury site (d). In estrogen-treated, lesioned animals p75NTR immunoreactivity is seen in cells associated with the walls of microvessels (e; g), identified by their curved nuclei (f; h) surrounding a lumen. Arrows indicate the same cells photgraphed under fluorescein (p75NTR; e,g) or UV illumination (DAPI nuclear dye; f,h). A double-labeled cell photographed under both fluoresein and UV is shown in (i). P75NTR labeling was not seen in microvessel walls in lesioned animals that received a control pellet (j) or in sections from estrogen-replaced lesioned animals that were incubated without a primary antibody (immunohistochemical control, l). Corresponding nuclear dye images are in k and m. Magnification bar: b-d = 160 μm; e-m = 25 μm.
Figure 3 IL-1β regulation by lesion and estrogen treatment: Olfactory bulb lesions significantly increased local IL-1β expression at 6 and 24 hours post injury (6 hpi, 24 hpi). Estrogen treatment (OVX+E) attenuated IL-lβ at 6 hpi, and this effect was more pronounced at 24 hpi. At 24 hpi, anti-p75NTR treatment abolished estrogen's effects on IL-1β. Anti-p75NTR had no effect on IL-1β expression in control pellet-replaced (OVX) animals. Bars represent means ± SEM, n = 6 per group. Sham: Sham lesion, Lesion+Vehicle: NMDA injection with pre immune serum, Lesion+Anti p75: NMDA injections with anti p75NTR antibodies. *:p < 0.05.
Figure 4 Regulation of IL-1β mRNA and the interleukin-converting enzyme: (A) mRNA from lesioned or sham injected animals, with estrogen or control pellets was reverse transcribed and amplified using primers specific for IL-1β and cyclophilin. PCR product was fractioned on an agarose gel and photographed under UV illumination. An image of these gels is shown here, with 4 representative animals from each group. At 24 h post injury, NMDA lesions increased IL-1β mRNA, however, estrogen-treated (OVX+E) animals were no different from the control-pellet (OVX) treated animals. (B) Since the mature form of IL-1β is obtained through cleavage of IL-lβ precursor protein by caspase-1, activity for this enzyme was also assayed. Neither NMDA lesions nor estrogen treatment caused any significant changes in caspase-1 activity at 24 h post injury. Histogram bars represent means ± SEM for the entire group (n = 6). *:p < 0.05.
Figure 5 JNK activity following olfactory bulb lesions: JNK activation was measured by Western blot assay using phospho-specific (p) antibodies, and pJNK2 expression was normalized to total JNK2 protein. JNK1 was constitutively active in all cases. Virtually no pJNK2 was seen in the sham-injected controls in either control or estrogen-replaced animals. Bulb lesions significantly increased pJNK2 at 6 h after the lesion. At 24 h, pJNK2 levels were still elevated in the control-replaced (OVX) lesioned animals but significantly attenuated in the estrogen-treated (OVX+E) lesioned group. Anti-p75NTR reversed the effects of estrogen on pJNK2 activation at 24 h, although it had no effect on the control-pellet group. Two representative examples are shown from each treatment condition in the Western blot images. Histogram bars represent means+SEM for the entire group (n = 6). Sham: Sham lesion, Lesion+Vehicle: NMDA injection with pre immune serum, Lesion+Anti p75: NMDA injections with anti-p75NTR antibodies. *:p < 0.05.
Figure 6 PGE2 levels in the lesioned olfactory bulb: PGE2 expression was not altered by lesion or hormone treatment at 6 h post injury (6 hpi). However, at 24 hpi, PGE2 expression increased 2-fold in the control-pellet (OVX) replaced lesioned groups, while this increase was not seen in the estrogen-treated (OVX+E) lesioned animals. However, anti-p75NTR treatment completely abolished estrogen's suppressive effects, making this group indistinguishable from the control pellet replaced, lesioned animals. Bars represent means ± SEM, n = 6 per group. Sham: Sham lesion, Lesion+Vehicle: NMDA injection with pre immune serum, Lesion+Anti p75: NMDA injections with anti-p75NTR antibodies. *:p < 0.05.
Figure 7 MMP-9 regulation: (A) MMP-9 mRNA was assayed by RT-PCR, and normalized to cyclophilin RNA. MMP-9 mRNA was significantly increased by lesion, and estrogen teatment resulted in a small but signficant decrease in MMP-9 mRNA expression. (B) MMP-9 activity was measured by gelatin zymography at 6 and 24 h post injury, and representative examples from each group are shown here. The location of the lytic area corresponded to the expected size of MMP-9 (96 kD), indicated by the filled-arrow head, as well as that of a positive control (conditoned media from endothelial cells). Line-arrow head indicates pre-stained protein size marker (86 kD). Virtually no MMP-9 activity was apparent in the sham-injected animals. At 24 h after lesion, olfactory bulb lesions increased MMP-9 activity, although this was attenuated in the estrogen-treated (OVX+E), lesioned animals. Estrogen-treated lesioned animals that received anti-p75NTR, however, had MMP-9 activity that was similar to that of the control-pellet (OVX) replaced lesioned animals. Histogram represents means+SEM of the measured lytic area for MMP-9 at the 24 h time point, n = 6 per group. Sham: Sham lesion, Lesion+Vehicle: NMDA injection with pre immune serum, Lesion+Anti p75: NMDA injections with anti p75NTR antibodies. *:p < 0.05.
Figure 8 NGF regulation by lesion and estrogen treatment: Olfactory bulb lesions significantly increased NGF expression at 24 hours post injury (24 hpi). Estrogen treatment (OVX+E) increased basal and lesion-induced expression of NGF. However, anti-p75NTR had no effect on NGF expression in control pellet-replaced (OVX) or estrogen-replaced (OVX+E) animals. Bars represent means ± SEM, n = 6 per group. Sham: Sham lesion, Lesion+Vehicle: NMDA injection with pre immune serum, Lesion+Anti p75: NMDA injections with anti p75NTR antibodies. *:p < 0.05.
Table 1 Summary P75NTR Western P75NTR Immuno RNA analysis Antip75 Expt 6 h Antip75 Expt 24 h
Sham Lesion NMDA Sham Lesion NMDA Sham Lesion NMDA Sham Lesion NMDA+ Veh* NMDA+Anti p75NTR Sham Lesion NMDA+ Veh* NMDA+Anti p75NTR
Control Pellet 6 6 5 6 6 6 6 6 6 6 6 6
Estrogen-Pellet 6 6 6 6 6 6 6 6 6 6 6 6
Analysis Western blots Immuno histochemistry RT-PCR analysis of IL-1β, MMP-9 IL-1β, pJNK, PGE2, MMP-9, IL-1β, pJNK, PGE2, MMP-9 caspase-3, caspase-1
* = Pre immune serum
==== Refs
Hurn PD Littleton-Kearney MT Kirsch JR Dharmarajan AM Traystman RJ Postischemic cerebral blood flow recovery in the female: effect of 17 beta-estradiol J Cereb Blood Flow Metab 1995 15 666 672 7790416
Simpkins JW Rajakumar G Zhang YQ Simpkins CE Greenwald D Yu CJ Bodor N Day AL Estrogens may reduce mortality and ischemic damage caused by middle cerebral artery occlusion in the female rat J Neurosurg 1997 87 724 730 9347981
Dubal DB Kashon ML Pettigrew LC Ren JM Finklestein SP Rau SW Wise PM Estradiol protects against ischemic injury J Cereb Blood Flow Metab 1998 18 1253 1258 9809515 10.1097/00004647-199811000-00012
Sohrabji F Peeples KW Marroquin OA Local and cortical effects of olfactory bulb lesions on trophic support and cholinergic function and their modulation by estrogen J Neurobiol 2000 45 61 74 11018768 10.1002/1097-4695(20001105)45:2<61::AID-NEU1>3.0.CO;2-L
Nordell VL Scarborough MM Buchanan AK Sohrabji F Differential effects of estrogen in the injured forebrain of young adult and reproductive senescent animals Neurobiol Aging 2003 24 733 743 12885581 10.1016/S0197-4580(02)00193-8
Vegeto E Belcredito S Etteri S Ghisletti S Brusadelli A Meda C Krust A Dupont S Ciana P Chambon P Maggi A Estrogen receptor-alpha mediates the brain anti inflammatory activity of estradiol Proc Natl Acad Sci U S A 2003 100 9614 9619 12878732 10.1073/pnas.1531957100
Rabbani O Panickar KS Rajakumar G King MA Bodor N Meyer EM Simpkins JW l7beta-estradiol attenuates fimbrial lesion-induced decline of ChAT-immunoreactive neurons in the rat medial septum Exp Neurol 1997 146 179 186 9225751 10.1006/exnr.1997.6516
Bruce-Keller AJ Keelink JL Keller JN Huang FF Camondola S Mattson MP Antiinflammatory effects of estrogen on microglial activation Endocrinology 2000 41 3646 3656 10.1210/en.141.10.3646
Drew PD Chavis JA Female sex steroids: effects upon microglial cell activation Neuroimmunol 2000 111 77 85 10.1016/S0165-5728(00)00386-6
Vegeto E Pollio G Ciana P Maggi A Estrogen blocks inducible nitric oxide synthase accumulation in LPS-activated microglia cells Exp Gerontol 2000 35 1309 1316 11113609 10.1016/S0531-5565(00)00161-3
Vegeto E Bonincontro C Pollio G Sala A Viappiani S Nardi F Brusadelli A Viviani B Ciana P Maggi A Estrogen prevents the lipopolysaccharide-induced inflammatory response in microglia J Neurosci 2001 21 1809 1818 11245665
Donovan MJ Miranda RC Kraemer R McCaffrey TA Tessarollo L Mahadeo D Sharif S Kaplan DR Tsoulfas P Parada L Neurotrophin and neurotrophin receptors in vascular smooth muscle cells. Regulation of expression in response to injury Am J Pathol 1995 147 309 324 7639328
Nemoto K Fukamachi K Nemoto F Miyata S Hamada M Nakamura Y Senba E Ueyama T Gene expression of neurotrophins and their receptors in cultured rat vascular smooth muscle cells Biochem Biophys Res Commun 1998 245 284 288 9535823 10.1006/bbrc.1998.8418
Nakahashi T Fujimura H Altar CA Li J Kambayashi J Tandon NN Sun B Vascular endothelial cells synthesize and secrete brain-derived neurotrophic factor FEBS Lett 2000 470 113 117 10734218 10.1016/S0014-5793(00)01302-8
Chaldakov GN Fiore M Stankulov IS Manni L Hristova MG Antonelli A Ghenev PI Aloe L Neurotrophin presence in human coronary atherosclerosis and metabolic syndrome: a role for NGF and BDNF in cardiovascular disease? Prog Brain Res 2004 146 279 289 14699970 10.1016/S0079-6123(03)46018-4
Tabakman R Lecht S Sephanova S Arien-Zakay H Lazarovici P Interactions between the cells of the immune and nervous system neurotrophins as neuroprotection mediators in CNS injury Prog Brain Res 2004 146 387 401 14699975
Crutcher KA Collins F Entorhinal lesions result in increased nerve growth factor-like growth-promoting activity in medium conditioned by hippocampal slices Brain Res 1986 399 383 389 3828773 10.1016/0006-8993(86)91533-7
Whittemore SR Larkfors L Ebendal T Holets VR Ericsson A Persson H Increased beta-nerve growth factor messenger RNA and protein levels in neonatal rat hippocampus following specific cholinergic lesions J Neurosci 1987 7 244 251 3806196
Lorez H Keller F Ruess G Otten U Nerve growth factor increases in adult rat brain after hypoxic injury Neurosci Lett 1989 98 339 344 2725951 10.1016/0304-3940(89)90425-4
Hammarberg H Lidman O Lundberg C Eltayeb SY Gielen AW Muhallab S Svenningsson A Linda H van der Meide PH Cullheim S Olsson T Piehl F Neuroprotection by encephalomyelitis: Rescue of mechanically injured neurons and neurotrophin production by CNS-infiltrating T and natural killer cells J Neurosci 2000 20 5283 5291 10884312
Taniuchi M Clark HB Schweitzer JB Johnson EM JR Expression of nerve growth factor receptors by Schwann cells of axotomized peripheral nerves: ultrastructural location, suppression by axonal contact, and binding properties J Neurosci 1988 8 664 681 2828568
Gorio A Vergani L Ferro L Prino G Di Giulio AM Glycosaininoglycans in nerve injury: II. Effects on transganglionic degeneration and on the expression of neurotrophic factors J Neurosci Res 1996 46 572 580 8951669 10.1002/(SICI)1097-4547(19961201)46:5<572::AID-JNR6>3.0.CO;2-G
Zhou XF Rush RA McLachlan EM Differential expression of the p75 nerve growth factor receptor in glia and neurons of the rat dorsal root ganglia after peripheral nerve transection J Neurosci 1996 16 2901 2911 8622121
Copray S Kust B Emmer B Young Lin M Liem R Amor S De Vries H Floris S Boddeke EJ Deficient p75 low-affinity neurotrophin receptor expression exacerbates experimental allergic encephalomyelitis in C57/BL6 mice J Neuroimmunol 2004 148 41 53 14975585 10.1016/j.jneuroim.2003.11.008
Barrett GL The p75 neurotrophin receptor and neuronal apoptosis Prog Neurobiol 2000 61 205 229 10704998 10.1016/S0301-0082(99)00056-8
Lee R Kermani P Teng KK Hempstead BL Regulation of cell survival by secreted proneurotrophins Science 2001 294 1945 1948 11729324 10.1126/science.1065057
Beattie MS Hermann GE Rogers RC Bresnahan JC Cell death in models of spinal cord injury Prog Brain Res 2002 137 37 47 12440358
Neumann H Misgeld T Matsumuro K Wekerle H Neurotrophins inhibit major histocompatibility class II inducibility of microglia: involvement of the p75 neurotrophin receptor Proc Natl Acad Sci U S A 1998 95 5779 5784 9576961 10.1073/pnas.95.10.5779
Flugel A Matsumuro K Neumann H Klinkert WE Birnbacher R Lassman H Otten U Wekerle H Anti-inflammatory activity of nerve growth factor in experimental autoimmune encephalomyelitis: inhibition of monocyte transendothelial migration Eur J Immunol 2001 31 11 22 11169433 10.1002/1521-4141(200101)31:1<11::AID-IMMU11>3.0.CO;2-G
Ye X Mehlen P Rabizadeh S VanArsdale T Zhang H Shin H Wang JJ Leo E Zapata J Hauser CA Reed JC Bredesen DE TRAF family proteins interact with the common neurotrophin receptor and modulate apoptosis induction J Biol Chem 1999 274 30202 30208 10514511 10.1074/jbc.274.42.30202
Dempsey PW Doyle SE He JQ Cheng G The signaling adaptors and pathways activated by TNF superfamily Cytokine Growth Factor Rev 2003 14 193 209 12787559 10.1016/S1359-6101(03)00021-2
Mamidipudi V Li X Wooten MW Identification of interleukin 1 receptor-associated kinase as a conserved component in the p75-neurotrophin receptor activation of nuclear factor-kappa B J Biol Chem 2002 277 28010 28018 12034707 10.1074/jbc.M109730200
Bake S Sohrabji F 17b-estradiol differentially regulates blood brain barrier permeability in young and aging female rats Endocrinology 2004 145 5471 5475 15471968 10.1210/en.2004-0984
Thornberry NA Lazebnik Y Caspases: enemies within Science 1998 281 1312 1316 9721091 10.1126/science.281.5381.1312
Turner CP Perez-Polo JR Regulation of the low affinity receptor for nerve growth factor, p75NGFR, in the olfactory system of neonatal and adult rat Int J Dev Neurosci 1992 10 343 359 1492589 10.1016/0736-5748(92)90025-U
Kallunki T Su B Tsigelny I Sluss HK Derijard B Moore G Davis R Karin M JNK2 contains a specificity-determining region responsible for efficient c-Jun binding and phosphorylation Genes & Dev 1994 8 2996 3007 8001819
Coffey ET Smiciene G Hongisto V Cao J Brecht S Herdegen T Courtney MJ c-Jun N-Terminal Protein Kinase (JNK) 2/3 Is specifically activated by stress, mediating c-Jun activation, in the presence of constitutive JNK1 activity in cerebellar neurons J Neurosci 2002 22 4335 4345 12040039
Cao Z Henzel WJ Gao X IRAK: a kinase associated with the interleukin-1 receptor Science 1996 271 1128 1131 8599092
Cao Z Xiong J Takeuchi M Kurama T Goeddel DV TRAF6 is a signal transducer for interleukin-1 Nature 1996 383 443 446 8837778 10.1038/383443a0
Vereker E O'Donnell E Lynch MA The inhibitory effect of interleukin-1beta on long-term potentiation is coupled with increased activity of stress-activated protein kinases J Neurosci 2000 20 6811 6819 10995825
Moon SK Cha BY Kim CH ERK1/2 mediates TNF-alpha-induced matrix metalloproteinase-9 expression in human vascular smooth muscle cells via the regulation of NF-kappaB and AP-1: involvement of the ras dependent pathway J Cell Physiol 2004 198 417 427 14755547 10.1002/jcp.10435
Plummer SM Holloway KA Manson MM Munks RJ Kaptein A Farrow S Howells L Inhibition of cyclo-oxygenase 2 expression in colon cells by the chemopreventive agent curcumin involves inhibition of NF-kappaB activation via the NIK/IKK signalling complex Oncogene 1999 18 6013 6020 10557090 10.1038/sj.onc.1202980
Johnson D Lanahan A Buck CR Sehgal A Morgan C Mercer E Bothwell M Chao M Expression and structure of the human NGF receptor Cell 1986 47 545 554 3022937 10.1016/0092-8674(86)90619-7
Radeke MJ Misko TP Hsu C Herzenberg LA Shooter EM Gene transfer and molecular cloning of the rat nerve growth factor receptor Nature 1987 325 593 597 3027580 10.1038/325593a0
Casaccia-Bonnefil P Gu C Khursigara G Chao MV p75 neurotrophin receptor as a modulator of survival and death decisions Microsc Res Tech 1999 45 217 224 10383114 10.1002/(SICI)1097-0029(19990515/01)45:4/5<217::AID-JEMT5>3.0.CO;2-5
Casaccia-Bonnefil P Carter BD Dobrowsky RT Chao MV Death of oligodendrocytes mediated by the interaction of nerve growth factor with its receptor p75 Nature 1996 383 716 719 8878481 10.1038/383716a0
Frade JM Barde YA Nerve growth factor: two receptors, multiple functions Bioessays 1998 20 137 145 9631659 10.1002/(SICI)1521-1878(199802)20:2<137::AID-BIES6>3.0.CO;2-Q
Rabizadeh S Oh J Zhong LT Yang J Bitler CM Butcher LL Bredesen DE Induction of apoptosis by the low-affinity NGF receptor Science 1993 261 345 348 8332899
Frade JM Rodriguez-Tebar A Barde YA Induction of cell death by endogenous nerve growth factor through its p75 receptor Nature 1996 383 166 168 8774880 10.1038/383166a0
Yoon SO Casaccia-Bonnefil P Carter B Chao MV Competitive signaling between TrkA and p75 nerve growth factor receptors determines cell survival J Neurosci 1998 18 3273 3281 9547236
Soilu-Hanninen M Ekert P Bucci T Syroid D Bartlett PF Kilpatrick TJ Nerve growth factor signaling through p75 induces apoptosis in Schwann cells via a Bcl-2-independent pathway J Neurosci 1999 19 4828 4838 10366617
Frago LM Leon Y de la Rosa EJ Gomez-Munoz A Varela-Nieto I Nerve growth factor and ceramides modulate cell death in the early developing inner ear J Cell Sci 1998 111 549 556 9454729
Dubreuil CI Winton MJ McKerracher L Rho activation patterns after spinal cord injury and the role of activated Rho in apoptosis in the central nervous system J Cell Biol 2003 162 233 243 12860969 10.1083/jcb.200301080
Gschwendtner A Liu Z Hucho T Bohatschek M Kalla R Dechant G Raivich G Regulation, cellular localization, and function of the p75 neurotrophin receptor (p75NTR) during the regeneration of facial motoneurons Mol Cell Neurosci 2003 24 307 322 14572455 10.1016/S1044-7431(03)00167-2
Chi OZ Barsoum S Wen Y Liu X Weiss HR 17beta-estradiol prevents blood-brain barrier disruption induced by VEGF Horm Metab Res 2004 36 272 276 15156404 10.1055/s-2004-814478
Chi OZ Liu X Weiss HR Effects of 17beta-estradiol on blood-brain barrier disruption during focal ischemia in rats Horm Metab Res 2002 34 530 534 12384831 10.1055/s-2002-34794
Johnson AB Sohrabji F Estrogen's anti-inflammatory effects on central and circulating immune cells vary with reproductive age Neurobiol Aging 2005
Kraling BM Wiederschain DG Boehm T Rehn M Mulliken JB Moses MA The role of matrix metalloproteinase activity in the maturation of human capillary endothelial cells in vitro J Cell Sci 1999 112 1599 1609 10212153
Puyraimond A Fridinan R Lemesle M Arbeille B Menashi S MMP-2 colocalizes with caveolae on the surface of endothelial cells Exp Cell Res 2001 262 28 36 11120602 10.1006/excr.2000.5069
Romanic AM White RF Arleth AJ Ohlstein EH Barone FC Matrix metalloproteinase expression increases after cerebral focal ischemia in rats: inhibition of matrix metalloproteinase-9 reduces infarct size Stroke 1998 29 1020 1030 9596253
Herrmann JL Menter DG Hamada J Marchetti D Nakajima M Nicolson GL Mediation of NGF-stimulated extracellular matrix invasion by the human melanoma low-affinity p75 neurotrophin receptor: melanoma p75 functions independently of trkA Mol Biol Cell 1993 4 1205 1216 8305739
Moalli PA Klingensmith WL Meyn LA Zyczynski HM Regulation of matrix metalloproteinase expression by estrogen in fibroblasts that are derived from the pelvic floor Am J Obstet Gynecol 2002 187 72 79 12114891 10.1067/mob.2002.124845
Sohrabji F Greene LA Miranda RC Toran-Allerand CD Reciprocal regulation of estrogen and NGF receptors by their ligands in PC12 cells J Neurobiol 1994 25 974 988 7525871 10.1002/neu.480250807
Sohrabji F Miranda RC Toran-Allerand CD Estrogen differentially regulates estrogen and nerve growth factor mRNAs in adult sensory neurons J Neurosci 1994 14 459 471 8301349
Gibbs RB Effects of estrogen on basal forebrain cholinergic neurons vary as a function of dose and duration of treatment Brain Res 1997 757 10 16 9200493 10.1016/S0006-8993(96)01432-1
Jezierski MK Sohrabji F Neurotrophin expression in the reproductively senescent forebrain is refractory to estrogen stimulation Neurobiol Aging 2001 22 309 319 11182481 10.1016/S0197-4580(00)00230-X
Ping SE Greferath U Barrett GL Estrogen treatment suppresses forebrain p75 neurotrophin receptor expression in aged, noncycling female rats J Neurosci Res 2002 69 51 60 12111815 10.1002/jnr.10273
Jover T Tanaka H Calderone A Oguro K Bennett MV Etgen AM Zukin RS Estrogen protects against global ischemia-induced neuronal death and prevents activation of apoptotic signaling cascades in the hippocampal CA1 J Neurosci 2002 22 2115 2124 11896151
Miyamoto N Mandai M Suzuma I Suzuma K Kobayashi K Honda Y Estrogen protects against cellular infiltration by reducing the expressions of E-selectin and IL-6 in endotoxininduced uveitis J Immunol 1999 163 374 379 10384138
Cuzzocrea S Santagati S Sautebin L Mazzon E Calabro G Serraino 1 Caputi AP Maggi A 17beta-estradiol antiinflammatory activity in carrageenan-induced pleurisy Endocrinology 2000 141 1455 1463 10746650 10.1210/en.141.4.1455
Waksman Y Hod I Friedman A Therapeutic effects of estradiol benzoate on development of collagen-induced arthritis (CIA) in the Lewis rat are mediated via suppression of the humoral response against denatured collagen type II (CII) Clin Exp Immunol 1996 103 376 383 8608634
Badger AM Blake SM Dodds RA Griswold DE Swift BA Rieman DJ Stroup GB Hoffman SJ Gowen M Idoxifene, a novel selective estrogen receptor modulator, is effective in a rat model of adjuvant-induced arthritis J Pharmacol Exp Ther 1999 291 1380 1386 10565864
Naslund MJ Strandberg JD Coffey DS The role of androgens and estrogens in the pathogenesis of experimental nonbacterial prostatitis J Urol 1988 140 1049 1053 3172358
Harris MT Feldberg RS Lau KM Lazarus NH Cochrane DE Expression of proinflammatory genes during estrogen-induced inflammation of the rat prostate Prostate 2000 44 19 25 10861753 10.1002/1097-0045(20000615)44:1<19::AID-PROS3>3.0.CO;2-S
Tchernitchin AN Galand P Oestrogen levels in the blood, not in the uterus, determine uterine eosinophilia and oedema J Endocrinol 1983 99 123 130 6631303
Quarmby VE Korach KS The influence of l7beta-estradiol on patterns of cell division in the uterus Endocrinology 1984 114 694 702 6697957
De M Wood GW Influence of oestrogen and progesterone on macrophage distribution in the mouse uterus J Endocrinol 1990 126 417 424 2212933
Kachkache M Acker GM Chaouat G Noun A Garabedian M Hormonal and local factors control the immunohistochemical distribution of imrnunocytes in the rat uterus before conceptus implantation: effects of ovariectomy, fallopian tube section, and injection Biol Reprod 1991 45 860 868 1805988
Kaushic C Frauendorf E Rossoll RM Richardson JM Wira CR Influence of the estrous cycle on the presence and distribution of immune cells in the rat reproductive tract Am J Reprod Immunol 1998 39 209 216 9526610
Jezierski MK Sohrabji F Region- and peptide-specific regulation of the neurotrophins by estrogen Brain Res Mol Brain Res 2000 85 77 84 11146109 10.1016/S0169-328X(00)00244-8
Jezierski MK Sturm AK Scarborough MM Sohrabji F NGF stimulation increases JNK2 phosphorylation and reduces caspase-3 activity in the olfactory bulb of estrogen-replaced animals Endocrinology 2001 142 2401 2404 11356687 10.1210/en.142.6.2401
Acarin L Gonzalez B Castellano B Neuronal, astroglial and microglial cytokine expression after an excitotoxic lesion in the immature rat brain Eur J Neurosci 2000 12 3505 3520 11029620 10.1046/j.1460-9568.2000.00226.x
Mount HT Elkabes S Dreyfus CF Black IB Differential involvement of metabotropic and p75 neurotrophin receptors in effects of nerve growth factor and neurotrophin-3 on cultured Purkinje cell survival J Neurochem 1998 70 1045 1053 9489724
Fiorentini C Guerra N Facchetti M Finardi A Tiberio L Schiaffonati L Spano P Missale C Nerve growth factor regulates dopamine D(2) receptor expression in prolactinoma cell lines via p75(NGFR)-mediated activation of nuclear factor-kappaB Mol Endocrinol 2002 16 353 366 11818506 10.1210/me.16.2.353
Pehar M Cassina P Vargas MR Castellanos R Viera L Beckman JS Estevez AG Barbeito L Astrocytic production of nerve growth factor in motor neuron apoptosis: implications for amyotrophic lateral sclerosis J Neurochem 2004 89 464 473 15056289
Maier SE Cramer JA West JR Sohrabji F Alcohol exposure during the first two trimersters equivalent alters granule cell number and neurotrophin expression in the developing rat olfactory bulb J Neurobiol 1999 41 414 423 10526319 10.1002/(SICI)1097-4695(19991115)41:3<414::AID-NEU9>3.0.CO;2-F
|
16156894
|
PMC1239918
|
CC BY
|
2021-01-04 16:03:47
|
no
|
BMC Neurosci. 2005 Sep 12; 6:58
|
utf-8
|
BMC Neurosci
| 2,005 |
10.1186/1471-2202-6-58
|
oa_comm
|
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-601616228110.1186/1471-2202-6-60Research ArticleDifferential effects of intragastric acid and capsaicin on gastric emptying and afferent input to the rat spinal cord and brainstem Holzer Peter [email protected] Evelin [email protected] Rufina [email protected] Department of Experimental and Clinical Pharmacology, Medical University of Graz, Universitätsplatz 4, A-8010 Graz, Austria2005 14 9 2005 6 60 60 6 5 2005 14 9 2005 Copyright © 2005 Holzer et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Hydrochloric acid (HCl) is a potential threat to the integrity of the gastric mucosa and is known to contribute to upper abdominal pain. We have previously found that gastric mucosal challenge with excess HCl is signalled to the rat brainstem, but not spinal cord, as visualized by expression of c-fos messenger ribonucleic acid (mRNA), a surrogate marker of neuronal excitation. This study examined whether gastric mucosal exposure to capsaicin, a stimulant of nociceptive afferents that does not damage the gastric mucosa, is signalled to both brainstem and spinal cord and whether differences in the afferent signalling of gastric HCl and capsaicin challenge are related to different effects on gastric emptying.
Results
Rats were treated intragastrically with vehicle, HCl or capsaicin, activation of neurons in the brainstem and spinal cord was visualized by in situ hybridization autoradiography for c-fos mRNA, and gastric emptying deduced from the retention of intragastrically administered fluid. Relative to vehicle, HCl (0.5 M) and capsaicin (3.2 mM) increased c-fos transcription in the nucleus tractus solitarii by factors of 7.0 and 2.1, respectively. Capsaicin also caused a 5.2-fold rise of c-fos mRNA expression in lamina I of the caudal thoracic spinal cord, although the number of c-fos mRNA-positive cells in this lamina was very small. Thus, on average only 0.13 and 0.68 c-fos mRNA-positive cells were counted in 0.01 mm sections of the unilateral lamina I following intragastric administration of vehicle and capsaicin, respectively. In contrast, intragastric HCl failed to induce c-fos mRNA in the spinal cord. Measurement of gastric fluid retention revealed that HCl suppressed gastric emptying while capsaicin did not.
Conclusion
The findings of this study show that gastric mucosal exposure to HCl and capsaicin is differentially transmitted to the brainstem and spinal cord. Since only HCl blocks gastric emptying, it is hypothesized that the two stimuli are transduced by different afferent pathways. We infer that HCl is exclusively signalled by gastric vagal afferents whereas capsaicin is processed both by gastric vagal and intestinal spinal afferents.
==== Body
Background
Gastric acid-related diseases are among the most prevalent mucosal disorders of the upper gastrointestinal tract. There is also evidence that hydrochloric acid (HCl) contributes to the pain associated with gastro-oesophageal reflux and peptic ulcer disease as well as to non-cardiac chest pain and functional dyspepsia [1-3]. Gastric chemonociception evoked by exposure of the rat stomach to excess HCl is mediated by vagal afferent neurons, given that the visceromotor pain response to intragastric (IG) administration of HCl is abolished by bilateral vagotomy, whereas the visceromotor response to distension remains unaltered [4]. This finding is consistent with our observation that IG administration of HCl is signalled to the nucleus tractus solitarii (NTS) of the rat brainstem, the central termination area of vagal afferents, as visualized by expression of messenger ribonucleic acid (mRNA) for the immediate early gene c-fos, whereas no induction of c-fos mRNA is seen in the spinal cord [5,6]. Similarly, gastric challenge with ammonium hydroxide induces c-fos mRNA and protein only in the brainstem, but not spinal cord, of the rat [7]. While both HCl and ammonium hydroxide injure the gastric mucosa at concentrations that cause near-maximal translation of the c-fos gene in the NTS [7], capsaicin is a chemical that excites gastrointestinal afferent neurons [8-10] without causing damage to the rat stomach [11]. This is because capsaicin stimulates afferent neurons by gating transient receptor potential ion channels of vanilloid type 1 (TRPV1), which are expressed by both vagal and spinal afferent neurons innervating the rat stomach and intestine [12-16].
The overall aim of this exploratory study was to test whether gastric mucosal challenge with capsaicin and excess HCl is differentially transmitted to the rat brainstem and spinal cord and whether the differential processing of the two stimuli takes place at the level of the upper gastrointestinal tract. Two sets of experiments were performed to address these questions. In the first study, IG administered capsaicin and HCl were compared in their effects on the expression of c-fos mRNA in the NTS and in the caudal thoracic spinal cord which receives the densest afferent input from the rat stomach [17,18]. It was in particular investigated whether IG administration of HCl and capsaicin has a distinct effect on neurons in specific laminae and nuclei of the dorsal spinal cord. The concentrations of HCl (0.5 M) and capsaicin (0.64 and 3.2 M) tested in these experiments were selected from previous experiments. Exposure of the rat gastric mucosa to HCl (0.5 M) causes a distinctive but submaximal induction of c-fos mRNA in the brainstem [5], while IG administration of 0.64 mM capsaicin is maximally effective in increasing gastric mucosal blood flow in a sensory neuron-dependent manner [19].
Since it was found that the afferent signalling of HCl and capsaicin to the NTS and spinal cord is different, the aim of the second study was to examine whether HCl and capsaicin influence gastric motility and emptying in a differential manner. It was reasoned that the magnitude of the c-fos response in the NTS and spinal cord depends both on the concentration of the chemicals and the duration of their presence in the stomach. It has previously been found that IG administration of excess HCl inhibits gastric emptying and alters intragastric pressure [5,20]. The findings of this study reveal that, unlike excess HCl, capsaicin does not inhibit gastric emptying. It is hypothesized, therefore, that gastric HCl challenge is exclusively signalled to the brainstem via vagal afferents, because it is retained in the stomach for a prolonged period of time, whereas both gastric vagal and duodenal spinal afferents respond to IG administration of capsaicin, the emptying of which into the duodenum is not retarded.
Results
Effects of HCl and capsaicin to induce c-fos mRNA in the NTS and spinal cord (study 1)
As illustrated previously [5,21], IG administration of 0.5 M HCl caused many neurons in the NTS to express c-fos mRNA when compared with IG administration of physiological saline. The number of c-fos mRNA-positive cells per section seen after IG administration of HCl was 7.0 times larger than after IG administration of physiological saline (Figure 1). The number of c-fos mRNA-positive neurons per NTS section counted after IG administration of vehicle tended to be higher than after administration of physiological saline, although this effect was statistically not significant (Figure 1). Relative to vehicle, capsaicin (0.64 and 3.2 mM) enhanced the number of c-fos mRNA-positive neurons per NTS section, this effect depending on the concentration of the drug. As can be seen in Figure 1, only the concentration of 3.2 mM capsaicin was able to significantly increase the induction of c-fos mRNA by a factor of 2.1. The distribution of c-fos mRNA-positive cells in the NTS after IG exposure to HCl (0.5 M) and capsaicin (3.2 mM) was uneven, the highest number of activated cells occurring in the ventromedial part of the NTS [22].
In agreement with previous findings [5], IG exposure to 0.5 M HCl failed to induce any expression of c-fos mRNA in the dorsal half of the caudal thoracic spinal cord. Thus, the total number of c-fos mRNA-positive cells per dorsal spinal cord section counted after IG exposure to physiological saline was 1.30 ± 0.29 (n = 4) and after IG exposure to HCl 1.33 ± 0.18 (n = 4). This lack of effect of HCl was also seen when the distribution of c-fos mRNA-positive cells to LI, LII, LIII, LIV, LV, AX and IMLN after IG administration of HCl was compared with that after IG administration of physiological saline (Figure 2A). IG administration of capsaicin (3.2 mM) likewise failed to significantly increase the expression of c-fos mRNA in the dorsal spinal cord, given that the total number of c-fos mRNA-positive cells per section counted after IG exposure to vehicle was 2.19 ± 0.32 (n = 6) and after IG exposure to capsaicin was 1.94 ± 0.23 (n = 7). Analysis of the distribution of c-fos mRNA-positive cells to LI, LII, LIII, LIV, LV, AX and IMLN revealed, however, that capsaicin caused a significant 5.2-fold increase of c-fos mRNA expression in LI, which went in parallel with a significant decrease in the formation of c-fos mRNA in LIII and LIV (Figure 2B). It should be noted that the level of c-fos transcription was very low, as typically less than 0.7 c-fos mRNA-positive cells per lamina were counted in the 0.01 mm sections of the unilateral dorsal spinal cord (Figure 2). For this reason, the experiments involving HCl and capsaicin were strictly run in parallel with those involving the respective control/vehicle solution (Figure 2).
Effects of HCl and capsaicin on intragastric pressure and gastric fluid recovery (study 2)
The baseline IGP measured before administration of any medium was between 400 and 500 Pa [20]. IG injection of a 2 ml fluid bolus increased IGP to a level whose magnitude was independent of whether the injected fluid was saline, HCl (0.35 M), vehicle or capsaicin (3.2 mM) as determined 2–3 min post-injection (Figure 3A). In contrast, the time course of the subsequent decline of IGP depended on the nature of the administered medium. Following injection of saline or vehicle, IGP decreased at a significantly faster rate than after injection of HCl or capsaicin, respectively (Figure 3B). Thus, in HCl-and capsaicin-exposed stomachs IGP did not significantly fall during the 30 min observation period post-injection, whereas in NaCl- and vehicle-exposed stomachs IGP significantly decreased to levels of about 65 % of the IGP measured 2–3 min post-injection (Figure 3B). Another effect of HCl was to enhance gastric fluid retention as deduced from a 100 % recovery of the injected fluid volume from the stomach 30 min post-injection (Figure 3C). In contrast, 30 min after administration of saline, vehicle or capsaicin only 30–60% of the injected fluid volume was regained (Figure 3C).
Discussion
The results of the current study show that IG administration of HCl and capsaicin to rats generates differential inputs to the NTS and thoracic spinal cord, which is associated with differential effects on gastric emptying. As described previously [5,7], gastric signalling to the brainstem and spinal cord was visualized by expression of the inducible gene c-fos at the mRNA level, a method that has been established as a standard tool in functional neuroanatomy to delineate the stimulus-evoked activation of neurons [23,24]. Transcription of the c-fos gene begins within minutes after neuronal excitation [23,24] and in the NTS appears to be maximal 45 min after gastric HCl challenge [5]. Although exposure to HCl (0.35–0.7 M) induces gastric mucosal injury in a concentration-related manner, there is evidence that the afferent signalling of gastric HCl challenge is not directly related to the formation of mucosal injury, because the expression of the c-fos gene in the NTS can be stimulated by IG concentrations of HCl that induce little, if any, epithelial damage [5,7]. It has, therefore, been hypothesized [5,7] that a massive increase of the H+ ion gradient across the gastric mucosal barrier is per se sufficient to drive H+ ions into the lamina propria where they can excite vagal afferent nerve fibres either directly [25,26] or indirectly via neuroactive factors released in the tissue.
The topographical distribution of c-fos mRNA-positive cells in the NTS was uneven but similar after IG administration of HCl and capsaicin. As previously shown by immunohistochemistry [22], the highest number of HCl-activated neurons was seen in the ventromedial part of the NTS. Apart from vagal gastric input [27], this area of the NTS also receives input from spinal lamina I neurons via the spinosolitary tract [28-30]. The relative contribution of the vagal and spinal inputs to this part of the NTS following chemical stimulation of the stomach remains to be determined.
The present study confirms that excess gastric HCl fails to induce c-fos mRNA and c-Fos protein in the dorsal horn of the posterior thoracic spinal cord [5,7] which receives the densest input from afferent neurons innervating the rat stomach [17,18]. Similar findings have been made following exposure of the rat gastric mucosa to ammonium hydroxide [7]. It thus appeared as if gastric challenge with noxious chemicals is signalled by vagal afferents only, a conjecture that was rejected by the current finding that afferent input from the capsaicin-exposed stomach is sent both to the spinal cord and brainstem. This result is in keeping with the expression of TRPV1, the capsaicin receptor, by both vagal and spinal afferent neurons innervating the rat gastrointestinal tract [12-16]. Retrograde tracing has shown that 80 and 71 % of the nodose and dorsal root ganglion neurons supplying the rat stomach, respectively, express TRPV1 [16]. When directly applied to the somata, capsaicin excites 90 % of the dorsal root ganglion neurons and 59 % of the nodose ganglion neurons projecting to the rat stomach [31]. While TRPV1 is easy to detect in the nodose ganglia, the level of TRPV1 expression in most vagal afferent nerve fibres within the stomach is below the immunohistochemical detection threshold [12]. This instance could explain why IG administration of capsaicin induces comparatively little expression of c-fos mRNA in the NTS. We do not think that the small effect of capsaicin is due to inadequate dosing, because it has previously been found that IG administration of 0.64 mM capsaicin is maximally effective in increasing gastric mucosal blood flow in a sensory neuron-dependent manner [19].
Like HCl, capsaicin administered into the rat stomach failed to significantly enhance the overall expression of c-fos mRNA in the dorsal half of the posterior thoracic spinal cord. However, regional analysis revealed that capsaicin caused neurons in the superficial lamina I of the dorsal horn to express c-fos mRNA, an effect that was not seen following IG administration of HCl. This finding obtained with capsaicin is consistent with the projection of visceral afferent neurons to lamina I and the superficial part of lamina II as well as to lamina V and area X of the rat and cat spinal cord [17,32]. Our data indicate that administration of capsaicin into the gastric lumen activates sensory neurons that project primarily to lamina I of the spinal cord. In view of this finding it can be ruled out that c-fos expression is an inadequate method to visualize chemoreceptor signalling from the gastric lumen to the spinal cord and that the failure of gastric HCl challenge to induce c-fos mRNA in the spinal cord represents a false negative result. Although the neural sensors whereby excess HCl is detected in the gastric lumen are not known, it is conceivable that both TRPV1 and acid-sensing ion channels (ASICs) such as ASIC3 are involved [33]. Since both TRPV1 and ASIC3 are expressed by a majority of dorsal root ganglion neurons supplying the rat stomach [16], it appears unlikely that IG HCl fails to induce c-fos mRNA in the spinal cord because the respective afferents do not bear the appropriate acid sensors.
Compared with the number of neurons expressing c-fos mRNA in the NTS, the number of c-fos mRNA-positive neurons in 0.01 mm sections of the spinal cord was very small. This is likely to reflect that the afferent input from the rat stomach to the spinal cord is minor relative to the spinal input from somatic tissues and that electrophysiologically characterized spinal afferents hardly innervate the mucosa of the gastrointestinal tract [34,35]. The low level of c-fos transcription in the laminae of the dorsal spinal cord made it compulsory to run the experiments involving HCl and capsaicin strictly in parallel with those involving the respective control/vehicle solution. We hypothesize that the apparently different distribution of c-fos mRNA-positive cells within the dorsal spinal cord of control rats, as shown in the two panels of Figure 2, may not only reflect inter-experiment variability but also the different nature of the control/vehicle solution: while the vehicle for HCl was physiological saline, the vehicle for capsaicin was saline containing ethanol and Tween 80.
The effect of IG capsaicin to increase c-fos mRNA induction in lamina I of the spinal cord was associated with a significant decrease in c-fos mRNA expression in laminae III and IV. We do not have any straight-forward explanation for this observation. Since laminae III and IV do not seem to receive direct input from the viscera [17,32], we hypothesize that the reduction of c-fos mRNA formation in these laminae is an indirect effect of capsaicin. Conceivably, visceral afferent input via lamina I neurons activates inhibitory pathways that depress the excitability of lamina III and IV neurons.
The afferent signalling of gastric mucosal exposure to capsaicin and excess HCl is determined not only by the concentration of the noxious chemical but also by the duration of its presence in the gastric lumen. It has previously been found that, relative to saline, IG administration of HCl to anaesthetized rats prolongs fluid retention in the stomach and delays adaptation of IGP [7,20]. HCl-induced gastric fluid retention results from inhibition of gastric emptying and enhanced gastric fluid, bicarbonate and mucus secretion [20,36], but analysis of the gastric contents was beyond the scope of this study. HCl-evoked inhibition of gastric emptying is mediated by neural reflexes that are initiated both in the stomach and duodenum [20,37-40]. The concentration of HCl tested for its gastropyloric motor effects in anaesthetized rats was reduced to 0.35 M, because anaesthesia weakens the gastric mucosal barrier to HCl and the IG concentration of HCl (0.5 M) tested for its effect on central c-fos expression induces extensive injury in anaesthetized rats but causes minor gastric damage in conscious animals [5]. As the experiments revealed, gastric exposure to HCl and capsaicin modified gastropyloric motility in a differential manner. While gastric emptying was blocked by HCl but left unaltered by capsaicin, the adaptation of IGP was prevented by both HCl and capsaicin. The effect of capsaicin to delay IGP adaptation may be related to its ability to induce contraction or relaxation of the rat gastric musculature, the type of response depending on the dose of capsaicin and the gastric region and muscle layer under study [19,41-43].
With regard to the disparate effect of IG HCl and capsaicin on spinal c-fos expression it was particularly important to note that, unlike HCl, capsaicin failed to enhance gastric fluid recovery, which means that gastric emptying occurred unimpaired and IG administered capsaicin was quickly transported into the upper small intestine. It could therefore be argued that the capsaicin-evoked c-fos response in the NTS, which was smaller than that to HCl, and spinal cord are due to capsaicin-evoked excitation of both gastric and intestinal afferents, whereas the excitatory effect of HCl is largely confined to gastric afferents. The failure of gastric HCl challenge to induce c-fos expression in the spinal cord cannot be explained by the reported ability of vagal afferents to activate descending pathways and thereby inhibit afferent input to the spinal cord [44,45], because bilateral chronic vagotomy fails to reveal any increase in spinal c-fos mRNA induction due to gastric HCl challenge [5]. There are other ways to explain the differential ability of IG HCl and capsaicin to induce c-fos mRNA in the NTS and spinal cord, but the analysis of these factors was beyond the scope of this study. For instance, there is evidence that capsaicin is little absorbed in the gastric wall [19], which would also explain why IG administered capsaicin is comparatively weak in stimulating gastric vagal afferents projecting to the NTS, whereas capsaicin transported to the upper small intestine may more easily reach and stimulate spinal afferent nerve terminals in the intestinal lamina propria.
Due to its explorative nature, the current study has its limitations. Thus, the differential effect of IG administered capsaicin and HCl on vagal and spinal afferent pathways is likely to depend not only on the gastric emptying rate and the absorption kinetics of HCl and capsaicin but also on the extent of mucosal injury and the magnitude of mucosal blood flow. While the integrity of the gastric mucosa is disturbed only by HCl [5,7,46] but not by capsaicin [11], gastric mucosal blood flow is elevated by both capsaicin [19] and backdiffusing HCl [46].
Conclusion
Gastric challenge with HCl and capsaicin is differentially signalled to the NTS and spinal cord, which indicates that the two stimuli are processed by disparate nociceptive afferent pathways. Since HCl inhibits gastric emptying, whereas capsaicin does not, it is inferred that the HCl-evoked afferent input to the NTS is transmitted by vagal afferents in the stomach, while the activation of NTS and spinal lamina I neurons by capsaicin is mediated both by vagal afferents in the stomach and by spinal afferents in the upper small intestine. Further experimentation is needed to determine how these findings relate to gastric chemonociception. In agreement with our c-fos data, nociception elicited by excess gastric HCl is mediated by vagal afferent neurons [4], and it awaits to be examined which afferent pathways relay nociception evoked by gastric capsaicin. Mechanonociception evoked by distension of the stomach is brought about by spinal afferents [4], although expression of c-fos is seen both in the spinal cord and, to a larger extent, in the brainstem [47].
Methods
Animals
The study was approved by an ethical committee at the Federal Ministry of Education, Science and Culture of the Republic of Austria and conducted according to the Directive of the European Communities Council of 24 November 1986 (86/609/EEC). The experiments were designed in such a way that the number of animals used and their suffering was minimized. Female age-matched Sprague-Dawley rats (Abteilung für Labortierkunde und -genetik, Medical University of Vienna, Himberg, Austria) weighing 180–220 g were used. They were housed in groups of four in plastic transparent cages under standard conditions; lights were on from 6:00 AM until 6:00 PM.
Experimental protocols
All experiments took place during the light phase between 8:00 and 12:00 AM. Twenty hours before the begin of the experiments the rats were deprived of food to ensure that the stomach was empty by the time of the experiments, while water was available ad libitum throughout this preparatory phase. In addition, the rats were placed in groups of two on a floor grid to prevent coprophagy. Two studies with different experimental protocols were carried out.
Study 1 was conducted with non-anaesthetized animals. Physiological saline (0.15 M NaCl), HCl (0.5 M), capsaicin (0.64 and 3.2 mM) or its vehicle were administered IG at a volume of 10 ml/kg through a soft infant feeding tube (outer diameter 2.2 mm; Portex, Hythe, UK). After 45 min the rats were euthanized by intraperitoneal injection of an overdose of pentobarbital (200 mg/kg; Intervet, Vienna, Austria) and their brainstem and spinal cord removed quickly. Capsaicin (Sigma, Vienna, Austria) was dissolved in a medium containing 10 % Tween 80, 10 % ethanol and 80 % physiological saline to give stock solutions of 2 and 10 mg/ml (6.4 and 32 mM) capsaicin. These stock solutions were then diluted with physiological saline to give test solutions of 0.64 and 3.2 mM capsaicin, respectively. The vehicle for capsaicin consisted of 1 % Tween 80, 1 % ethanol and 98 % physiological saline.
Study 2 was performed with animals that were anaesthetized with phenobarbital (230 mg/kg intraperitoneally; Sigma) and placed on a thermostated table to maintain their rectal temperature at 37 degrees Celsius [20]. The rats were then fitted with a tracheal cannula to facilitate spontaneous breathing. A cannula in the left jugular vein was used for continuous infusion of physiological saline (1.5 ml/h) to avoid dehydration. After a midline laparotomy an IG catheter (outer diameter: 2.2 mm) was inserted in the stomach via the oesophagus, and the stomach flushed [20]. With its tip being positioned in the corpus region, the catheter was used to record intragastric pressure (IGP) via a pressure transducer as well as to inject fluid into and drain it from the stomach [20]. This method of IGP measurement has been described and validated in a previous study [20]. After an equilibration period of 30 min, a 2 ml fluid bolus was slowly injected into the stomach over a period of 5 s and left in the stomach for a period of 30 min after which the stomach was drained and the weight of the recovered fluid determined. The recovery of fluid from the stomach (an indirect measure of gastric emptying) was expressed as a percentage of the weight of the fluid administered into the stomach [20]. Each rat was subjected to 4 injection/recovery trials at intervals of 15 min during which the stomach was left empty. First, two priming trials with saline were carried out, followed either by a test trial with saline and a test trial with HCl (0.35 M) or by a test trial with vehicle and a test trial with capsaicin (3.2 mM). IGP was averaged for the periods of 2–3 min, 9–10 min and 29–30 min post-injection. Since as described before [20] the peak rise of IGP post-injection varied because of differences in injection speed, the IGP averaged during the period of 2–3 min post-injection was taken as 100 % and the IGP recorded during the subsequent observation periods expressed as a percentage of that reference value.
In situ hybridization autoradiography
The brainstem and spinal cord were quickly removed and frozen on powdered dry ice. Coronal sections (0.01 mm) were cut serially from the brainstem at the rostrocaudal extension of the area postrema and the caudal thoracic spinal cord with a cryostat [5,6,22]. Every sixth section was processed for in situ hybridization with an oligodeoxyribonucleotide probe labelled at the 3' end with [35S]deoxyadenosine 5'(α-thio)triphosphate as described previously [6]. The sections were dipped in Ilford K5 photographic emulsion and, after 18–25 days of exposure in sealed boxes at 4 degrees Celsius, the autoradiograms were developed and the sections counterstained with haematoxylin and coverslipped [6]. The specificity of the procedure was proved by the absence of any hybridization signal when control sections were hybridized with a mixture of labelled probe with a 100-fold excess of unlabelled ('cold') probe.
The autoradiograms were examined in a coded manner with a light microscope (Axiophot, Zeiss, Oberkochen, Germany) coupled to a computerized image analysis system (Imaging, St. Catharines, Ontario, Canada). Cells were considered c-fos mRNA-positive when their grain density was at least 10 times higher than the background [6]. In order to enhance the reliability of the quantitative results, 5 brainstem sections and 7–10 spinal cord sections from each animal were evaluated. These sections were selected such that they were 0.05 mm apart from each other so as to avoid that the same cells were counted twice. The c-fos mRNA-positive cells per section were counted unilaterally in the NTS at the level of the area postrema and in the dorsal half of the spinal cord at the caudal thoracic level (T8–T12). These structures were identified according to Molander and Grant [48] and Paxinos and Watson [49]. In the dorsal half of the spinal cord, the distribution of c-fos mRNA-positive cells to laminae I-V (LI-LV), area X (AX) around the central canal and the intermediolateral nucleus (IMLN) was evaluated according to the rat spinal cytoarchitecture described by Molander et al. [50] and Molander and Grant [48]. Particular care was taken to count equivalent sections in the rostro-caudal axis when different treatment groups were compared with each other [5]. All counts per section for a given area in each animal were averaged to give the number of c-fos mRNA-positive cells per section in that particular area. These average values from each animal were then used to calculate the mean number of positive cells per section in the respective areas of each experimental group [5,6].
Statistics
All data are presented as means ± SEM, n referring to the number of rats in the respective group. After the normal distribution of the experimental data was revealed by the Kolmogorov Smirnov test, significant differences between the experimental groups were evaluated with Student's t test, one-way analysis of variance or, when repeated measurements were taken, one-way analysis of variance for repeated measures followed by Dunn's test (SPSS 11.5, SPSS, Chicago, IL). Probability values of P < 0.05 were regarded as significant.
List of abbreviations used
ASIC, acid-sensing ion channel; ANOVA, analysis of variance; AX, area X around the central canal of the spinal cord; HCl, hydrochloric acid; IG, intragastric; IGP, intragastric pressure; LI-LV, laminae I-V of the dorsal spinal horn; IMLN, intermediolateral nucleus; mRNA, messenger ribonucleic acid; NTS, nucleus tractus solitarii; Pa, Pascal; TRPV1, transient receptor potential ion channel of vanilloid type 1
Authors' contributions
PH conceived the study following discussion with EP and RS, coordinated the experiments and drafted the manuscript in close consultation with RS. RS supervised the in situ hybridization experiments, and EP performed the gastric motility study, carried out the statistical analysis of the data and drafted the figures. All authors read and approved the final manuscript.
Acknowledgements
This study was supported by the Austrian Scientific Research Funds (FWF grant L25-B05) and the Zukunftsfonds Steiermark (grant 262). The authors thank Milana Jocic for her expert performance of the in situ hybridization experiments.
Figures and Tables
Figure 1 Number of c-fos mRNA-positive cells per section (0.01 mm) in the unilateral NTS determined 45 min after IG administration of NaCl (0.15 M), HCl (0.5 M), vehicle (Veh) and capsaicin (Cap, 0.64 and 3.2 mM). Means + SEM, n as indicated. * P < 0.05 versus Veh, ** P < 0.01 versus NaCl.
Figure 2 Number of c-fos mRNA-positive cells per section (0.01 mm) in various laminae and areas of the unilateral dorsal half of the caudal thoracic spinal cord determined 45 min after IG administration of (A) NaCl (0.15 M), HCl (0.5 M), (B) vehicle (Veh) and capsaicin (Cap, 3.2 mM). The graphs show the counts for laminae I-V (LI-LV), area X (AX) and the intermediolateral nucleus (IMLN). Means + SEM, n as indicated. * P < 0.05, ** P < 0.01 versus Veh.
Figure 3 Effects of intragastric injection of NaCl (0.15 M), HCl (0.35 M), vehicle (Veh) and capsaicin (Cap, 3.2 mM) on the (A) initial rise of intragastric pressure (IGP), (B) subsequent time course of IGP and (C) fluid recovery from the stomach. NaCl, HCl, Veh and Cap were injected as 2 ml bolus. The values shown in panel A represent the IGP reached 2–3 min post-injection. The IGP values measured during the periods 9–10 min and 29–30 min post-injection (panel B) are expressed as a percentage of the initial IGP rise recorded 2–3 min post-injection, and the gastric volume recovery (panel C) measured 30 min after bolus injection is expressed as a percentage of the injection volume (2 ml). Means + SEM, n = 6–13. ** P < 0.01 versus IGP measured 2–3 min post-injection; ++ P < 0.01 versus NaCl.
==== Refs
Kang JY Yap I Acid and gastric ulcer pain J Clin Gastroenterol 1991 13 514 516 1744385
Quartero AO Numans ME de Melker RA Hoes AW de Wit NJ Dyspepsia in primary care: acid suppression as effective as prokinetic therapy. A randomized clinical trial Scand J Gastroenterol 2001 36 942 947 11521984 10.1080/003655201750305459
Sarkar S Thompson DG Woolf CJ Hobson AR Millane T Aziz Q Patients with chest pain and occult gastroesophageal reflux demonstrate visceral pain hypersensitivity which may be partially responsive to acid suppression Am J Gastroenterol 2004 99 1998 2006 15447763 10.1111/j.1572-0241.2004.40174.x
Lamb K Kang YM Gebhart GF Bielefeldt K Gastric inflammation triggers hypersensitivity to acid in awake rats Gastroenterology 2003 125 1410 1418 14598257 10.1016/j.gastro.2003.07.010
Schuligoi R Jocic M Heinemann A Schöninkle E Pabst MA Holzer P Gastric acid-evoked c-fos messenger RNA expression in rat brainstem is signaled by capsaicin-resistant vagal afferents Gastroenterology 1998 115 649 660 9721162
Michl T Jocic M Heinemann A Schuligoi R Holzer P Vagal afferent signaling of a gastric mucosal acid insult to medullary, pontine, thalamic, hypothalamic and limbic, but not cortical, nuclei of the rat brain Pain 2001 92 19 27 11323123 10.1016/S0304-3959(00)00467-X
Danzer M Jocic M Samberger C Painsipp E Bock E Pabst MA Crailsheim K Schicho R Lippe IT Holzer P Stomach-brain communication by vagal afferents in response to luminal acid backdiffusion, gastrin, and gastric acid secretion Am J Physiol Gastrointest Liver Physiol 2004 286 G403 G411 14592947 10.1152/ajpgi.00308.2003
Su X Wachtel RE Gebhart GF Capsaicin sensitivity and voltage-gated sodium currents in colon sensory neurons from rat dorsal root ganglia Am J Physiol 1999 277 G1180 G1188 10600815
Blackshaw LA Page AJ Partosoedarso ER Acute effects of capsaicin on gastrointestinal vagal afferents Neuroscience 2000 96 407 416 10683581 10.1016/S0306-4522(99)00547-3
Rong W Hillsley K Davis JB Hicks G Winchester WJ Grundy D Jejunal afferent nerve sensitivity in wild-type and TRPV1 knockout mice J Physiol 2004 560 867 881 15331673 10.1113/jphysiol.2004.071746
Lippe IT Pabst MA Holzer P Intragastric capsaicin enhances rat gastric acid elimination and mucosal blood flow by afferent nerve stimulation Br J Pharmacol 1989 96 91 100 2924079
Patterson LM Zheng H Ward SM Berthoud HR Vanilloid receptor (VR1) expression in vagal afferent neurons innervating the gastrointestinal tract Cell Tissue Res 2003 311 277 287 12658436
Ward SM Bayguinov J Won KJ Grundy D Berthoud HR Distribution of the vanilloid receptor (VR1) in the gastrointestinal tract J Comp Neurol 2003 465 121 135 12926020 10.1002/cne.10801
Horie S Yamamoto H Michael GJ Uchida M Belai A Watanabe K Priestley JV Murayama T Protective role of vanilloid receptor type 1 in HCl-induced gastric mucosal lesions in rats Scand J Gastroenterol 2004 39 303 312 15125461 10.1080/00365520310008647
Kadowaki M Kuramoto H Takaki M Combined determination with functional and morphological studies of origin of nerve fibers expressing transient receptor potential vanilloid 1 in the myenteric plexus of the rat jejunum Auton Neurosci 2004 116 11 18 15556833 10.1016/j.autneu.2004.08.005
Schicho R Florian W Liebmann I Holzer P Lippe IT Increased expression of TRPV1 receptor in dorsal root ganglia by acid insult of the rat gastric mucosa Eur J Neurosci 2004 19 1811 1818 15078554 10.1111/j.1460-9568.2004.03290.x
Neuhuber WL Sandoz PA Fryscak T The central projections of primary afferent neurons of greater splanchnic and intercostal nerves in the rat. A horseradish peroxidase study Anat Embryol 1986 174 123 144 3706772 10.1007/BF00318344
Green T Dockray GJ Characterization of the peptidergic afferent innervation of the stomach in the rat, mouse, and guinea-pig Neuroscience 1988 25 181 193 2455875 10.1016/0306-4522(88)90017-6
Holzer P Livingston EH Saria A Guth PH Sensory neurons mediate protective vasodilatation in rat gastric mucosa Am J Physiol 1991 260 G363 G370 2003604
Holzer P Painsipp E Jocic M Heinemann A Acid challenge delays gastric pressure adaptation, blocks gastric emptying and stimulates gastric fluid secretion in the rat Neurogastroenterol Motil 2003 15 45 55 12588468 10.1046/j.1365-2982.2003.00382.x
Jocic M Schuligoi R Schöninkle E Pabst MA Holzer P Cooperation of NMDA and tachykinin NK1 and NK2 receptors in the medullary transmission of vagal afferent input from the acid-threatened rat stomach Pain 2001 89 147 157 11166470 10.1016/S0304-3959(00)00357-2
Danzer M Samberger C Schicho R Lippe IT Holzer P Immunocytochemical characterization of rat brainstem neurons with vagal afferent input from the stomach challenged by acid or ammonia Eur J Neurosci 2004 19 85 92 14750966 10.1111/j.1460-9568.2004.03109.x
Hughes P Dragunow M Induction of immediate-early genes and the control of neurotransmitter-regulated gene expression within the nervous system Pharmacol Rev 1995 47 133 178 7784478
Schulman H Hyman S Zigmond MJ, Bloom FE, Landis SC, Roberts JL, Squire LR Intracellular signaling Fundamental Neuroscience 1999 San Diego: Academic Press 269 316
Clarke GD Davison JS Mucosal receptors in the gastric antrum and small intestine of the rat with afferent fibres in the cervical vagus J Physiol 1978 284 55 67 731565
Hillsley K Grundy D Sensitivity to 5-hydroxytryptamine in different afferent subpopulations within mesenteric nerves supplying the rat jejunum J Physiol 1998 509 717 727 9596794 10.1111/j.1469-7793.1998.717bm.x
Norgren R Smith GP Central distribution of subdiaphragmatic vagal branches in the rat J Comp Neurol 1988 273 207 223 3417902 10.1002/cne.902730206
Menetrey D Gannon A Levine JD Basbaum AI Expression of c-fos protein in interneurons and projection neurons of the rat spinal cord in response to noxious somatic, articular, and visceral stimulation J Comp Neurol 1989 285 177 195 2503547 10.1002/cne.902850203
Hammond DL Presley R Gogas KR Basbaum AI Morphine or U-50,488 suppresses Fos protein-like immunoreactivity in the spinal cord and nucleus tractus solitarii evoked by a noxious visceral stimulus in the rat J Comp Neurol 1992 315 244 253 1545011 10.1002/cne.903150210
Gamboa-Esteves FO Kaye JC McWilliam PN Lima D Batten TF Immunohistochemical profiles of spinal lamina I neurones retrogradely labelled from the nucleus tractus solitarii in rat suggest excitatory projections Neuroscience 2001 104 523 538 11377852 10.1016/S0306-4522(01)00071-9
Sugiura T Dang K Lamb K Bielefeldt K Gebhart GF Acid-sensing properties in rat gastric sensory neurons from normal and ulcerated stomach J Neurosci 2005 25 2617 2627 15758172 10.1523/JNEUROSCI.2894-04.2005
Cervero F Connell LA Distribution of somatic and visceral primary afferent fibres within the thoracic spinal cord of the cat J Comp Neurol 1984 230 88 98 6096416 10.1002/cne.902300108
Holzer P Acid-sensitive ion channels in gastrointestinal function Curr Opin Pharmacol 2003 3 618 625 14644014 10.1016/j.coph.2003.06.008
Sengupta JN Gebhart GF Johnson LR Gastrointestinal afferent fibers and sensation Physiology of the Gastrointestinal Tract 1994 Third New York: Raven Press 483 519
Ozaki N Gebhart GF Characterization of mechanosensitive splanchnic nerve afferent fibers innervating the rat stomach Am J Physiol Gasrointest Liver Physiol 2001 281 G1449 G1459
Hatakeyama Y Matsuo M Tomoi M Ohtsuka M Shimomura K Luminal dilution caused by certain mild irritants and capsaicin contributes to their gastric mucosal protection Am J Physiol 1995 268 G200 G206 7864117
Cervero F McRitchie HA Neonatal capsaicin does not affect unmyelinated efferent fibers of the autonomic nervous system: functional evidence Brain Res 1982 239 283 288 7093683 10.1016/0006-8993(82)90853-8
Forster ER Green T Elliot M Bremner A Dockray GJ Gastric emptying in rats: role of afferent neurons and cholecystokinin Am J Physiol 1990 258 G552 G556 2333968
Raybould HE Hölzer H Duodenal acid-induced inhibition of gastric motility and emptying in rats Am J Physiol 1993 265 G540 G546 8214075
Lu YX Owyang C Duodenal acid-induced gastric relaxation is mediated by multiple pathways Am J Physiol 1999 276 G1501 G1506 10362654
Holzer-Petsche U Seitz H Lembeck F Effect of capsaicin on gastric corpus smooth muscle of the rat in vitro Eur J Pharmacol 1989 162 29 36 2470602 10.1016/0014-2999(89)90600-6
Lefebvre RA de Beurme FA Sas S Relaxant effect of capsaicin in the rat gastric fundus Eur J Pharmacol 1991 195 131 137 1676677 10.1016/0014-2999(91)90390-C
Takeuchi K Niida H Matsumoto J Ueshima K Okabe S Gastric motility changes in capsaicin-induced cytoprotection in the rat stomach Japn J Pharmacol 1991 55 147 155
Randich A Gebhart GF Vagal afferent modulation of nociception Brain Res Rev 1992 17 77 99 1327371 10.1016/0165-0173(92)90009-B
Qin C Chandler MJ Miller KE Foreman RD Responses and afferent pathways of C1–C2 spinal neurons to gastric distension in rats Auton Neurosci 2003 104 128 136 12648614 10.1016/S1566-0702(03)00002-X
Holzer P Livingston EH Guth PH Sensory neurons signal for an increase in rat gastric mucosal blood flow in the face of pending acid injury Gastroenterology 1991 101 416 423 2065919
Traub RJ Sengupta JN Gebhart GF Differential c-fos expression in the nucleus of the solitary tract and spinal cord following noxious gastric distention in the rat Neuroscience 1996 74 873 884 8884783 10.1016/0306-4522(96)00173-X
Molander C Grant G Paxinos G Spinal cord cytoarchitecture The Rat Nervous System 1995 Second San Diego: Academic Press 39 45
Paxinos G Watson C The Rat Brain in Stereotaxic Coordinates Third Edition on Compact Disc 1997 San Diego: Academic Press
Molander C Xu Q Grant G The cytoarchitectonic organization of the spinal cord in the rat. I. The lower thoracic and lumbosacral cord J Comp Neurol 1984 230 133 141 6512014 10.1002/cne.902300112
|
16162281
|
PMC1239919
|
CC BY
|
2021-01-04 16:03:48
|
no
|
BMC Neurosci. 2005 Sep 14; 6:60
|
utf-8
|
BMC Neurosci
| 2,005 |
10.1186/1471-2202-6-60
|
oa_comm
|
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-531614656910.1186/1475-925X-4-53ResearchClassification of the extracellular fields produced by activated neural structures Richerson Samantha [email protected] Mark [email protected] Danielle [email protected] Mark M [email protected] Department of Biomedical Engineering, Bucknell University, Lewisburg, Pa 17837 USA2 Department of Physics, Bucknell University, Lewisburg, Pa 17837 USA3 Department of Neurology, Geisinger Medical Center, 100 N Academy Rd, Danville, Pa 17822 USA2005 7 9 2005 4 53 53 17 6 2005 7 9 2005 Copyright © 2005 Richerson et al; licensee BioMed Central Ltd.2005Richerson et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Classifying the types of extracellular potentials recorded when neural structures are activated is an important component in understanding nerve pathophysiology. Varying definitions and approaches to understanding the factors that influence the potentials recorded during neural activity have made this issue complex.
Methods
In this article, many of the factors which influence the distribution of electric potential produced by a traveling action potential are discussed from a theoretical standpoint with illustrative simulations.
Results
For an axon of arbitrary shape, it is shown that a quadrupolar potential is generated by action potentials traveling along a straight axon. However, a dipole moment is generated at any point where an axon bends or its diameter changes. Next, it is shown how asymmetric disturbances in the conductivity of the medium surrounding an axon produce dipolar potentials, even during propagation along a straight axon. Next, by studying the electric fields generated by a dipole source in an insulating cylinder, it is shown that in finite volume conductors, the extracellular potentials can be very different from those in infinite volume conductors. Finally, the effects of impulses propagating along axons with inhomogeneous cable properties are analyzed.
Conclusion
Because of the well-defined factors affecting extracellular potentials, the vague terms far-field and near-field potentials should be abandoned in favor of more accurate descriptions of the potentials.
Cable PropertiesDipole MomentQuadrapole MomentAction Potential
==== Body
Background
The most commonly employed neurophysiologic techniques used to diagnose and monitor the status of the nervous system in humans involve recording extracellular fields generated by time dependent changes in transmembrane potentials in axons, dendrites or cell bodies [1,2]. In order to understand these extracellular fields, it is important to classify the various types of field and their generators. Historically, evoked potentials were classified into "far-field" potentials and "near-field" potentials according to the criteria in Table 1[3-6]. These definitions are quite different from those employed in classical electromagnetic theory [7]. In electromagnetic theory, far-field potentials refer to the dominant component of the electromagnetic field in the range where the distance between the observation point and the generator is much larger than both the wavelength of the radiation and the size of the generator. The goals of this paper are to set forward expressions for extracellular fields generated under a number of circumstances and to use this information to produce a physically based classification scheme for the different types of extracellular field that may be recorded. The first step in this process will be a detailed description of the extracellular potentials generated as an impulse travels down a generalized neural structure. This will provide an understanding of those geometric properties of a generalized axon which are associated with the production of dipolar and quadrupolar fields at large distances from the axon. Following this, the effects that an inhomogeneous extracellular electrical environment has on the recorded extracellular field will be explored in a discussion of two problems: an impulse passing near a plane conducting boundary and an impulse passing near a conducting sphere. The effects of montage selection and low frequency filtering on these extracellular recordings will be investigated. For comparison, the effects that boundary conditions have on the distribution of extracellular fields will also be explored. Finally, the extracellular fields generated by an impulse traveling down an axon whose physical properties change abruptly will be investigated within the cable model. The time course of changes seen in the extracellular fields in the presence of inhomogeneities will lead to a detailed discussion of the frequency spectrum of extracellular potentials. In each of these cases, a theoretical discussion will be presented followed by the results of illustrative simulations.
Table 1 Classical Properties of Near and Far Field Potentials
Property Far-Field Near-Field
Latency Relatively Independent of Recording Electrode Position Strongly dependent on Position of Recording Electrode
Distribution on Skin Broad Narrow
Polarity Positive Negative or Positive
Recording Monopolar Monopolar or Bipolar
It should be noted that in many parts of the paper the terms charge, dipole, or quadrupole will be used although it is more proper to refer to point current sources, dipolar or quadrupolar current sources. This does not alter any of the fundamental conclusions in this paper.
Model derivations and simulations
A. Extracellular fields and axonal geometry
Understanding how the geometry of a generalized axon affects the extracellular fields produced when it is depolarized is particularly instructive. Previous work by Plonsey and Rosenfalck was particularly useful in defining extracellular fields for active fibers of finite and infinite length in infinite homogenous media [8-10]. However, these models used previously assume fibers have circular cross sections, are straight, and are located in uniform conducting media. Holt and Koch [36] have studied this problem from the viewpoint of the cable equations. The general solution we present here is based on the geometry of surfaces, a simplified version of which has been applied previously in analyzing magnetic stimulation of a bent neuron [11,14]. Any surface, such as the surface of the generalized axon, can be characterized by a vector function of two variables [12]. In this development, these two variables will be called s and θ. It will be instructive to think of s as the distance along the length of the axon and θ as the angular position around the axon. The extracellular field generated when a small region of a nerve along its long axis is depolarized is due to relatively localized movements of charges. Thus, the extracellular fields in an infinite volume conductor can be calculated using the multipole expansion, the first terms of which are the field generated by the net dipole moment and the net quadrupole moment of the source. Appendix A (see Additional file #1) demonstrates the calculation of the net dipole and quadrupole moments produced by depolarization of the nerve in the case in which the trans membrane potential Vm(s, θ) = Vm(s) is independent of θ. This leads to the following expression for the dipole moment per unit activated length:
and the expression for the magnitude of the dipole moment per unit length:
In the above expressions, and are the unit tangent and normal vectors to the curve describing the curve of the centroids of the axon and s is arc-length along this curve. κ*(s) is the curvature [12] of at s. In addition:
where a(s, θ) is the distance from the centroid at s to the point (s, θ) on the surface. This is the mean square radius of the axon at position s.
One implication of the above equations is that the dipole moment produced by each segment of the axon depends only on the curvature of the curve of centroids and the change in the mean square radius with distance. It does not depend on the detailed shape of the axon. It is also important to realize that the component of the dipole moment produced by changing axonal radius is directed along the tangent to the centroid curve X*(s) and the component produced by the curvature of the axon are oriented along the normal to this same curve.
The quadrupole moment tensor produced when a small axonal segment is depolarized can be evaluated in a similar way although with substantially more algebra (See appendix A (see Additional file #1)) :
where τ*(s) is the torsion [12] of and the and are integrals of the cube of the radius multiplied by sin(θ) and cos(θ) as noted in appendix A (see Additional file #1). This has a number of simple implications. First, whenever the axon cross section has inversion symmetry around the centroid (a(s, θ) = a(s, θ + π)), then and are both zero and:
In order to illustrate these points, a simple simulation using the leading-trailing [13] dipole model was carried out to compute the expected extracellular fields generated as a function of time when a nerve impulse traverses a bend in an axon as shown in Figure 1. The leading and trailing dipoles are each oriented along the local tangent to the axon and are separated by a constant distance. Figure 2 shows an example of the extracellular fields demonstrating that when the recording electrodes are near the axon (small values of R) peaks are seen as the traveling impulse passes nearest the electrodes. However, far from the axon (large values of R), a peak is recorded at the point where the impulse traverses the bend. As expected, far from the source, the dipolar potential generated at the curve in the axon which declines as for large distances dominates the quadrupolar potential which declines as .
Figure 1 Illustration of the geometry used in simulating the extracellular fields generated when a nerve impulse propagates through a 180° bend with radius a = 60 mm. The separation between the leading and trailing dipoles is taken as 6 mm. The recording electrode position from which the tracings are taken is located at z = -60.7 mm. The leading and trailing dipoles [13] are oppositely oriented and parallel to the tangent of the nerve.
Figure 2 Tracings of the recorded extracellular potentials as a function of time. The tracing from the electrode closest to the axon (R = 10 mm) shows a typical quadrupolar traveling wave as the impulse passes closest to the recording electrode. As the recording electrode is placed progressively further from the axon (R = 245 mm and R = 1019 mm) the quadrupolar potential from the action potential passing near the electrode is dwarfed by the dipolar potential generated as the impulse travels through the bend. The impulse begins at t = 0 at z = -100 mm and by t = 10 reaches z = -100 again. The impulse reaches the location of the recording electrode at t = 1.1. The velocity of the travelling action potential was taken as 39 m/sec.
B. Changes in local extracellular electrical environment
Once the effects of axon geometry on the recorded extracellular potentials have been explored, it is natural to explore the effects that changes in the extracellular electrical environment may have.
It has been suggested that the presence of inhomogeneities in the extracellular space can change the fundamental characteristics of the extracellular fields, producing "far-field" potentials (A discussion of the inverse problem of stimulating a nerve near a region with a localized change in conductivity can be found in Roth [14] and a discussion of the general effects of inhomogeneities in conductivity can be found in Geselowitz [15].). In order to understand these effects, it is helpful to begin by studying the fields generated by a small conducting sphere of radius a placed at position in an arbitrary electric potential . An approximate expression for the potential produced by the presence of the sphere far from the sphere is (see Appendix B (see Additional file #1)):
Therefore, as long as the electric field that would have been present at the center of the sphere in its absence is non-zero, a dipolar potential can be generated by this conducting sphere. This is true even if that field is due to a quadrupole or higher order source so that the presence of inhomogeneities can produce extracellular potentials that decay much more slowly than those of the driving field. More generally, it is known [16] that the dipole moment induced in any dielectric (or conducting) object by an external electric field is:
where v is the volume of the object and is a constant tensor that depends on the geometry of the object and the difference between the dielectric constant (or conductivity) of the object and the medium. For a sphere with dielectric constant (ε) (or conductivity (σ)) differing from that of the remainder of the environment:
where, ε0 is the dielectric constant of the rest of the medium and δij is the Kronecker delta. A similar expression holds for the case of a conducting sphere with ε replaced by σ and ε0 replaced by σ0, the conductivity of the medium outside the sphere.
In contrast to the results obtained above, it is important to note that dipolar fields are NOT generated when a nerve impulse characterized by a quadrupolar source travels near a plane conducting boundary. The effect of the plane boundary on the extracellular fields can be simulated by image charges placed at the location that their optical images would have in the boundary plane. Thus, if the source charge density is quadrupolar, the image charges are quadrupolar.
It is instructive to compute the actual extracellular fields generated both when an impulse passes near a conducting sphere as in Figure 3 and near a conducting plane boundary as in Figure 4. These simulations were based on the leading/trailing dipole model with the leading and trailing dipoles oppositely oriented parallel and antiparallel to the z axis separated by a distance of 6 mm. The method of images [7] was used to compute the fields that satisfy the appropriate boundary conditions outside the sphere after breaking each dipole into closely spaced charges of opposite charge. Figure 5 shows the potential as a function of nerve impulse location in the case of the conducting sphere for electrodes near and far from the axon. It is clear that when the recording electrodes are near the axon (R small) the peak in the potential is seen as the impulse traverses the point nearest the recording electrodes while, when the electrode is further from the axon (large R), the peak in the extracellular potential appears as the impulse passes nearest to the conducting sphere. This is what would be expected on the basis of the dipolar potential generated as the impulse approaches the sphere.
Figure 3 Illustration of the model used for the leading/trailing dipole simulation of the extracellular fields near a conducting sphere. R is the distance of the recording electrodes from the center of the axon. The distance between the leading and trailing dipoles is taken as 6 mm. The nerve is stimulated on the left side (negative z values) and propagates toward the right side. The sphere is taken as highly conducting with a radius of 1 mm and is placed 1.1 mm below the nerve. The center of the sphere is taken at z = 0.
Figure 4 Illustration of the model used for the leading/trailing dipole simulation of the extracellular fields near a plane boundary. R is the distance of the recording electrodes from the center of the axon. The distance between the leading and trailing dipoles is taken as 6 mm. The nerve is stimulated on the left side (negative z values) and propagates toward the left side. The region of high conductivity is on the right z > 0 and the region of lower extracellular conductivity is on the left z < 0.
Figure 5 Plots of the extracellular potential recorded at an electrode a distance R from the axon located at z = -59 mm as the impulse travels from z = -100 mm to z = 100 mm. The geometry is that of the spherical conductor placed as in Figure 4. When the electrodes are located close to the axon (R = 10 mm), the peak in the potential occurs as the nerve impulse passes near the electrodes. For electrodes far from the axon (R = 110 mm or R = 1014 mm) the peak extracellular potential is recorded as the impulse travels nearest the sphere (z = 0). The values of the potential are arbitrary and are for purposes of comparison of the different graphs within this figure only.
Figure 6 shows the extracellular fields produced as an impulse passes through a plane conducting boundary. It is clear that the presence of the plane conducting boundary is not associated with a peak in the extracellular field as the impulse traverses the boundary as in the case of the impulse near the conducting sphere. Again, this has its origin in the fact that the potential drops as at large distances from the axon in the case of the conducting sphere and for the plane conducting boundary.
Figure 6 Plots of the extracellular potential recorded at an electrode (located at z = -59 mm) a distance R from the axon as the impulse travels from z = -100 mm to z = 100 mm. The geometry is that of the plane conducting boundary placed as in Figure 5. When the electrodes are located close to the axon (R = 10 mm), the peak in the potential occurs as the nerve impulse passes near the electrodes. For electrodes far from the axon (R = 1014 mm) a step change in potential is seen as the impulse passes through the boundary between the two regions. The values of the potential are arbitrary and are for purposes of comparison of the different graphs within this figure only.
C. Effects of finite volume conductors
Since practical recordings are not carried out in an infinite volume conductor, it is useful to consider a simple qualitative model demonstrating the effects that placing a charge in a finite volume conductor can have on the recorded extracellular fields (Appendix C (see Additional file #1)). In particular, the special case of charges placed inside a cylindrical volume conductor that is insulating except at its two ends is very instructive. Assuming that all charges are confined to a finite segment of the cylinder z0 - Δ <z <z0 + Δ, the potential outside the region in which the charges exist can be written:
It is instructive to consider the mean value of the potential along segments perpendicular to the long axis of the cylinder:
These integrals are not well defined in an infinite volume and so the results are restricted to finite cylindrical volume conductors. Integrating the Poisson equation over the radial and angular variables yields an expression for the mean potentials in the charge free regions and integrating the Poisson equation across the region containing charges produces the following expression for the averaged electrical potential:
In this equation, Q is the total charge contained in the region and dz is the component of the dipole moment of the charge along the long axis of the cylinder. a<and b<are constants chosen to satisfy the boundary conditions. This suggests that, far enough from the sources, the actual potential has a linear dependence on the axial coordinate. The form of the above equation suggests that if the source is dipolar there may be a step change in the potential across the region containing the charges that is proportional to the dipole moment. A quadrupolar field would be expected to be least influenced by the presence of the finite cylindrical volume conductor. These results are similar to those obtainable using the Green's function for the appropriate cylindrical boundary conditions [17].
The critical question is to define the regime in which these results apply. Some arguments relating to this question are presented in Appendix C (see Additional file #1). They suggest, for an infinitely long cylinder, that when the axial coordinate of the point of observation is many cylinder diameters from the charges (z >> a), the potentials develop the linear behavior discussed above. Similarly, when the point of observation is much less than a cylinder diameter away from the charge, it is expected that the potential will possess similarities to that seen in an infinite volume conductor.
In order to understand the effects of finite volume conductors in more detail, a series of finite element simulations were performed using FEMLAB (Comsol, Natick MA). In each of these simulations the Laplace equation was solved using an algebraic preconditioner followed by a conjugate gradients equation solver. Results were checked for stability to changes in grid sizes. In order to allow for reasonable solutions with manageable grid sizes, finite sized spheres were used in place of point charges. Figure 7 illustrates the distribution of potential from a conducting sphere of size 0.1 held at a potential of unity placed in the center of a cylinder of length 10 and radii ranging from 1 to 100. The ends of the cylinder were grounded and the remainder of the cylinder was considered to be insulating. This clearly demonstrates that at roughly half a cylinder radius from the sphere, the potentials deviate from the values that they would have in an infinite volume conductor and demonstrate linear changes with z. Figure 8 demonstrates the potential produced by a dipole simulated by two conducting spheres of diameter 0.25 at +/-0.5 are held at potentials of +/-1. Smaller spheres and smaller spacing between spheres required prohibitively large grids for solution. It is clear that the transition to the linear behavior discussed above occurs only for the thin cylinder. Finally, Figure 9 shows the potential produced by a quadrupole simulated by three spheres of diameter 0.25 centered at 0.75, 0, -0.75 with charges 1,-2,1. It is clear that the finite volume conductor has much less effect on the potential generated by the quadrupole than those generated by the dipole or the monopole as expected.
Figure 7 The potentials generated by a conducting sphere of radius 0.1 mm held at constant potential 1 placed in cylindrical volume conductors of differing sizes. The flat ends of the cylinder are at ground potential and the curved surface is taken as an insulator. Note that the cylinder has length 10 mm and extends from -5 to 5 on either side of the point z = 0.
Figure 8 The potentials generated by a "dipole" formed from two conducting spheres of radius 0.25 mm centered around z = +/-0.5 mm held at constant potential 1 and -1 respectively placed in cylindrical volume conductors of differing sizes. The flat ends of the cylinder are at ground potential and the curved surface is taken as an insulator.
Figure 9 The potentials generated by a "quadrupole" formed from three conducting spheres of radius 0.25 mm centered around z = +/-0.75 mm and z = 0 held at constant potential 1,1,-2 respectively placed in cylindrical volume conductors of differing sizes. The flat ends of the cylinder are at ground potential and the curved surface is taken as an insulator.
Figure 10 further illustrates the properties of the volume conductor that affect the extracellular fields. It shows the results of finite element simulations for a case in which a thin (radius = 1, length = 10) insulating cylinder is attached to a larger (radius = 5) insulating sphere as would be the case if an arm were attached to a larger torso. Specifically this figure demonstrates the potential recorded from the z = -10 end of the cylinder and the z = +10 end of the sphere for a dipole source placed at different locations along its long axis. As expected, the potentials at the ends of the cylindrical region exhibits a step change as the dipole passes from the cylindrical end of the figure into the spherical end. The behavior of the potential in the spherical region is more complex.
Figure 10 The effects of more complex geometries of the volume conductor on the electric potentials recorded. All results refer to the following geometry. A fully insulated cylinder of radius 1 and length 10 extending from z = -10 to z = 0. This cylinder is attached to an insulated sphere of radius 5 with the two intersecting at z = 0. The plots show the potential at the flat cylindrical end (z = -10) and the end of the z = 10 spherical edge as a function of the axial location of the center of a dipole consisting of two spheres held at potentials of 1 and -1 as in Figure 8. The step in the voltage at the cylindrical end of the volume as the dipole moves through the boundary between the two regions is evident.
The sensitivity of the recorded extracellular potentials to the shape, size and character of the volume conductor in which the sources are immersed is evident in these calculations. They are also very sensitive to the detailed structure of the boundary conditions imposed. These observations are consistent with the results obtained in a number of experiments performed by Jewett [6,18] who recorded changes in extracellular potentials from isolated nerves as the impulse propagated into volume conductors of different sizes and shapes. These observations on the importance of finite volume effects are also consistent with the conclusions drawn from finite element models of Cunningham [19] who evaluated extracellular fields created in a 2-D finite element model of the hand, arm and torso. However, as the geometry of the system becomes more complex, the interpretation of results also becomes more complex. Mapping studies of the electric fields generated by artificial dipole and quadrupole sources by Dumitru and King [20,21] reproduce the above theoretical and computational results and in particular demonstrate that dipole sources in cylinders are associated with large regions of constant potential away from the source. However, these authors state that quadrupolar sources do produce fields far from the source when inside a cylinder under certain circumstances. This may result from the fact that in these simulations discrete electrodes were used to generate the quadrupolar source.
Clinical studies [22-24] on the distribution of the potentials generated by median nerve stimulation in humans do demonstrate some potentials with a latency that is independent of electrode position that are broadly distributed over the skin. The spatial distribution of these potentials does not appear to be quadrupolar or bipolar in nature as they seem to decrease in amplitude very slowly with distance from the putative source. It is thus likely that boundary effects play a significant role in the generation of these potentials.
D. Changes in cable properties of an axon
The effects that a localized alteration in the cable properties of an axon has on the generated extracellular fields are also of interest [25]. As long as the axon remains cable-like, straight and with a constant radius, equation (2) shows that only quadrupolar potentials will be generated even if cable properties of the axon change along its length. However, as in the case of the impulse approaching the plane boundary, the effective quadrupole moment will change near the point at which the cable properties change.
In order to illustrate these effects, a simple simulation was undertaken to find the transmembrane potential as a function of time for the situation in which a stimulator moving with a constant velocity injects a constant amount of current through the membrane. It is assumed that the membrane resistance quadruples at the point z = 100. The equations describing this simple model and the method of computing the extracellular fields is discussed in Appendix D (see Additional file #1).
Figure 11 contains a plot of the recorded potential at z = 25 as the impulse passes by this location as a function of the distance of the recording electrodes from the axon. This clearly demonstrates that, when the recording electrodes are close to the axon, only the traveling wave is recorded as the impulse passes near the axon. Far from the axon, much of what is seen in the figure is apparently generated as the impulse reaches the interface between the two regions of altered membrane resistance. In fact, much of this is a function of the time scale used to plot the data. As will become apparent in the next section, the frequency of propagating extracellular potentials recorded declines as the point of observation moves further from the axon. Thus, far enough from the axon, the propagating potential will be difficult to visualize although the potentials that occur as the impulse encounters a boundary having higher frequency components are easier to see. A detailed analysis of the potential seen at the time the impulse reaches the boundary supports this interpretation since it does fall as at large distances from the axon and is quadrupolar.
Figure 11 Plot of the potential generated at various distances from the stimulated non-uniform axon described equations (D.1) and (D.2) as a function of time for electrodes with varying distances from the axon. The recording electrode is located at the point z = 25 mm. In this graph, the abscissa is the location of the stimulator as it moves along the axon. The downward arrow denotes the location (z = 100 mm) where the space constant of the axon suddenly doubles and the upward arrow denotes the location (z = 25 mm) of the recording electrode. Note that for recording electrodes near the axon only the traveling wave is noted but when the electrodes are placed far from the axon, only potentials generated as the impulse reaches the boundary are evident.
F. Spectral analysis of extracellular potentials
As in the above section, it is clear that the frequency of the extracellular potentials produced as an action potential moves through a complex medium is an important factor in interpreting recordings. Thus, an understanding of the spectral content of the potentials recorded from a travelling action potential is important. This problem is discussed in detail in Appendix E (see Additional file #1) where it is shown that the power spectrum at angular frequency ω of the potential recorded a distance R from an axon along which an action potential propagates at a constant velocity v is given by:
where α is the quadrupole moment associated with the action potential, β is the spatial extent of the action potential and:
It is important to note that (12) implies that, the spectral response is the product of two factors, a "form factor" related to the spectral content of the impulse itself and a "structure factor" related to the distribution of the potential from a quadrupolar source. In Appendix E (see Additional file #1) it is shown that 90% of the contribution of the "structure factor" to the total power in the recorded extracellular potential occurs over the frequency range:
Table 2 contains a list of the frequencies that correspond to these values for different recording electrode to axon distances, R, for an impulse travelling at a velocity of 40 m/sec. The critical observation is that higher frequencies are recorded near the axon and only low frequencies far from the axon. On the other hand, contributions from the "form factor" are independent of R but do depend on the spatial extent of the action potential. Ninety percent of the power in the "form factor" occurs for:
Table 2 Cutoff/Peak Frequencies For Action Potentials
R fmax (Hz) fpeak (Hz)
0.1 cm 74,000 1,262 1,273
1 cm 7,400 1,262 716
10 cm 740 1,262 95.4
100 cm 74 1,262 9.54
Taking β = 6 mm and v = 40 meters/second, it is possible to tabulate the cutoff frequencies for the structure and form factors as well as the location of the peak of the power spectrum as in Table 2. It is clear, as expected, that close to the axon, the highest frequency is determined by the shape of the action potential but, at large distances, the frequency content of the recorded action potential is more strongly determined by the distance from the axon. This is confirmed in Figure 12 which demonstrates the power spectra expected from (12) as a function of distance from the axon.
Figure 12 Plot of the power spectrum of extracellular fields generated by an action potential propagating along an axon as a function of distance, R, from the axon. It is assumed that the potential moves with a velocity of 40 meters/second and that the length of activated neuron is 6 mm. The maximum frequency for which there is significant power clearly declines as the distance between the axon and the recording electrode increases.
This argument demonstrates that applying a low frequency filter when recording electrodes are placed far from the source can remove the responses generated as the action potential traverses the axon. However, the frequency spectrum of the potentials generated when an impulse reaches any boundary must be similar to that of the "form factor" only since these potentials do not propagate down the axon (The "structure factor" component arises only out of propagation of the impulse as is evident since only this term contains the propagation velocity.). Thus, as in Table 2, if the distance between the recording electrodes and the axon is sufficiently great, a low frequency filter may eliminate the propagating potentials to a much greater degree than the potentials generated at interfaces. This may obscure the true nature of the generators of the recorded extracellular fields.
G. The effect of recording montage
It should be noted that all of the simulated extracellular potentials discussed above represent the absolute value of the potential at the recording point. This corresponds to the use of a referential recording montage. In many clinical situations, bipolar recordings of nerve action potentials are made in which the difference in potential between two closely placed electrodes is recorded. The question is whether there is any theoretical disadvantage to either of these recording techniques for certain types of extracellular fields. Figure 13 shows the logarithm of the absolute value of the potential produced by a quadrupole in an infinite volume conductor and in a cylindrical volume conductor computed as in section C above. The thin lines refer to reference recordings and the thicker lines refer to bipolar recordings. As suggested by Stegeman, both the bipolar and reference montages record potentials that vary with the location of the electrode when recordings are made in an infinite volume conductor while the bipolar recording in the cylindrical volume conductor produces only constant values when the electrode is more than 2 cylinder radii from the quadrupole. Thus, if recordings of a propagating action potential are important in a finite volume conductor, then reference recordings would be required. If only the changes that appear as a potential reaches some inhomogeneity then a bipolar montage would be adequate as seen by comparing the results of Figure 14 to those of Figures 5 and 6.
Figure 13 Plot of the potentials recorded from the quadrupole of Figure 9 in an infinite volume conductor and an insulating cylindrical volume conductor of radius 1 extending from z = -5 to z = 5. Recordings made with a monopolar electrode (reference recording) and a bipolar recording are shown. The logarithm of the absolute value of the potential is plotted rather than the potential in order to better show the behavior of the potential far from the source. Note that plots extend only from z = 1 to z = 5 although the cylinder extends from -5 to 5. This demonstrates that bipolar recordings far from the source yield a constant value independent of electrode location.
Figure 14 Plots of the potentials recorded in the same simulations as in Figures 5 and 6 using a bipolar montage with the two electrodes placed a short distance from one another in the z direction. This demonstrates essentially similar behavior to that seen with the monopolar recording.
H. Radiation fields
Although electromagnetic radiation is not typically considered in clinical applications, changes in charge distributions over time produce radiation fields that typically decline in amplitude as for large distances from the source and so could be significant when the recording electrodes are very far from the source. However, the detailed considerations of appendix F (see Additional file #1) indicate that these fields are miniscule in all potential situations.
Discussion
In this paper, a number of theoretical mechanisms underlying the generation and recording of extracellular potentials during propagation of a nerve impulse down a generalized neural structure were presented and illustrated with data from simulations. This approach provides a much greater degree of insight into the generation of extracellular fields than the use of simple analytic approaches, finite element simulations, or clinical recordings in isolation. In particular, although necessarily approximate and idealized, the theoretical calculations allow for predictions of the general categories of field types that can be encountered and estimates of their magnitude.
When a neural membrane is depolarized the change in transmembrane potential is associated with a dipole moment density oriented perpendicular to the surface of the membrane that is proportional to the change in potential. Because of the symmetry of the ideal cylindrical axon, the net dipole moment associated with the activation disappears and so the propagating nerve impulse is typically a quadrupolar field far from the axon. However, when the cylindrical symmetry is broken by bending the nerve, a net dipole moment along the normal to the axon may be generated. In addition, if there are changes in the axon radius along the length of the cylinder there will be dipole moments generated directed along the tangent to the axon.
The theoretical results and simulations discussed above are in agreement with the experimental results obtained by Dupree and Jewett [26] which showed a peak in the extracellular potential that is generated when frog sciatic nerve was bent. These recordings were however made with electrodes always within 200 mm of an axon placed in a finite volume conductor. Because of the relative proximity of the electrodes to the source the full transition from recording only traveling waves to recording only potentials as the impulse enters the bend is not seen in this experiment.
In addition, the above theoretical results are applicable to situations in which an axon terminates in a sealed end. This is equivalent to allowing the axon radius to change from its baseline value to zero at the end of the axon. As described by (2) a net dipole moment will appear at such points and a peak will be recorded in the extracellular potentials. Equation (2) does not suggest that a net dipole moment will appear in a cylindrical axon with constant radius that is cleanly cut (ie not sealed) at the end. Dumitru and Jewett [13] suggest that this should occur because, once the impulse reaches the end of the axon, there is no longer "neural tissue to support the leading dipole". They speculate that at this point only the fields generated by the trailing dipole appear. This is not the case, since the leading and trailing dipoles are not physical entities and are only used to represent the field generated by the depolarized membrane surface. In fact, as the impulse reaches the cut end, the equivalent leading dipole must remain at the end of the axon while the trailing dipole gradually reaches the end of the axon. Thus, the quadrupole moment of the impulse will decline linearly to zero as the impulse reaches the cut end of the nerve since the total area of depolarization will diminish gradually. Thus, no peak in extracellular potential should be recorded as the impulse reaches a pure "cut end". However, if there is any change in axon diameter at the end, a large potential may be generated.
As breaking the cylindrical symmetry of an axon by distorting the axon itself can produce net dipolar fields, so can placing an axon in an environment that is not cylindrically symmetric. In particular, placing a sphere of altered conductivity near the axon destroys cylindrical symmetry and results in a dipolar potential even when extracellular fields generated by the axon itself would be quadrupolar. Placing an infinite conducting plane perpendicular to the path of the traveling impulse, however, does not destroy the symmetry of the depolarized axonal segments and so, although the effective quadrupole moment of the propagating action potential changes as the impulse passes through this barrier, no dipolar potentials are generated. Similarly, when a propagating impulse reaches a point on the axon at which there is a sudden change in the cable properties, there is a change in the quadrupole moment but no dipole potential is generated.
The results obtained in this section should be compared to the experiments of Nakanishi [27] who placed nerves through multiple partitions and found that, with one electrode in each partitioned segment, potentials were generated that could be related to the passage of the impulse from one compartment into another. The amplitude of this potential was correlated with the impedance between the partitions, a measure of the size of the gap in the partition. This also suggested that the geometry of the partitions was critical to the development of the responses. This seems plausible and recordings with electrodes either far from the axon or in multiple locations to detect the angular dependence of the potential could be helpful in determining whether the true character of the responses are dipolar or quadrupolar.
The above simulations of the potentials produced as an impulse passes through a plane conducting barrier are different from those found by Stegeman [28,29] who computed the potentials produced by a simulated impulse traveling in an insulating cylindrical shell filled with media of different conductivity. In this model, an action potential travels at constant velocity along the long axis (z axis) of the cylinder at its center and encounters a stepwise change in conductivity at the point z = 0. Their simulations suggested that, as the nerve impulse passes the point z = 0, it generates a stepwise constant shift in the electric potential at all points z > 0 although no or minimal change is seen for in the region of space z < 0. Although there are sudden changes in the potential in model described above as the impulse passes through the boundary between the regions of differing conductivity, all of the potentials do decline with distance from the source as would be expected from a quadrupolar source. This difference between Stegeman's model and the model of an action potential approaching a plane conducting barrier as discussed above relates to the unusual types of electric field that are generated in a finite cylindrical volume conductor. As discussed in section C, when sources are placed in an insulating cylinder, the recorded potential is similar in many ways to its value in an infinite volume conductor when the recording electrodes are very close to the source. However, neither dipolar or quadrupolar fields are recorded more than 2 or 3 cylinder radii from the source but the potential becomes a linear function of the axial coordinate. This behavior is specific to insulating cylindrical volume conductors. Although there are distortions of the potentials in a circular volume conductor and in a cylindrical volume conductor that is not perfectly insulating, many of the characteristics of the potentials recorded in an infinite volume conductor persist. Field distributions produced primarily as a result of boundary effects can be identified in actual recordings when the potential does not drop as or or when the expected angular behavior expected with a dipolar source, cos(θ), or a quadrupolar source, 3 cos2(θ)-1, are not evident.
The effect of recording montage on the recorded potentials was discussed. In any recordings of extracellular potentials in an infinite volume conductor, similar information is recorded from closely spaced electrodes and with a distant reference because the potentials drop off with distance in a predictable manner. However, the potentials generated by sources within a finite volume conductor with insulating boundary conditions become linear functions of distance far from the source and so bipolar recordings will not record changes in the location of the sources over time while monopolar or reference recordings will.
The spectral properties of the generated extracellular fields were also elucidated with the result that the frequency spectrum generated by a traveling impulse is cutoff at a frequency which is inversely proportional to the distance of the recording electrodes from the axon. Put differently, the frequency of potentials recorded far from a traveling quadrupolar action potential drops as the distance between the recording electrodes and the axon increases. On the other hand, the frequency spectrum generated when an impulse reaches a boundary is independent of the distance from the generator. In this case the frequency is cutoff at roughly the size of the nerve impulse divided by the conduction velocity. Understanding this frequency behavior is important in that it forms the scientific basis for choosing filters to optimally record potentials with different origins. For instance, if recording of traveling potentials far from the axon, is desired, it will be important to keep low frequency (high pass) filters as low as possible. However, if the goal of an experiment is to analyze the changes in the local environment that a nerve impulse encounters, it is best to set the low frequency filter high enough to reduce the amplitude of the traveling potentials [30].
Finally, in a discussion of the possibility that radiation fields are generated during the propagation of neural impulses or when an impulse encounters a localized change in conductivity, it was demonstrated that such fields are extremely tiny. The main reasons for this are the fact that the rate of change in charge distributions is slow and the radiated power increases as the fourth power of the frequency. The radiated power also depends on the fourth power of the size of the generator and neural structures are typically small.
Many of the previous studies into the extracellular fields generated by action potentials were heavily reliant on actual recordings and on finite element simulations of the extracellular fields. This resulted in the empirical classification shown in Table 1. Both of these techniques are valuable but they do not give a full sense of the types of field generated that analytic computations can. The arguments presented in this paper strongly suggest that a better classification for extracellular fields should be that described in Table 3. This classification is based primarily on whether the recorded potentials are dipolar, quadrupolar or related to finite volume effects and whether they are generated by a traveling wave of excitation or generated when the traveling wave reaches a point where there is either a change in the axon geometry, axon cable properties or the local extracellular environment. This approach avoids the ambiguity associated with the terms near-field and far-field potential which have in the last 30 years been used in many different contexts by different authors.
Table 3 Summary of Extracellular Field Types Far From An Axon
Field Type Causes Far-Field Behavior Angular Generator Type Comment
Radiation Change in Dipole Direction Moving Dipole Through Region of Altered Electrical Properties cos(θ) -Free Particle. Complex Angular Dependence for Transition Radiation Traveling Clinically Insignificant
Dipole Change in Axon Radius (Dipole Tangent To Axon) cos(θ) Stationary Includes "Sealed End" Axons
Axon curvature (Dipole Normal To Axon) cos(θ) Stationary
Localized Changes in Extracellular Electrical Properties Cos(θ) Stationary Only When Cylindrical Symmetry is Broken
Quadrupolar Impulse Travelling Along Uniform, Straight, Homogenous Axon in Isotropic Electrical Environment. 3 cos2(θ)-1 Travelling
Changes in Cable Properties. 3 cos2(θ)-1 Stationary
Localized Changes in Extracellular Electrical Properties 3 cos2(θ)-1 Stationary For Instance the Plane Conducting Boundary
Boundary Fields Generated Because of the Finite Volume In Which The Generator is Immersed Broadly Distributed Depends Critically on Boundary Shape and Size Stationary or Travelling Potentials Linear Far From Source For A Cylindrical Volume Conductor
Authors' contributions
SR participated in coordination of the study, designed and oversaw the modeling, and helped to draft the manuscript. DP and MI carried out the simulations. MS conceived of the study, participated in its design and coordination, aided in the derivations, and helped to draft the manuscript. All authors read and approved the final manuscript.
Figure 15 Plot of the function g(ω)in equation (E.14).
Supplementary Material
Additional File 1
Appendices; with regards to p24, please see Figure 15
Click here for file
Acknowledgements
Danielle Perry and Mark Ingram were sponsored by the NSF REU grant to the department of Physics at Bucknell University. The authors also wish to thank Robert Marchitelli who also contributed to the calculations in this paper under the sponsorship of the NSF REU grant to the department of physics at Bucknell. Finally, we wish to thank Dr. Marty Ligare from the department of Physics at Bucknell University for his help in organizing this project and his ongoing assistance in all aspects of this project. Finally, we wish to thank one of the anonymous reviewers for comments that have significantly improved this manuscript.
==== Refs
Chiappa KH Evoked Potentials in General Medicine 1997 Philadelphia; Lippincott-Raven
Plonsey R Barr RC Bioelectricity: a quantitative approach 2000 New York: Kluwer Academic
Yamada T The anatomic and physiologic bases for median nerve somatosensory evoked potentials Neurologic Clinics North America 1988 6 704 733
Cracco RQ Cracco JB Somatosensory evoked potential in man: far field potentials Electroenceph Clin Neurophys 1976 41 460 466
Frith RW Benstead TJ Daube JR Stationary waves recorded at the shoulder after median nerve stimulation Neurology 1986 36 1458 1464 3762965
Jewett DL Deupree DL Bommannan D Far field potential generated by action potentials of isolated frog sciatic nerves in a spherical volume Electroenceph Clin Neurophys 1990 75 105 117
Jackson JD Classical Electrodynamics 1999 3 New York: Wiley
Plonsey R The active fiber in a volume conductor IEEE Trans Biomed Eng 1974 21 371 381 4461667
Plonsey R Action potential sources and their volume conductor fields Proc IEEE 1975 65 601 611
Rosenfalck P Intra- and extracellular potential fields of active nerve fibers Thromb Diath Haemorrh Suppl 1969 321 1 168 5383732
Abdeen MA Stuchly MA Modeling of magnetic field stimulation of bent neurons IEEE Trans Biomed Eng 1994 41 1092 1095 8001998
Struik DJ Lectures on Classical Differential Geometry 1961 Reading, Mass: Addison Wesley
Dumitru D Jewett DL Far-Field Potentials Muscle and Nerve 1993 16 237 254 8446122
Roth BJ Mechanisms for electrical stimulation of excitable tissue Crit Rev Biomed Eng 1994 22 253 305 8598130
Geselowitz DB On the theory of the electrocardiogram Proc IEEE 1989 77 857 876
Landau LD Lifshitz EM Electrodynamics of Continuous Media 1960 New York: Pergammon Press
Ili S Jeffrey DJ A note on Laplace's equation inside a cylinder 14296212
Jewett DL Deupree DL Far-field potentials recorded from action potentials and from a tripole in a hemicylindrical volume Electroenceph Clin Neurophys 1989 72 439 449
Cunningham K Halliday AM Jones SJ Simulation of 'stationary' SAP and SEP phenomena by 2-dimensional potential field modeling Electroencephalogr Clin Neurophysiol 1986 65 416 28 2429821
Dumitru D King JC Far-field potentials in circular volumes: evidence to support the leading/trailing dipole model Muscle and Nerve 1992 15 101 105 1732753
Dumitru D King JC Far-field potential production by quadrupole generators in cylindrical volume conductors Electroenceph Clin Neurophys 1993 88 421 431
Kimura J Ishida T Suzuki S Kudo Y Matsuoka H Yamada T Far-field recording of the junctional potential generated by median nerve volleys at the wrist Neurology 1986 36 1451 7 3762964
Kimura J Yamada T Physiologic mechanisms underlying the generation of far-field potentials Electroencephalogr Clin Neurophysiol Suppl 1990 41 13 21 2289420
Yamada T Machida M Oishi M Kimura A Kimura J Stationary negative potentials near the source vs. positive far-field potentials at a distance Electroenceph Clin Neurophys 1985 60 509 524
Tasaki I Field J Conduction of the nerve impulse Handbook of Physiology: Neurophysiology I 1959 chapter 3 Washington DC: American Physiologic Society
Deupree DL Jewett DL Far-field potentials due to action potentials traversing curved nerves, reaching cut ends and crossing boundaries between cylindrical volumes Electroenceph Clin Neurophys 1988 70 355 362
Nakanishi T Origin of action potential recorded by fluid electrodes Electroenceph Clin Neurophys 1983 55 114 115
Stegeman DF Oosterom AV Colon EJ Far-field evoked potential components induced by a propagating generator: computational evidence Electroenceph Clin Neurophys 1987 67 176 187
Stegeman DF Dumitru D King JC Roeleveld K Near and Far-Fields: Source Characteristics and the Conducting Medium in Neurophysiology J Clin Neurophys 1997 14 429 442
Gobbele R Waberski TD Dieckhofer A Kawohl W Klostermann F Curio G Buchner H Patterns of disturbed impulse propagation in multiple sclerosis identified by low and high frequency somatosensory evoked potential components J Clin Neurophys 2003 20 283 290
Morse PM Feshbach H Methods of Theoretical Physics 1953 New York: McGraw-Hill
Frank IM Transition radiation and optical properties of matter Sov-Phys-Usp 1966 8 729 742
Lam J Transition radiation from an anisotropic dielectric layer with particular reference to the biological membrane J Phys A 1969 2 529 539
Landau LD Lifshitz EM The classical theory of fields 1975 New York: Pergammon
Amir A Jewett DL Far field action potentials generated by inhomogeneities in a volume conductor Engineering in Medicine and Biology Society 1992 2458 2460
Holt GR Koch C Electrical Interactions via the Extracellular Potential Near Cell Bodies Journal of Computational Neuroscience 1999 6 169 184 10333161
|
16146569
|
PMC1239920
|
CC BY
|
2021-01-04 16:37:34
|
no
|
Biomed Eng Online. 2005 Sep 7; 4:53
|
utf-8
|
Biomed Eng Online
| 2,005 |
10.1186/1475-925X-4-53
|
oa_comm
|
==== Front
Cancer Cell IntCancer Cell International1475-2867BioMed Central London 1475-2867-5-291615330310.1186/1475-2867-5-29Primary ResearchEpigenetic inactivation and aberrant transcription of CSMD1 in squamous cell carcinoma cell lines Richter Toni M [email protected] Benton D [email protected] Steven B [email protected] Dept of Otolaryngology - Head and Neck Surgery, Washington University School of Medicine, Box 8115, 660 S. Euclid Ave., St. Louis, MO 63110, USA2005 9 9 2005 5 29 29 16 6 2005 9 9 2005 Copyright © 2005 Richter et al; licensee BioMed Central Ltd.2005Richter et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The p23.2 region of human chromosome 8 is frequently deleted in several types of epithelial cancer and those deletions appear to be associated with poor prognosis. Cub and Sushi Multiple Domains 1 (CSMD1) was positionally cloned as a candidate for the 8p23 suppressor but point mutations in this gene are rare relative to the frequency of allelic loss. In an effort to identify alternative mechanisms of inactivation, we have characterized CSMD1 expression and epigenetic modifications in head and neck squamous cell carcinoma cell lines.
Results
Only one of the 20 cell lines examined appears to express a structurally normal CSMD1 transcript. The rest express transcripts which either lack internal exons, terminate abnormally or initiate at cryptic promoters. None of these truncated transcripts is predicted to encode a functional CSMD1 protein. Cell lines that express little or no CSMD1 RNA exhibit DNA methylation of a specific region of the CpG island surrounding CSMD1's first exon.
Conclusion
Correlating methylation patterns and expression suggests that it is modification of the genomic DNA preceding the first exon that is associated with gene silencing and that methylation of CpG dinucleotides further 3' does not contribute to inactivation of the gene. Taken together, the cell line data suggest that epigenetic silencing and aberrant splicing rather than point mutations may be contributing to the reduction in CSMD1 expression in squamous cancers. These mechanisms can now serve as a focus for further analysis of primary squamous cancers.
==== Body
Background
CUB and Sushi Multiple Domains 1 (CSMD1) was cloned as a candidate tumor suppressor or progression gene from a region of human chromosome 8 deleted in tumors of the upper aerodigestive tract, prostate, ovary and bladder [1-7]. Deletion of 8p23.2 or reduced expression of CSMD1 has been associated with poor prognosis in head and neck squamous cell carcinomas and in prostate cancers [2,5,8].
CSMD1, consisting of 70 exons spread over two megabases of 8p23.2, encodes a rare 11.5 kb transcript most abundantly expressed in the brain [1]. It is the founding member of a novel, evolutionarily highly conserved gene family whose proteins contain multiple domains thought to be sites of protein-protein or protein-ligand interactions and whose structure suggests that they may be transmembrane receptors or adhesion proteins [9,10].
Tumor suppressor genes are expected to be inactivated in cancers either genetically by mutations or epigenetically by modification of their promoters. While CSMD1 transcripts are detectable in upper aerodigestive tract epithelium, preliminary analysis of several head and neck squamous cell carcinoma cell lines suggested that CSMD1 expression was lost in these lines [1]. Although the region containing CSMD1 is frequently deleted in head and neck squamous cell carcinomas and prostatic adenocarcinomas [3,11-15], point mutations in the gene are relatively rare in primary squamous cancers [16] and in squamous cell carcinoma cell lines (Schmidt, Richter and Scholnick, unpublished). Nonsense or splice junction mutations in CSMD1 have not been reported and not enough is known about the function of the protein to accurately assess the effect of the few missense mutations that have been detected. Thus, if CSMD1 is inactivated in tumors, alternative mechanisms for gene silencing must be operating.
In this paper, we demonstrate that while most squamous cell carcinoma cell lines do not express full length CSMD1 transcripts, nearly all produce abnormal transcripts unlikely to encode functional CSMD1 proteins. Methylation of the DNA preceding CSMD1's first exon is correlated with reduction in the level of expression and cell lines expressing at low levels do not appear to elongate the full 11.5 kb transcript. Other anomalies of expression include incorrect splicing and the use of cryptic promoters. Our data suggest that activation of these promoters may result from the global demethylation of the genome associated with tumorigenesis (reviewed by Ehrlich [17]).
Taken together these data demonstrate that mechanisms other than point mutation are responsible for the aberrant CSMD1 expression in head and neck squamous cell carcinoma cell lines, and these data suggest potential targets for further investigation in primary tumors.
Results
CSMD1 promoter methylation in HNSCC cell lines is correlated with expression levels
Preliminary evidence suggested that CSMD1 expression is lost in head and neck squamous cell carcinomas [1] but that point mutations were rare [[16], and Schmidt, Richter and Scholnick, unpublished]. To date, only two of the 20 cell lines we have tested for CSMD1 expression, UPCI:SCC066 and PCI-13, express large transcripts initiated at the normal CSMD1 promoter. These data suggest that a mechanism(s) other than point mutation must be responsible for the loss of expression. CSMD1's first exon is embedded in a 3.7 kb CpG island (data from the UCSC genome browser [18]) suggesting that promoter methylation might epigenetically silence the gene. To test this hypothesis, we surveyed 32 head and neck cancer cell lines for CSMD1 promoter methylation using the Combined Bisulfite Restriction Analysis (COBRA) technique described by Xiong and Laird (Methods) [19]. COBRA analysis of the three amplicons diagrammed in Figure 1 suggested that 28 of the cell lines (87%) had more promoter methylation than did normal upper aerodigestive epithelium (data not shown).
Figure 1 Positions of the amplicons used for COBRA and bisulfite sequencing relative to CSMD1's first exon and the CpG island. Amplicon 1 extends from -395 to -112 bp, amplicon 2 from +175 to +396 bp, and amplicon 3 from +398 to +718 bp relative to the first base of the transcript. The region between amplicons 1 and 2 is so GC rich that no workable PCR primers could be designed to amplify it after bisulfite conversion. Transcription of CSMD1 is from left to right in the figure.
We selected nine of these cell lines for high resolution analysis of promoter methylation by sequencing of clones from bisulfite converted genomic DNA. This approach has the distinct advantage of allowing determination of the state of all the CpG dinucleotides within an amplicon on an allele by allele basis. Amplicons 1 and 2 have 19 and 20 CpG dinucleotides, respectively. Amplicon 3 could not be examined by this technique because it is unclonable after bisulfite conversion. The methylation data were correlated to CSMD1 expression levels as measured by quantitative RT-PCR using an amplicon spanning exons 1 and 2 (Methods). A pool of cDNA from five normal oropharyngeal epithelium specimens served as a basis for comparison to the cell lines.
Our data from amplicon 1 demonstrate a clear relationship between methylation and the level of expression (Figure 2). The bisulfite sequencing data confirm that there is relatively little promoter methylation in normal tissue (clones 1–20, Figure 2A). This is also the case in cell line UPCI:SCC066 which expresses a large CSMD1 transcript from the normal promoter at a level approximately 33% of that of normal tissue (clones 32–39, Figure 2B). PCI-13, our highest expressing line at 125% of normal epithelium, displays two distinct patterns of promoter methylation with some clones heavily methylated (clones 21–24, and 30) and others with no methylation (clones 25–39 and 31; Figure 2B). This pattern is consistent with either heterozygosity for methylation or the co-existence of 2 distinct populations within the cell line, one heavily methylated and one unmethylated. We cannot distinguish between these two possibilities using the currently available data.
Figure 2 Methylation status of each of the 19 CpG dinucleotides in amplicon 1. Each CpG dinucleotide within the amplicon is represented by a column and its position in base pairs relative to the CpG island is given at the top of the figure. Positions were measured from the end of the CpG nearest the 5' end of CSMD1 as defined by data obtained from the UC Santa Cruz genome browser (island sequence beginning CGGTGTGCGGCGTGAGCTTCCCCCACCCGAG...). Each sequenced clone is represented by a row and is numbered sequentially starting from the top of the figure. Open symbols (○) indicate that cytosine residue of the dinucleotide was unmethylated, closed symbols (●) indicate that it was methylated, and "x" indicates that the identity of the base could not be determined from that sequencing reaction. The level of expression is presented as a percentage relative to five pooled samples of normal UPPP epithelium. These are not the same UPPP samples used for measurement of promoter methylation in normal tissue. A) normal oropharyngeal epithelium from five individuals undergoing uvulopalatopharyngoplasty. B) Methylation status of two cell lines that express large CSMD1 transcripts from the normal promoter (PCI-13 and UPCI:SCC066), C) cell lines with low expression of the CSMD1 transcript from the normal promoter (PCI-2, PCI-1, PCI-100, 094, 041, UPCI:SCC056, and UPCI:SCC104).
The remaining cell lines express CSMD1 at a level half that of UPCI:SCC066 or less (ranging from 17% to 1% of normal epithelium) and they exhibit considerably greater methylation of amplicon 1 (clones 40–76, Figure 2C). Cell lines with more amplicon 1 methylation tend to express the gene at lower levels but the relationship is not strictly quantitative (Figure 2C).
In contrast, our data revealed no relationship between expression level and methylation of amplicon 2 (located towards the 3' end of exon 1, Figure 1). For example, all of the 10 clones of amplicon 2 sequenced from PCI-13 were methylated at 19 or 20 of their 20 CpG dinucleotides. UPCI:SCC066, on the other hand, has nearly no methylation in amplicon 2 with only a single methylated CpG dinucleotide detected in one clone out of the nine sequenced. Amplicon 2 ranges from completely unmethylated to heavily methylated in the seven remaining cell lines (data not shown).
Low transcript levels are accompanied by a failure to elongate the full CSMD1 transcript
On the surface, the quantitative RT-PCR data presented in Figure 2 suggest that the cell lines we consider low expressing might still have up to 17% of the normal level of CSMD1 transcript. A survey of 20 cell lines using a battery of RT-PCR primer pairs located throughout the 11.5 kb transcript reveals that this is not the case. These lines included OKF6-TERT1, a TERT immortalized, p16 deficient but untransformed oral keratinocyte cell line [20].
Our data suggest that the low expressing cell lines shown in Figure 2C express considerably more of the 5' end of the 11.5 kb transcript than they do exons further 3', a phenomenon well illustrated by cell line PCI-100. This line expresses the exon 1/exon 2 amplicon at approximately 15% of the level of normal epithelium. In contrast, we had previously reported that CSMD1 transcripts were not detectable in this line by combined RT-PCR and Southern blotting using three sets of intron spanning primers [1]. The most 5' of those amplicons spans exons 9 through 26.
Analysis with additional primer pairs resolves this apparent paradox by demonstrating that the amount of transcript declines sharply and reproducibly as one examines progressively more 3' exons (Figure 3). The same effect is seen using either oligo-dT or random hexamer primed cDNA. No transcript of this structure has been detected in normal epithelium nor have we detected any sequence alterations in PCI-100 that would explain why the full transcript is not expressed. It is not clear whether PCI-100 produces a small number of discrete size classes of transcript, if individual transcripts terminate at random points within the very large introns in this part of the gene (the first 10 introns average ~150 kb in length), or if the short transcripts result from the elevated activity of a previously undetected posttranscriptional control mechanism.
Figure 3 RT-PCR data demonstrating the preferential loss of longer CSMD1 transcripts in cell line PCI-100. Exons 1 and 2 were amplified with prm1542 (gggacccgatgctatgagagggaag) and prm1482 (cccgtgaggaaaccctgggctct), exons 2 through 4 with prm2078 (gggcgagcgcaataggatacagtt) and prm1382 (ggatggcgtggccttccaagatgtag), exons 4 through 6 with prm1392 (agctgcctccctggctacatcttgg) and prm1405 (cttggaactgagcgttaaatcctttg), and exons 7 through 11 with prm1463 (tgaaaaaggcgattgagttgaagtc) and prm1421 (gaccgatctggtgtctcccaccttc). "STD" indicates a lane of DNA standards whose sizes are given on the left side of the figure, "HFB" = human fetal brain, "blank" indicates a control reaction with water substitute for cDNA.
Inactivation of CSMD1 by aberrant splicing
Two cell lines, UPCI:SCC066 and PCI-13, express large CSMD1 transcripts initiated at the normal promoter. Subsequent finer scale analysis demonstrates that PCI-13's transcript lacks exons 4 and 5, resulting in a frameshift-induced nonsense codon in exon 6 (Figure 4). Sequencing of the PCI-13 RT-PCR product demonstrates the direct juxtaposition of wildtype exons 3 and 6 and that the transcript contains no novel sequences or splices that would prevent the frameshift. UPCI:SCC066 produces two transcripts, a normal one that includes exons 4 and 5 and another that lacks them (Figure 4). RT-PCR of human fetal brain cDNA reveals very low levels of an RT-PCR product corresponding in size to that expected from the internally deleted transcript. This band is not readily visible at the exposure used for Figure 4A. We have not detected a similar sized PCR product in RNA from oropharyngeal epithelium but this may reflect the fact that CSMD1 transcripts are ~10x less abundant in oropharyngeal epithelium than they are in fetal brain (data not shown).
Figure 4 A) RT-PCR demonstrating deletion of exons 4 and 5 from the CSMD1 transcripts of cell lines PCI-13 and UPCI:SCC066. RT-PCR with a forward primer in exon 3 (prm2080, gggatttcagctgccctcctctat) and a reverse primer in exon 6 (prm1405, cttggaactgagcgttaaatcctttg) yields a 623 bp product from upper aerodigestive tract (UPPP), human fetal brain (HFB) and UPCI:SCC066 RNA. Exclusion of exons 4 and 5 from the transcripts of PCI-13 and UPCI:SCC066 reduces the size of the PCR product to 220 bp. The sizes of the DNA standards (STD) are indicated on the left side of the figure. The blank reaction contained water instead of cDNA. "UPPP" = normal oropharyngeal epithelium isolated from uvulopalatopharyngoplasty surgical discards. B) Idiogram showing the organization of exons 3 through 6 in the human genome. The sizes of the introns were obtained from the UCSC genome browser. The sequences of the splice junctions are provided above the idiogram with intronic bases in lowercase and exonic bases in uppercase. The structure of the UPCI:SCC066 and PCI-13 CSMD1 transcripts are shown below the idiogram.
The transcripts lacking exons 4 and 5 appear to result from aberrant splicing rather than somatic deletion of these two exons or mutations of their splicing consensus sequences. Exons 4 and 5 can be amplified from PCI-13 genomic DNA and sequencing of those PCR products demonstrates that both their coding sequences and consensus splice sites are wildtype.
Activation of cryptic promoters in cancer cell lines
The RT-PCR survey revealed a second transcriptional anomaly exhibited by 4 cell lines: SCC9, 041, PCI-1 and PCI-2. Like PCI-100, these lines express low levels of the very 5' end of the transcript and even lower levels of more 3' exons within the first half of the transcript. However, these lines are distinct in expressing higher levels of the 3' half of the transcript, suggesting that alternative promoters in the middle of the gene may be used. SCC9 was chosen for further study because it expresses the 3' half of the transcript at a level dramatically higher than normal for oropharyngeal epithelium. Northern blotting detects a comparatively abundant 6.4 kb truncated transcript as well as smaller amounts of an 8.7 kb transcript in SCC9 (Figure 5A). The other three cell lines express their truncated CSMD1 transcripts at lower levels (data not shown). Sequence analysis of CSMD1 cDNA clones from SCC9 demonstrates that many transcripts are improperly spliced, resulting in retention of intronic sequences and/or deletion of exonic sequences (data not shown). In particular, retention of sequences from intron 40 is common. The high frequency of faulty splicing may explain the broadness of the 6.4 kb CSMD1 band in Figure 5A and suggests that the 8.7 kb transcript may also be incompletely or improperly spliced.
Figure 5 A) Northern blot showing the truncated CSMD1 transcript expressed by the SCC9 cell line. Approximately 8 μg of poly-A+ RNA was fractionated on a denaturing 1% agarose gel, transferred to a filter and probed with the full length human CSMD1 cDNA. The image was acquired by phosphorimager scanning. B) Structure of the aberrant SCC9 transcript and the corresponding segment of the wildtype CSMD1 transcript showing the relationship between the novel cryptic promoter and exon 37. The partial Alu element contains only the first 120 base pairs of the normal ~300 bp Alu sequence.
5' RACE [21] reveals that the SCC9 message is initiated just upstream of an Alu element in intron 36 (Figure 5B). Only the 5'-most 120 base pairs of the Alu element are present in the genome. RT-PCR using a forward primer specific for the novel sequences of the SCC9 transcript (prm1904, cgtttagttcgacacacttcatgt) demonstrates that cell lines 041, PCI-1 and PCI-2 do not initiate their CSMD1 transcripts at the same point, suggesting that other cryptic promoters are active in these lines. The sequence of this novel exon has been entered in Genbank as accession number DQ093422.
DNA methyltransferase inhibitors activate the same cryptic promoter used in cell line SCC9
Expression from epigenetically silenced promoters can sometimes be restored by treatment with inhibitors of DNA methyltransferase or histone deacetylase activity ([22]). We selected two low expressing cell lines with promoter methylation, UPCI:SCC104 and 094, for treatment with various concentrations of 5-azacytidine or 5-aza-2'-deoxycytidine (5-aza-dC) as well as combinations of either of those drugs with the histone deacetylase inhibitor trichostatin A. These treatments did not reactivate the silenced CSMD1 promoter. COBRA analysis of genomic DNA from the treated cells suggested that the drugs did not robustly affect methylation of the CSMD1 promoter even at levels high enough to be toxic to the cells. These experiments did however shed light on the cryptic promoter used in SCC9 cells and on the interpretation of experiments using methyltransferase inhibitors.
Treatment of cell line 094 with relatively high doses of 5aza-dC results in the expression of the 3' end of the CSMD1 transcript. This transcript was not detected in control 094 cells undergoing mock drug treatment (Figure 6) nor was it detected in 094 cells growing under normal culture conditions (data not shown). RT-PCR of cDNA from drug-treated 094 cells using the primer developed from the novel 5' exon of SCC9's truncated transcript (prm1904, see above) yielded a product identical in size to that amplified from SCC9 cDNA (Figure 6). The identity of the product was confirmed by DNA sequencing which also revealed that the drug-induced 094 transcript was more faithfully spliced than the transcript expressed in SCC9 (data not shown).
Figure 6 5aza-dC induced activation of CSMD1 from the same cryptic promoter active in cell line SCC9. Cultures of cell line 094 were treated with the indicated micromolar concentrations of 5aza-dC, and used for RNA extraction. Cell line SCC9 was grown in the absence of the drug and used as a control. RT-PCR was performed with primers prm1904 (cgtttagttcgacacacttcatgt) and prm1880 (cactggaaggagagcacgtcgttcac). prm1904 is specific for the cryptic first exon first used in SCC9, prm1880 is homologous to CSMD1 exon 38. The presence or absence of reverse transcriptase in the RT reaction is indicated by a + or -, respectively. The blank reaction contained water instead of input cDNA. The sizes of the markers are indicated on the left side of the figure.
Discussion
Our data clearly demonstrate that expression of normal CSMD1 transcripts is rare in head and neck squamous cell carcinoma cell lines. Of the HNSCC cell lines examined, only UPCI:SCC066 appears to express a normal transcript from the expected promoter. Even that cell line produces a second species of aberrantly spliced transcript lacking internal exons. Our data suggest that epigenetic modification of the DNA 5' of the transcription start site may contribute to the down-regulation of CSMD1. In addition, a low level of expression appears to be associated with production of prematurely terminated transcripts. This degree of complexity might be expected from a 2 megabase, 70 exon gene.
Methylation of a specific region of the CpG island, -395 to -112 bp relative to the transcriptional start site (amplicon 1), appears to be correlated with the activity of the normal CSMD1 promoter. In contrast, methylation of amplicon 2, located within the first exon, shows no such relationship. Our data suggest that the relationship between the amount of methylation in amplicon 1 and the level of expression may not be strictly quantitative. Differences between cell lines with amplicon 1 methylation could arise through a number of mechanisms, for example, variations in the levels of transcription factors between cell lines. In cases where there is considerable heterogeneity in the methylation pattern within a cell line like PCI-100, alleles with less methylation may be expressed at higher levels than those more heavily methylated (compare clones 51 and 52 to clone 58 in Figure 2C). Alternatively, the presence of methylation in amplicon 1 could be a qualitative but not strictly quantitative indicator of methylation of a critical segment of the promoter not discovered in this study.
The normal CSMD1 promoter was not reactivated by drugs that inhibit DNA methyltransferases and histone deacetylases, nor did the drugs abolish CSMD1 promoter methylation, even at toxic doses. Not all genes with promoter methylation respond to such treatments [23]. These drug treatments did, however, provide a potential explanation for the use of a normally cryptic promoter by cell line SCC9. The CSMD1 transcript in this line is initiated near a partial Alu element. 5-aza-dC treatment of cell line 094 activates the same cryptic promoter. This suggests that cryptic promoters may be naturally activated by general hypomethylation of the genome in cancer cells and the subsequent release of repetitive elements from epigenetic repression (reviewed by Ehrlich [17]). The SCC9 transcript does not appear to encode a functional protein but, with a very large gene like CSMD1, there is a potential for some abnormally initiated transcripts to encode truncated proteins with dominant negative properties.
The second ramification of this finding is for the interpretation of data obtained by treating cells with methyltransferase inhibitors. Detection of CSMD1 transcripts solely with primers mapping to the 3' end of the gene could have been erroneously interpreted as representing reactivation of the normal promoter. It seems imperative that such experiments demonstrate that transcripts detected after drug treatment are actually initiated at the normal promoter.
Aberrant splicing also appears to play a role in the production of defective CSMD1 transcripts. Loss of splicing fidelity has been proposed as a characteristic of cancer cells [24,25] and this would be consistent with the variety of misspliced transcripts we detected from SCC9. However, the removal of exons 4 and 5 from the CSMD1 transcript in PCI-13 may reflect a more specific phenomenon than a general inability to splice large introns; this line is still capable of splicing large introns as evidenced by its successful splicing of exon 3 to exon 6, eliminating an intron of over 666 kb. The failure to include exons 4 and 5 may be due to inactivation of a splicing enhancer in intron 3, or to less efficient splicing due to the fact that exons 4 and 5 do not begin with the consensus G residue (Figure 4B) [26].
Conclusion
Taken together, our data suggest that CSMD1 function is lost in head and neck squamous cell carcinoma cell lines through a variety of mechanisms other than point mutagenesis. Epigenetic modifications of amplicon 1 and defective splicing appear to be fruitful areas to explore in primary head and neck squamous cancers.
Methods
Cell lines and tissue samples
DNA from HNSCC cell lines UMSCC9, UMSCC35, UMSCC37, UMSCC38, UMSCC45, UMSCC49, UMSCC65, UMSCC68, and UMSCC76 was provided by Dr. Thomas Carey, University of Michigan [27]. Dr. Ruud Brakenhoff, Vrije Universitat, provided cell lines 040, 041, and 094 [28]; Dr. Theresa Whiteside, University of Pittsburgh, provided cell lines PCI-1, PCI-2, PCI-4B, PCI-13, PCI-30, PCI-50, PCI-51, PCI-52, PCI-100 [29], SCC4, SCC9 [30,31], and UPCI:SCC068, UPCI:SCC74, UPCI:SCC104, UPCI:SCC182, UPCI:SCC203, and UPCI:SCC220 [developed by Dr. Susanne Gollin, University of Pittsburgh, 32]. Dr. Gollin provided cell lines UPCI:SCC056, UPCI:SCC066, and UPCI:SCC114 [14,16,32]. The immortal but untransformed keratinocyte line OKF6-TERT1 was obtained from Dr. James Rheinwald, Harvard University [20].
Normal oropharyngeal epithelium was isolated from discarded tissue from uvulopalatopharyngoplasties (UPPP) collected anonymously with the approval of the Washington University Human Studies Committee.
Cell Culture and Tissue Preparation
Squamous cell carcinoma cell lines were grown in DMEM or DMEM:F-12, 1:1 Mixture (BioWhittaker) containing 10% fetal bovine serum (Sigma). DMEM medium was supplemented with 1X MEM Nonessential Amino Acids (BioWhittaker). Upper aerodigestive tract epithelium was separated from the rest of the UPPP specimen by digestion with Dispase II (Roche) using a protocol adapted from Oda and Watson [33].
Nucleic acid preparation and bisulfite conversion
Genomic DNA was isolated by using either Nucleospin Tissue kits (Clontech), QIAamp DNA Blood Mini kits (Qiagen) or Trizol (Invitrogen) according to the manufacturers' instructions. Total RNA isolation, synthesis of first strand cDNA, RT-PCR and 5' RACE PCR were performed essentially as previously described [1]. Poly-A+ RNA for Northern blotting was selected from total RNA using Oligotex beads (Qiagen). Northern blotting and hybridization were performed as previously described [1]. cDNA synthesis was primed using oligo dT or random primers and extended by either Thermoscript or Superscript III reverse transcriptase (Invitrogen). PCR was run in Perkin-Elmer 480 or Applied Biosystems 9700 thermal cyclers for 35 cycles unless otherwise noted. Images of ethidium bromide stained gels were captured with a Gel-Doc imaging station (Biorad). Quantitative PCR was run in an Applied Biosystems 5700 thermal cycler using SYBR Green Master Mix (Applied Biosystems). Primers prm2426 (gtgtggagtatctgcagacatga) and prm2427 (ctggactaagcctccacagttct) were used to amplify a 132 base segment spanning CSMD1's first and second exons. An amplicon from human 18S RNA was used as a basis for comparisons across cell lines (primers prm2396, ttcggaactgaggccatgat and prm2397, tttcgctctggtccgtcttg). Calculations were performed using the ΔΔCt method in GeneAmp 5700 SDS software (version 1.3) and Microsoft Excel. Quantitation was based on the average values obtained from duplicate reactions. The level of CSMD1 expression in normal oropharyngeal epithelium was determined from pooled cDNA from five UPPP specimens.
We used the CpGenome DNA Modification kit (Intergen) for bisulfite conversion of the genomic DNA according to the manufacturer's protocol, with the following exception. Incubation of the conversion reaction was carried out in a thermal cycler for six cycles each consisting of three minutes at 94°C followed by three hours at 50°C (Christina Menke and Paul Goodfellow, personal communication).
Analysis of CSMD1 Promoter Methylation
Methylation of three segments of the CSMD1 CpG island was examined using the Combined Bisulfite Restriction Analysis technique (COBRA) [19]. All three segments were amplified by using the nested primers and PCR conditions listed in Table 1. Amplicon 1 extends from -395 to -112 bp, amplicon 2 from +175 to +396 bp, and amplicon 3 from +398 to +718 bp relative to the first base of the transcript. A small region surrounding the transcription start site (-111 to +174 bp) could not be examined because no PCR primers could be designed from its extremely GC-rich sequence. The first round PCR used 2 μl of bisulfite converted genomic DNA in a final volume of 10 μl. Subsequent amplifications with nested primers used 4 μl of first round product as template in reactions with a final volume of 20 μl. All PCR was carried out for 35 cycles. A portion of the second round PCR product was run on a 1.5% agarose gel, stained with ethidium bromide, and quantified using the ImageQuant software package (v1.2 for Macintosh, Molecular Dynamics) so that equal amounts of each could be used in restriction digests.
Table 1 PCR primers used for COBRA analysis of CSMD1 CpG island
PCR round forward reverse annealing temp [MgCl2] (mM) amplicon size (bp) # of CpG's # BstU1 sites # Taq1 sites
Amplicon 1
1 st prm1998 taagttaggtagggggttgtttt prm1999 aaccactacaaaactaaactact 45°C 1.5 626
2 nd prm2000 ggaagggagattaaaggatgg prm2001 aaactcaaccatccttacccacaa 58°C 1.5 298 19 1 4
Amplicons 2 and 3
1 st prm1942 gagtagtttagttttgtagtggt prm1943 tattaaattcctttctccttaaca 45°C 2.0 594
2 nd prm2006 agtagtttagttttgtagtggtt prm2007 caatcatatctacaaatactcc 45°C 2.0 225 20 2 1
2 nd prm1954 tgtggagtatttgtagatatgattg prm1968 cctttctccttaacaccctatacta 55°C 1.5 388 31 4 1
Restriction digests for COBRA were performed with either BstU I or Taqα I (5 or 10 units per reaction, respectively; New England Biolabs) for 4 hours in a final volume of 10 μl. Taqα I digests were performed only when no methylation was detected with BstU I. BstU I digests also included an internal control DNA fragment to confirm complete digestion. This DNA fragment contains a single BstU I site and was amplified from a cloned CSMD1 cDNA using primers prm2020 (agatcccccagtgtctccctgtgt) and prm2021 (actgctggtgccgtggtaatgact). The control PCR product is 1019 bp long and is digested to two fragments of 605 and 414 bp by BstU I. Digestion products were fractionated on a 10% polyacrylamide gel, stained with ethidium bromide, and visualized with a Gel-Doc video imaging workstation (Bio-Rad).
High resolution analysis of methylation was performed by sequence analysis of individual clones from amplicons 1 and 2. DNA from amplicon 3 proved unclonable and gel electrophoresis suggests that its very AT rich sequence results in a bent DNA configuration. PCR products were purified using the Nucleospin Extraction columns (Clontech) and inserted into the pCR2.1-TOPO vector using the TOPO TA Cloning kit (Invitrogen) according to the manufacturer's instructions. Plasmid DNA from individual colonies was isolated using the Nucleospin Plus Plasmid Miniprep kit (Clontech) and sequenced with a reverse vector primer (agcggataacaatttcacacagga) using fluorescence based sequencing with Big Dye Terminator mix (Applied Biosystems).
Treatment of cultured cells with DNA methyltransferase inhibitors
Cell line 094 was treated with 5-aza-2'-deoxycytidine (5aza-dC) (Sigma) dissolved in DMSO. Two 100 mm cell culture dishes containing 5 × 105 cells were established for each of the drug concentrations tested. Cells were grown for 72 hours in media containing DMEM:F-12, 1:1 Mixture (BioWhittaker) with 1X MEM Nonessential Amino Acids (BioWhittaker) and 10% fetal bovine serum and then switched to media containing 5aza-dC at concentrations of 0 μm, 5 μM, 25 μM, or 100 μM. Cells were fed daily for 4–5 days and then both plates were harvested in 3 ml of Trizol (Invitrogen) for isolation of RNA and DNA according to the manufacturer's instructions. RT-PCR used for detection of CSMD1 transcripts in these treated cells was run for 40 cycles.
Abbreviations
5aza-dC = 5-aza-2'deoxycytidine, COBRA = Combined Bisulfite Restriction Analysis, CSMD1 = Cub and sushi multiple domains 1, HNSCC = head & neck squamous cell carcinoma, RT-PCR = reverse transcription – polymerase chain reaction, SCC = squamous cell carcinoma, UPPP = uvulopalatopharyngoplasty.
Competing interests
The author(s) declare that they have no competing interests
Authors' contributions
TMR performed the DNA methylation analysis, and parts of the transcript survey. BDT cloned and characterized the novel first exon expressed in cell line SCC9. SBS performed parts of the transcript survey and characterized the deletion of exons 4 and 5 in PCI-13. All three authors participated in the analysis of the data and in the writing of the manuscript.
Acknowledgements
The authors wish to thank Dr. Paul Goodfellow and Christina Menke for their considerable help with the COBRA techniques, Drs. Theresa Whiteside, Susanne Gollin, Ruud Brakenhoff and Thomas Carey for SCC cell lines or cell line DNAs, Dr. James Hartmann for the UPPP specimens and Dr. Paul Goodfellow for his critical comments on the manuscript. This project was supported by grant CA58473 from the National Institutes of Health to SBS.
==== Refs
Sun PC Uppaluri R Schmidt AP Pashia ME Quant EC Sunwoo JB Gollin SM Scholnick SB Transcript map of the 8p23 putative tumor suppressor region Genomics 2001 75 17 25 11472063 10.1006/geno.2001.6587
Scholnick SB Haughey BH Sunwoo JB El-Mofty SK Baty JD Piccirillo JF Zequeira MR Chromosome 8 allelic loss and the outcome of patients with squamous cell carcinoma of the supraglottic larynx Journal of the National Cancer Institute 1996 88 1676 1682 8931613
Washburn JG Wojno KJ Dey J Powell IJ Macoska JA 8pter-p23 deletion is associated with racial differences in prostate cancer outcome Clinical Cancer Research 2000 6 4647 4652 11156215
Bockmühl U Ishwad CS Ferrell RE Gollin SM Association of 8p23 deletions with poor survival in head and neck cancer Otolaryngology–Head and Neck Surgery 2001 124 451 455 11283506 10.1067/mhn.2001.114794
Paris PL Andaya A Fridlyand J Jain AN Weinberg V Kowbel D Brebner JH Simko J Watson JE Volik S Albertson DG Pinkel D Alers JC Van Der Kwast TH Vissers KJ Schroder FH Wildhagen MF Febbo PG Chinnaiyan AM Pienta KJ Carroll PR Rubin MA Collins C Van Dekken H Whole genome scanning identifies genotypes associated with recurrence and metastasis in prostate tumors Human Molecular Genetics 2004 13 1303 1313 15138198 10.1093/hmg/ddh155
Wright K Wilson PJ Kerr J Do K Hurst T Khoo SK Ward B Chenevix-Trench G Frequent loss of heterozygosity and three critical regions on the short arm of chromosome 8 in ovarian adenocarcinomas Oncogene 1998 17 1185 1188 9764830 10.1038/sj.onc.1202028
Muscheck M Sükösd F Pesti T Kovacs G High density deletion mapping of bladder cancer localizes the putative tumor suppressor gene between loci D8S504 and D8S264 at chromosome 8p23.3 Laboratory Investigation 2000 80 1089 1093 10908154
Henshall SM Afar DEH Hiller J Horvath LG Quinn DI Rasiah KK Gish K Willhite D Kench JG Gardiner-Garden M Stricker PD Scher HI Grygiel JJ Agus DB Mack DH Sutherland RL Survival analysis of genome-wide gene expression profiles of prostate cancers identifies new prognostic targets of disease relapse Cancer Res 2003 63 4196 4203 12874026
Lau WL Scholnick SB Identification of two new members of the CSMD gene family Genomics 2003 82 412 415 12906867 10.1016/S0888-7543(03)00149-6
Shimizu A Asakawa S Sasaki T Yamazaki S Yamagata H Kudoh J Minoshima S Kondo I Shimizu N A novel giant gene CSMD3 encoding a protein with CUB and sushi multiple domains: a candidate gene for benign adult familial myoclonic epilepsy on human chromosome 8q23.3-q24.1 Biochemical & Biophysical Research Communications 2003 309 143 154 12943675 10.1016/S0006-291X(03)01555-9
Sunwoo JB Holt MS Radford DM Deeker C Scholnick SB Evidence for multiple tumor suppressor genes on chromosome arm 8p in supraglottic laryngeal cancer Genes, Chromosomes and Cancer 1996 16 164 169 8814448 10.1002/(SICI)1098-2264(199607)16:3<164::AID-GCC2>3.0.CO;2-X
Wu CL Roz L Sloan P Read AP Holland S Porter S Scully C Speight PM Thakker N Deletion mapping defines three discrete areas of allelic imbalance on chromosome arm 8p in oral and oropharyngeal squamous cell carcinomas Genes, Chromosomes and Cancer 1997 20 347 353 9408750 10.1002/(SICI)1098-2264(199712)20:4<347::AID-GCC5>3.0.CO;2-1
Sunwoo JB Sun PC Gupta VK Schmidt AP Scholnick SB Localization of a putative tumor suppressor gene in the sub-telomeric region of chromosome 8p Oncogene 1999 18 2651 2655 10353609 10.1038/sj.onc.1202838
Ishwad CS Shuster M Bockmühl U Thakker N Shah P Toomes C Dixon M Ferrell RE Gollin SM Frequent allelic loss and homozygous deletion in chromosome band 8p23 in oral cancer International Journal of Cancer 1999 80 25 31 9935228 10.1002/(SICI)1097-0215(19990105)80:1<25::AID-IJC6>3.0.CO;2-G
Sun PC Schmidt AP Pashia ME Sunwoo JB Scholnick SB Homozygous deletions define a region of 8p23.2 containing a putative tumor suppressor gene Genomics 1999 62 184 188 10610711 10.1006/geno.1999.6020
Toomes C Jackson A Maguire K Wood J Gollin S Ishwad C Paterson I Prime S Parkinson K Bell S Woods G Markham A Oliver R Woodward R Sloan P Dixon M Read A Thakker N The presence of multiple regions of homozygous deletion at the CSMD1 locus in oral squamous cell carcinoma question the role of CSMD1 in head and neck carcinogenesis Gene Chromosomes Cancer 2003 37 132 140 12696061 10.1002/gcc.10191
Ehrlich M DNA methylation in cancer: too much, but also too little Oncogene 2002 21 5400 5413 12154403 10.1038/sj.onc.1205651
Kent WJ Sugnet CW Furey TS Roskin KM Pringle TH Zahler AM Haussler D The human genome browser at UCSC Genome Res 2002 12 996 1006 12045153 10.1101/gr.229102. Article published online before print in May 2002
Xiong Z Laird PW COBRA: a sensitive and quantitative DNA methylation assay Nucleic Acids Research 1997 25 2532 2534 9171110 10.1093/nar/25.12.2532
Dickson MA Hahn WC Ino Y Ronfard V Wu JY Weinberg RA Louis DN Li FP Rheinwald JG Human keratinocytes that express hTERT and also bypass a p16(INK4a)-enforced mechanism that limits life span become immortal yet retain normal growth and differentiation characteristics Mol Cell Biol 2000 20 1436 1447 10648628 10.1128/MCB.20.4.1436-1447.2000
Frohman MA Dush MK Martin GR Rapid production of full-length cDNAs from rare transcripts: amplification using a single gene-specific oligonucleotide primer Proceedings of the National Academy of Sciences of the United States of America 1988 85 8998 9002 2461560
Karpf AR Jones DA Reactivating the expression of methylation silenced genes in human cancer Oncogene 2002 21 5496 5503 12154410 10.1038/sj.onc.1205602
Costello JF Fruhwald MC Smiraglia DJ Rush LJ Robertson GP Gao X Wright FA Feramisco JD Peltomaki P Lang JC Schuller DE Yu L Bloomfield CD Caligiuri MA Yates A Nishikawa R Su Huang H Petrelli NJ Zhang X O'Dorisio MS Held WA Cavenee WK Plass C Aberrant CpG-island methylation has non-random and tumour-type-specific patterns Nat Genet 2000 24 132 138 10655057 10.1038/72785
Lee MP Feinberg AP Aberrant splicing but not mutations of TSG101 in human breast cancer Cancer Research 1997 57 3131 3134 9242438
Wang Z Lo HS Yang H Gere S Hu Y Buetow KH Lee MP Computational analysis and experimental validation of tumor-associated alternative RNA splicing in human cancer Cancer Research 2003 63 655 657 12566310
Shapiro MB Senapathy P RNA splice junctions of different classes of eukaryotes: sequence statistics and functional implications in gene expression Nucleic Acids Research 1987 15 7155 7174 3658675
Lansford CD Grenman R Bier H Somers KD Kim SY Whiteside TL Clayman G Welkoborsky HJ Carey TE Masters JRW, Palsson BO Head and Neck Cancers Human Cell Culture 1999 2 Dordrecht , Kluwer Academic Publishers 185 255
Hermsen MAJA Joenje H Arwert F Welters MJP Braakhuis BJM Bagnay M Westerveld A Slater R Centromeric breakage as a major cause of cytogenetic abnormalities in oral squamous cell carcinoma Genes, Chromosomes and Cancer 1996 15 1 9 8824719 10.1002/(SICI)1098-2264(199601)15:1<1::AID-GCC1>3.0.CO;2-8
Heo DS Snyderman C Gollin SM Pan S Walker E Deka R Barnes EL Johnson JT Herberman RB Whiteside TL Biology, cytogenetics, and sensitivity to immunological effector cells of new head and neck squamous cell carcinoma lines Cancer Research 1989 49 5167 5175 2766286
Rheinwald JG Beckett MA Defective terminal differentiation in culture as a consistent and selectable character of malignant human keratinocytes Cell 1980 22 629 632 6160916 10.1016/0092-8674(80)90373-6
Rheinwald JG Beckett MA Tumorigenic keratinocyte lines requiring anchorage and fibroblast support cultures from human squamous cell carcinomas Cancer Research 1981 41 1657 1663 7214336
Huang X Gollin SM Raja S Godfrey TE High-resolution mapping of the 11q13 amplicon and identification of a gene, TAOS1, that is amplified and overexpressed in oral cancer cells Proceedings of the National Academy of Sciences of the United States of America 2002 99 11369 11374 12172009 10.1073/pnas.172285799
Oda D Watson E Human oral epithelial cell culture I. Improved conditions for reproducible culture in serum-free medium In Vitro Cell Dev Biol 1990 26 589 595 2358421
|
16153303
|
PMC1239921
|
CC BY
|
2021-01-04 16:24:16
|
no
|
Cancer Cell Int. 2005 Sep 9; 5:29
|
utf-8
|
Cancer Cell Int
| 2,005 |
10.1186/1475-2867-5-29
|
oa_comm
|
==== Front
Chiropr OsteopatChiropractic & Osteopathy1746-1340BioMed Central London 1746-1340-13-191614657810.1186/1746-1340-13-19Case ReportFostering critical thinking skills: a strategy for enhancing evidence based wellness care Jamison Jennifer R [email protected] School of Chiropractic, Murdoch University, South Street, Perth, Western Australia, 6849, Australia2005 8 9 2005 13 19 19 19 6 2005 8 9 2005 Copyright © 2005 Jamison; licensee BioMed Central Ltd.2005Jamison; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Chiropractic has traditionally regarded itself a wellness profession. As wellness care is postulated to play a central role in the future growth of chiropractic, the development of a wellness ethos acceptable within conventional health care is desirable.
This paper describes a unit which prepares chiropractic students for the role of "wellness coaches". Emphasis is placed on providing students with exercises in critical thinking in an effort to prepare them for the challenge of interfacing with an increasingly evidence based health care system.
Methods
This case study describes how health may be promoted and disease prevented through development of personalized wellness programs. As critical thinking is essential to the provision of evidence based wellness care, diverse learning opportunities for developing and refining critical thinking skills have been created. Three of the learning opportunities are an intrinsic component of the subject and, taken together, contributed over 50% of the final grade of the unit. They include a literature review, developing a client wellness contract and peer evaluation. In addition to these 3 compulsory exercises, students were also given an opportunity to develop their critical appraisal skills by undertaking voluntary self- and unit evaluation. Several opportunities for informal self-appraisal were offered in a structured self-study guide, while unit appraisal was undertaken by means of a questionnaire and group discussion at which the Head of School was present.
Results
Formal assessment showed all students capable of preparing a wellness program consistent with current thinking in contemporary health care. The small group of students who appraised the unit seemed to value the diversity of learning experiences provided. Opportunities for voluntary unit and self-appraisal were used to varying degrees.
Unit evaluation provided useful feedback that led to substantial changes in unit structure.
Conclusion
Students have demonstrated themselves capable of applying critical thinking in construction of evidence based wellness programs. With respect to unit design, selective use of learning opportunities highlighted the desirability of using obligatory learning opportunities to ensure exposure to core constructs while student feedback was found to provide useful information for enriching unit review.
It is hoped inclusion of critical thinking learning opportunities in the undergraduate chiropractic curriculum will contribute to the development of an evidence based ethos in chiropractic care.
Chiropracticcritical thinking skillswellness
==== Body
Background
Health care has long been regarded as an art and a science. In contemporary conventional health care the 'science' dimension has increasingly come to dominate the 'art' of health care. At the undergraduate level this has been expressed as enhanced emphasis in the training of future physicians on searching and critically evaluating the available literature utilizing electronic and other databases [1]. At the level of the health care system allopathic disciplines are encouraging critical and empirical evaluation of alternative medical techniques [2,3]. Evidence based medicine {EBM} has become the new health care mantra and is largely pursued through critical evaluation of individual research studies, systematic reviews of studies in a particular area or practice, evidence-based practice guidelines outlining standards for the profession, and evidence-based systems of care focusing on implementation [4]. In each of these pursuits critical thinking emerges as a requisite skill.
Despite chiropractic's philosophy of vitalism contrasting sharply with the "mechanistic" foundations of orthodox medicine, there are some in the chiropractic profession who welcome this development. Not only may the development of evidence-based guidelines in chiropractic practice insulate against malpractice lawsuits, they may also improve relations between chiropractic and the health care system and better enable the chiropractic profession achieve is foremost goal of serving as a portal of entry into the health care system with chiropractors functioning as primary contact practitioners.
In addition to chiropractic functioning at the community-health care system interface [5], the chiropractic profession considers itself a provider of wellness care and this is subsumed under the mantel of maintenance care [6]. "Maintenance" or "wellness" care involves regular, ongoing visits that is not correlated directly to symptomatology. However George B. McClelland, DC, Chairman ACA Board of Governors has stated "Philosophically the idea of regular spinal manipulative therapy opposes the concept of wellness" [7]. Furthermore it has been suggested that: "...the proposition of chiropractic as a "wellness profession" is not defensible." [8]. Conventional health care would concur given that there are those in the chiropractic profession whose practice of wellness care is limited to correcting subluxations. While the notion that mechanical and functional disorders of the spine, expressed as subluxations, can degrade health and correction of spinal disorders by adjustments may restore health is fundamental to chiropractic thinking, there is no scientifically acceptable data to support this belief. Furthermore, wellness care calls for a holistic approach and the desirability for the chiropractic profession to explore a more comprehensive approach to wellness care is apparent given the Institute of Alternative Futures report Future of Chiropractic Revisited: 2005 to 2010, which suggested possible growth scenarios for chiropractic were as "wellness coaches" or as "healthy life doctors" with a wellness mindset.
If chiropractic is to evolve as a wellness profession in an increasingly evidence based health care system, it would seem necessary that it critically appraise its current wellness practices and adopt a schema in which its practitioners serve as motivators and educators. One initiative which may contribute to this end is to include in undergraduate education units which encourage critical thinking in the context of health promotion and disease prevention. Murdoch university provides their third year chiropractic students with just such a learning opportunity.
Critical thinking skills are thoughtfully being incorporated into the curriculum of nursing [9,10] and medical programs [11], at both under- and post graduate levels [12-14].
Critical thinking is regarded as purposeful, self-regulatory judgment. In addition to evaluating whether arguments are strong, weak or relevant, critical thinking involves inferring degrees of truth from given data; recognizing unstated assumptions underlying assertions; deducing whether conclusions necessarily follow from given statements and interpreting and weighing evidence to decide if generalizations are warranted [15]. It is commonly accepted that critical thinking can be taught. Diverse learning opportunities have been shown to facilitate the development and acquisition of this skill ranging from concept mapping [10], through critical questioning workshops [11] and systematic literature reviews [13] to problem based learning [14]. Problem based learning programs create scenarios in which prior knowledge is activated in a meaningful context thereby encouraging elaboration and organization of knowledge [16]. Students in problem based curricula demonstrate an enhanced ability to apply science based concepts to their explanations [17]. While problem based learning appears to be particularly useful for refining reasoning skills, integration of critical thinking in all areas of learning has been found a useful strategy for fostering this ability [18].
This paper describes how a preclinical unit has been structured to include diverse learning opportunities for applying critical thinking skills in the context of wellness. It illustrates how students can be given opportunities to practice critical thinking as a prelude to practicing evidence based health care.
Case Presentation
Unit Design
Health Promotion and Nutritional Management is a subject taught in the third year of a 5 year chiropractic program at Murdoch University. The broad aims of this unit are to:
1. Provide the student with a strategy for implementing personal wellness programs in clinical practice.
2. Enable the student to critically explore the contribution of lifestyle interventions, including the use of nutrients in therapeutic doses, in health promotion, disease prevention and management.
3. Alert the student to the early signs and symptoms suggestive of some lifestyle modifiable diseases prevalent in primary practice.
The learning objectives are to:
• Enhance wellness through recruitment of wellness triggers; identification and reduction of lifestyle risk factors; promotion of fitness; and provide early diagnosis and management, using lifestyle interventions and nutritional therapy, of selected diseases prevalent in primary practice.
• Empower patients to take increased personal responsible for their health care through formulation of wellness contracts by performing a personal health status appraisal; screening patients to ascertain their risk of prevalent diseases; negotiating health goals through examination of patient's perceived and professionally assessed health needs; determining potential barriers, including cultural, socio-economic factors, to implementation of health promotion and disease prevention strategies; negotiating a health promotion and disease prevention plan; implementing a personalized health management program; monitoring patient progress and modify the health contract, as required.
• Analyze the patient's preferred interaction style and adapt ones mode of clinical care as required.
• Critically appraise relevant literature and apply evidence-based problem solving to promote wellness.
• Implement a self-care wellness program.
The unit provides a classroom learning experience which runs for 6 weeks, and a structured self-learning guide, complemented by WebCT, a computer based learning platform, which runs for 13 weeks of the semester. The unit has been designed to enhance active and encourage independent learning and provides 5 distinct opportunities for developing and refining critical thinking skills. The 5 critical thinking opportunities provided ranged from client health assessment, peer evaluation and literature review, which together contribute almost 60% of the final grade, to voluntary self-assessment and finally unit evaluation.
1 – Self-Assessment
The self-assessment learning experiences are embedded in the structured self-study learning guide. The learning guide has been structured to provide students with a opportunity to undertake continuous formative self-assessment. Figure 1 shows the template used in the structured self-directed learning guide and depicts the guideposts to the self-assessment critical appraisal opportunities provided by the challenge and review questions and self-care tasks. The factual content of the unit is covered in 25 discrete topics each of which contains a unique learning template. For each topic the student is provided with self-assessment opportunities to:
Figure 1 Acquiring good habits.
• Critically review their learning by completing challenge and review questions based on the content of that topic. The student has the opportunity to monitor their grasp and recall of factual information.
• Apply the information provided in that topic to their lifestyle and formulate a personal wellness program. The student is given the opportunity to preview construction of a wellness program in a non-threatening environment and simultaneously embrace a self-care system based on a lifetime of health choices.
2 – A Client Wellness Program
Students who chose to prepare a personal wellness program are particularly well prepared when required to formulate the formal client wellness program. Formulating a wellness program for a client passes through a number of critical thinking steps. Students are required to undertake critical appraisal of a client's lifestyle with respect to their good and bad habits and, given their family history, ascertain the client's health risk. They are then required to identify health needs and, in negotiation with the client, develop a list of wellness goals. The next steps are to make the client aware of diverse strategies for achieving these goals, help them select and then implement those strategies appropriate to their lifestyle. The student is then required to monitor the client's wellness program and adapt the program as needed to meet ongoing client successes, failures and changing needs. See Figure 2.
Figure 2 Preparation of client wellness program.
3 – Peer Evaluation
The peer evaluation task is closely linked to the wellness program. Students are asked to appraise the wellness contract prepared by another student. They are encouraged to analyze all aspects of the program with a view to making useful suggestions on how the program may be improved. See figure 3. Marks are scored for constructive criticisms that provide feedback which enhances the learning of the program originator and potentially improves the wellness outlook of the client.
Figure 3 Guidelines for peer assessment.
4 – Literature Appraisal
The ability to assess the scientific validity of information is increasingly recognized as an essential competence in a profession which is increasingly embracing the notion of evidence based practice. It is therefore imperative that students are given opportunities to critically evaluate the literature. For this exercise students are required to rank evidence according to the system developed by the Canadian Task Force and the US Preventive Services Task Force [19,20]. The guidelines for the nutritional literature review included as part of the students' formal in this unit can be found in Figure 4[21].
Figure 4 Critiquing the research literature.
Along with the client wellness program and its critique, the students' literature review contributes over half of the total grade for the unit.
5 – Unit Appraisal
In contrast to peer-, client- and literature assessment, students are given an optional opportunity to critically appraise the unit. Unit appraisal takes two forms. An informal questionnaire survey of student opinion initiated by the lecturer, see Figure 5, and a formal group discussion. All students are invited to participate in the group discussion which forms part of the formal School's assessment of the unit. The Head of School is present for and leads these discussions.
Figure 5 The questionnaire.
Results
Summative student assessment found students could competently prepare a client wellness program. Analysis of client wellness programs submitted for formal assessment confirmed that students had mastered the skills required to achieve this objective. All students demonstrated the ability to appraise their client's lifestyle, prepare and monitor a wellness program Most students were demonstrably competent to ascertain their client's individual disease risk or health hazard as based on a family history and lifestyle. All but 2 students commented on the preferred behaviour style of their client and took this into consideration when formulating their wellness program. A few students took their own preferred behaviour style into consideration and analysed how this may be modified to best suit the client.
In contrast to their success at developing a wellness program, formal assessment of the peer appraisal assignment suggested they found critiquing a wellness program more demanding than constructing one. While all students provided satisfactory comment on the structure and content of an others wellness program, some students faltered when required to provide useful information for refining the initial program.
Formal assessment of the students' critical appraisal of the literature found all students capable of searching the literature and extracting relevant papers. Furthermore, most students were able to compare and discuss conflicting research reports and many showed themselves capable of commenting on potential biases resulting from flaws in research design. However, few categorized the level of evidence provided according to the schema proposed by the Canadian and US Preventive Taskforces.
In contrast to the above compulsory critical thinking opportunities, few students availed themselves of the opportunities offered for unit assessment. The unit survey provided insight into the students' appraisal of the unit as a whole as well as specifically provided feedback on their evaluation of various critical thinking opportunities. Of a class of some 60 students, a total of 22 completed the survey. Consistent with the ethos of independent learning, attendance is optional except when students are required to present their critique of the nutrition literature. The unit survey was completed by 17 students who voluntarily attended lectures and by a further 5 students who were required to do their class presentation on the day of the survey.
Half the students participating selected lectures as their most preferred learning style, a finding verified when ranked preferences were analyzed on a Likert type scale. Figure 6 describes the overall unit rating. Eighteen students regarded the unit as highly relevant to their future practice as a chiropractor, 3 were uncertain and 1 felt it was irrelevant. The students' self-assessment of their critical reading/learning opportunity is reported in Figure 7 which provides an overview of the perceived usefulness of the study guide, the essential reading and study questions. Linking study questions with the unit's content provided an opportunity for active learning and critical interpretation of new information. It also provided an opportunity for self-assessment. Two students indicated they had not attempted any of the study questions.
Figure 6 Overall unit rating.
Figure 7 Appraisal of the Structured Self-study guide.
A Likert type scale was used to ascertain which of the learning experiences students perceived as most valuable. Students who indicated they hadn't performed or who had attended less than half of the sessions offered for a particular activity were deemed unqualified to comment and excluded from analysis of that activity. A score of 5 per student was allocated to each activity rated as an excellent learning experience, 4 was allocated for an activity rated as good, 3 for a fair learning experience and 2 per student for activities rated as a waste of time. The score derived was then divided by the number of respondents to that item and the final score was used to rank learning experiences. On this arbitrary scale the most valued learning experiences, WebCT challenge and study questions, each achieved a total of 3.8; the least appreciated, student presentation, a value of 2.57. Figure 8 shows how students appraised the popular WebCT challenge compared to the self-care and student presentation learning experiences. The WebCT challenge provided students with a formative self-assessment opportunity to evaluate the acquisition of factual knowledge which would be later tested in formal summative examination of the unit. Despite this imperative, 7 students had not used the WebCT challenge, similarly 7 had not implemented any self-care tasks.
Figure 8 Appraisal of Diverse Learning Opportunities.
This trend extended to student presentation. Five {5} respondents indicated they had attended less than half the possible student presentations. Student presentations emerged, both in the questionnaire and in small group evaluation of the unit, to be regarded as 'a waste of time'. Clarification identified that although students found the literature search and data analysis to be useful, the classroom format was regarded as 'boring' and too time consuming. This perspective was confirmed by the group of 6 students who attended the formal unit assessment conducted by the Head of School. Despite the negative classroom learning experience, the students attending the formal unit evaluation indicated they regarded the ability to critically analyze the literature an important component of their training. Furthermore, as shown in Figure 9, two out of 3 respondents felt they had the analytical skills to assess the scientific validity of information if they were provided with details of the research methods used, a perception was verified on formal assessment.
Figure 9 Perceived ability after 5 weeks: Students perception of learning.
Based on the learning they had experienced during the first 5 weeks of the semester, students were asked whether they believed themselves capable of preparing a client wellness contract. Figure 9 shows the majority of students judged themselves capable of evaluating a client's good habits, determining and changing a client's bad habits and assessing and performing a non-invasive health hazard appraisal. Formal assessment confirmed their optimism. In contrast the confidence of respondents with regard to their ability to undertake peer evaluation, see Figure 10, was not confirmed on formal assessment.
Figure 10 Perceived ability after 5 weeks: Confidence to undertake critical appraisal.
Discussion
While it is unclear whether the correction of subluxations makes a unique contribution to wellness, it is apparent that care beyond an adjustment is required if chiropractors are to take the role of 'wellness coaches' or "healthy life doctors" in conventional health care. Wellness is a growth industry and the scientific basis of many wellness practices is uncertain. Critical thinking is fundamental to and regarded an important educational objective in the preparation of health professionals as evidence based carers [22]. Problem based learning scenarios have been found to be conducive to developing critical thinking skills in the classroom [14-18] and on the internet [23]. This paper described how by combining classroom interaction with paper based and internet self-study opportunities various learning opportunities have been created to enhance critical thinking in a wellness context.
Upon completion of the unit, formal assessment found students capable of formulating and administering a client wellness program, undertaking peer review and critically appraising the literature. These findings were largely consistent with the perceptions of the small group of students who chose to evaluate the unit. While any extrapolation of the results of the unit evaluation to the whole class is precluded due to the small size of the participating group, the results of this exercise did provide useful information for future planning. Marked discrepancies emerged with respect to the preferred learning opportunities of different students in the respondent group. Given that the majority of students completing the unit assessment were voluntarily attending a classroom learning experience, it was perhaps not surprising that overall they indicated a clear preference for lecture based learning. It seems not unreasonable to surmise that at least some of their colleagues, who chose to omit classroom learning, preferred a more independent scenario. When structuring a unit it may therefore be prudent to consider providing diverse learning scenarios for acquiring similar knowledge, skills and attitudes to cater for the learning needs of different individuals. Another red flag which emerged from this study is the necessity to incorporate compulsory learning opportunities. Although WebCT and self-study questions were the learning opportunities most favored by the majority of respondents, there were those who had not utilized these learning measures. While students with different learning styles may be expected to avail themselves of different learning opportunities, it should be noted that students were aware that these self-assessment learning experiences covered content in a format similar to the proposed end of semester examination. As some students, despite this incentive chose to omit these learning experiences the need for compulsory completion of selected learning task seems advantageous. In unit planning, it would certainly seem desirable to ensure that knowledge and skills considered fundamental to chiropractic practice are included in diverse obligatory tasks.
Consistent with the ethos of student centred learning, student unit evaluation provides useful feedback for future planning. In this instance, unit modifications in response to criticisms leveled at the format of the student presentations promises to enrich the unit for future students. While retaining the central theme of demonstrating proficiency in critically appraising the literature, the delivery mode will be modified from student presentation to student debate. For example, instead of being asked to discuss the scientific basis for the use of Echinacea, the challenge will be for 2 teams to use scientifically justifiable arguments for and against the statement "Echinacea can be used to prevent the common cold".
Conclusion
This paper described diverse learning experiences designed to enhance critical thinking skills in the context of wellness. By using various modalities in diverse problem solving formats the classroom, internet and a study guide have been combined to create independent, structured self-learning situations. Results of summative student assessment showed students capable of developing a personalized client wellness program consistent with current thinking in conventional health care. By providing a diversity of critical thinking learning opportunities, the more fundamental of which are compulsory, it is hoped that this unit will contribute to the graduation of chiropractors better prepared to interface as 'wellness coaches' or 'healthy life doctors' within an evidence based health care system.
==== Refs
Cohen L McMaster's pioneer in evidence-based medicine now spreading his message in England CMAJ 1996 154 388 90 8564911
Vickers AJ Message to complementary and alternative medicine: evidence is a better friend than power BMC Complement Altern Med 2001 1 1 11346455 10.1186/1472-6882-1-1
Koutouvidis N CAM and EBM: arguments for convergence J R Soc Med 2004 97 39 40 14702370 10.1258/jrsm.97.1.39
Villanueva-Russell Y Evidence-based medicine and its implications for the profession of chiropractic Soc Sci Med 2005 60 545 61 15550303 10.1016/j.socscimed.2004.05.017
Jamison JR The chiropractic doctor J Manipulative Physiol Ther 1985 8 245 9 4078504
Jamison JR Rupert R Maintenance care: towards a global description J Canadian Chiropractic Assoc 2001 25 100 105
Personal communication
Nelson CF Lawrence D Triano JJ Bronfort G Perle SM Metz D Chiropractic as spine care: a model for the profession Chiropractic & Osteopathy 2005 13 9 (6 July 2005) 16000175 10.1186/1746-1340-13-9
Baker CM Problem-based learning for nursing: integrating lessons from other disciplines with nursing experiences J Prof Nurs 2000 16 258 66 11033935 10.1053/jpnu.2000.9461
Wheeler LA Collins SK The influence of concept mapping on critical thinking in baccalaureate nursing students J Prof Nurs 2003 19 339 46 14689390 10.1016/S8755-7223(03)00134-0
Loy GL Gelula MH Vontver LA Teaching students to question Am J Obstet Gynecol 2004 191 1752 6 15547559 10.1016/j.ajog.2004.07.071
Abraham RR Upadhya S Torke S Ramnarayan K Clinically oriented physiology teaching: strategy for developing critical-thinking skills in undergraduate medical students Adv Physiol Educ 2004 28 102 4 15319191 10.1152/advan.00001.2004
Lang TA The value of systematic reviews as research activities in medical education Acad Med 2004 79 1067 72 15504773 10.1097/00001888-200411000-00011
Morales-Mann ET Kaitell CA Problem-based learning in a new Canadian curriculum J Adv Nurs 2001 33 13 9 11155104 10.1046/j.1365-2648.2001.01633.x
Thorpe K Loo RC Critical-thinking types among nursing and management undergraduates Nurse Education Today 2003 23 566 574 14554110 10.1016/S0260-6917(03)00102-3
Charlin B Mann K Hansen P The many faces of problem-based learning: A framework for understanding and comparison Medical Teacher 1998 20 323 330 10.1080/01421599880300
Hmelo CE Cognitive consequences of problem-based learning for the early development of medical expertise Teaching & Learning in Medicine 1998 10 92 100 10.1207/S15328015TLM1002_7
Chenoweth L Facilitating the process of critical thinking for nursing Nurse Educ Today 1998 18 281 92 9847713 10.1016/S0260-6917(98)80045-2
Canadian Task Force The periodic health examination Can Med Assoc J 1979 121 1193 254 115569
Report of the US Preventive Services Task Force: Guide to clinical preventive services 1996 2 Baltimore: Williams & Wilkins
Jamison JR Clinical guide to nutrition and dietary supplements in disease management 2003 Edinburgh: Churchill Livingstone 6 Box 1.1
Kawashima A Petrini MA Study of critical thinking skills in nursing students and nurses in Japan Nurse Educ Today 2004 24 286 92 15110438 10.1016/j.nedt.2004.02.001
McGrath D Teaching on the front lines: using the Internet and problem-based learning to enhance classroom teaching Holist Nurs Pract 2002 16 5 13 11845766
|
16146578
|
PMC1239922
|
CC BY
|
2021-01-04 16:38:23
|
no
|
Chiropr Osteopat. 2005 Sep 8; 13:19
|
utf-8
|
Chiropr Osteopat
| 2,005 |
10.1186/1746-1340-13-19
|
oa_comm
|
==== Front
CytojournalCytoJournal1742-6413BioMed Central London 1742-6413-2-161617429110.1186/1742-6413-2-16Case ReportDisseminated primary diffuse leptomeningeal gliomatosis: a case report with liquid based and conventional smear cytology Bilic Masha [email protected] Cynthia T [email protected] Zoran [email protected] Rana S [email protected] Department of Pathology & Laboratory Medicine, Medical University of South Carolina, Charleston, SC, USA2 Department of Radiology, Medical University of South Carolina, Charleston, SC, USA2005 20 9 2005 2 16 16 18 7 2005 20 9 2005 Copyright © 2005 Bilic et al; licensee BioMed Central Ltd.2005Bilic et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Primary diffuse leptomeningeal gliomatosis is a rare neoplasm confined to the meninges without evidence of primary tumor in the brain or spinal cord parenchyma. Cerebrospinal fluid diversion via ventriculoperitoneal shunt may be used as a therapeutic modality. Herein, we describe the first report of cytologic findings of a case of this neoplasm with shunt-related peritoneal metastasis.
Case presentation
A 19-year-old male presented with a 6-month history of severe headaches. He had bilateral papilledema on physical exam. Cerebrospinal fluid examination was negative. Four months later a ventriculoperitoneal shunt was placed. Shortly thereafter, he was diagnosed with primary diffuse leptomeningeal gliomatosis based on the biopsy of an intradural extramedullary lesion adjacent to the lumbar spinal cord at a referral cancer center. The histology featured an infiltrating growth pattern of pleomorphic astrocytes with diffuse positivity for glial fibrillary acidic protein. A couple of months later he presented at our institution with ascites and an anterior peritoneal mass. Repeat cerebrospinal fluid cytology and fine needle aspiration of the mass confirmed disseminated gliomatosis. Cytologic characteristics included clusters of anaplastic cells of variable size, high nuclear to cytoplasm ratio and scant to moderate cytoplasm. Occasional single bizarre multinucleated cells were seen with eccentric "partial wreath-like" nuclei, clumped chromatin and prominent nucleoli. Patient expired 13 months after initial presentation.
Conclusion
Disseminated primary diffuse leptomeningeal gliomatosis should be considered in the differential diagnosis of chronic aseptic meningitis and in the presence of a peritoneal tumor in patients with ventriculoperitoneal shunts. Immunocytochemistry may be of diagnostic value.
cytologyprimary diffuse leptomeningeal gliomatosisventriculoperitoneal shuntperitoneal metastasis
==== Body
Introduction
Primary diffuse leptomeningeal gliomatosis (PDLG) is a rare uniformly fatal neoplastic condition. It is characterized by widespread infiltration of the meninges by tumor thought to arise from heterotopic glial cell nests, without evidence of primary tumor within the brain or spinal cord parenchyma [1]. Glial heterotopia is defined as nests or linear arrays of glioneuronal tissue in the meninges. While their pathogenesis remains debatable, such heterotopic nests have been noted in the subarachnoid space in about 1% of unselected autopsies [2]. The incidence is substantially higher in patients with congenital anomalies, particularly those of the central nervous system [2,3]. Neoplastic transformation of extramedullary heterotopic glial tissue is a rare event, usually diagnosed at autopsy. One of the earliest reports of this entity was in 1936 by Bailey [4]. Since then, fewer than two dozen cases overall, and fewer than ten cases predominantly affecting the meninges around the spinal cord have been reported in the English language literature [1].
Patients with PDLG most often present with symptoms and signs of intracranial hypertension, such as headache and papilledema. Examination of the cerebrospinal fluid (CSF) characteristically shows raised opening pressure, elevated protein concentration and lymphocyte pleocytosis. Diagnosis is often delayed mainly because of the nonspecific nature of the symptoms and CSF findings, such that empirical antituberculous therapy is frequently tried in patients with PDLG [1,5]. There appears to be no cure, however, variable clinical and radiologic remissions lasting up to 4 years from presentation have been achieved utilizing aggressive chemotherapeutic and radiation regimens [1,6,7]. Cytologic examination of CSF in PDLG is most often negative. One series reported only 1 of 8 cases of PDLG having "atypical cells" on CSF cytology, without further elaboration of cytologic findings [8]. This is in contrast to a positive antemortem CSF cytologic diagnosis in 8 out of 12 patients with secondary meningeal spread from intracranial malignant glioma [9].
To reduce the intracranial pressure, some patients undergo a CSF diversion procedure in the form of ventriculo-peritoneal (VP) or alternate route shunting [5,6]. Peritoneal tumor dissemination by way of VP shunt is a known complication in patients with different primary central nervous system (CNS) tumors, most often germinomas [10,11]. According to our review of the literature, this complication has not been previously reported in PDLG.
Herein we describe the cytologic features of CSF and fine needle aspiration (FNA) of PDLG with VP shunt-related peritoneal spread.
Case Report
A 19 year old college freshman presented, in January of 2004, with a 6 month history of severe recurring global headaches. He had no past medical or travel history. Physical examination was notable for bilateral papilledema. His workup included magnetic resonance imaging (MRI) of the brain and intracranial vessels, which initially did not reveal any abnormalities. Initial examination of the CSF revealed an opening pressure of 50 cm H2O (normal: 10–20 cm), protein of 778 mg/dL (normal: 15–45 mg/dL), glucose of 44 mg/dL (normal: >45 mg/dL), an absolute WBC count of 24/mm3 with 55% lymphocytes and 45% macrophages. A full microbiology workup was negative for organisms. The CSF was negative by polymerase chain reaction (PCR) for enteroviruses and West Nile virus. Flow cytometry showed no evidence of lymphoma or leukemia. Cytologic examination of the CSF was negative for malignant cells, showing predominantly reactive monocytes. The patient was initially managed symptomatically with analgesics and diuretics. By May of 2004, he developed cranial nerve palsies necessitating placement of a VP shunt. The patient sought a medical second opinion at another institution, where a diagnostic biopsy was obtained of an intradural extramedullary lesion at T12/cauda equina via vertebral laminectomy.
The biopsy showed an infiltrating growth pattern of pleomorphic astrocytes. The nuclei were hyperchromatic with irregular borders. Blood vessels were numerous. Dense, wavy collagen bands were noted throughout the tumor, and seemed particularly apparent in the vicinity of blood vessels (Figure 1A). Necrosis and psammoma bodies were absent. The neoplastic cells had variable amounts of fibrillary cytoplasm which was diffusely positive for glial fibrillary acidic protein (GFAP) (DakoCytomation, Carpinteria, CA, 1:3000 dil; Figure 1B). A diagnosis of glial neoplasm consistent with leptomeningeal gliomatosis was made. Subsequently, the patient returned for care at our institution. At this time, MRI showed peripheral contrast enhancement of the meninges surrounding the entire spinal cord (Figure 2A) as well as prominent thickening and enhancement of the cauda equina, consistent with diffuse leptomeningeal disease. Brain MRI study showed only subtle leptomeningeal and subarachnoid spread, without evidence of a primary tumor. He received aggressive craniospinal axis radiation and temozolomide chemotherapy which resulted in clinical improvement for a period of about two months. By August of 2004 the patient developed bilateral lower extremity paraplegia and recurrent seizures. Five months later, he developed massive ascites and an anterior peritoneal mass (Figure 2B). A computed tomography (CT)-guided FNA of the peritoneal mass was performed. The FNA specimen was processed as air-dried Diff-Quik (DQ) stained and alcohol-fixed Papanicolaou (Pap) stained direct smears, one ThinPrep (TP) slide and a cell block. A diagnosis of VP shunt-related peritoneal spread of PDLG was made.
Figure 1 A. Biopsy of an intradural extramedullary lesion at T12/cauda equina, showing highly pleomorphic cells with a hint of fibrillary cytoplasm growing in an infiltrating fashion (H&E, ×100). B. Diffuse cytoplasmic staining of tumor cells with antibody for GFAP (GFAP, ×200).
Figure 2 A. Postcontrast sagittal T1-weighted MR image of the cervical spine shows diffuse leptomeningeal enhancement along the surface of the spinal cord. B. Axial contrast-enhanced CT image through the abdomen demonstrates extensive ascites and an enhancing midline mass (arrows) adjacent to the shunt catheter (arrowhead).
DQ-stained smears showed pleomorphic malignant cells in poorly formed, somewhat cohesive clusters with haphazard cellular arrangement. The nuclei were hyperchromatic with great variation in size. Cytoplasmic borders were indistinct (Figure 3). The TP showed scattered single bizarre, very large multinucleated neoplastic cells with "partial wreath-like" or "horse shoe" nuclear configuration, coarse chromatin clumping, prominent nucleoli, and fibrillary cytoplasmic tails (Figure 4A). Clusters or nests of smaller, mononuclear neoplastic cells were also seen featuring high nuclear to cytoplasm ratio, scant homogeneous cytoplasm, occasional nuclear clefting and more even chromatin distribution (Figure 4B). Scattered cytoplasmic GFAP positivity was present in the cell block material from ascites collected 3 days prior to the FNA. Core biopsy of the peritoneal mass was confirmatory.
Figure 3 CT-guided FNA of anterior peritoneal mass, showing a somewhat cohesive large cluster of highly pleomorphic cells (Diff-Quik, ×200).
Figure 4 Cytomorpholgy of primary diffuse leptomeningeal gliomatosis involving the peritoneum. A. Single bizzare multinucleated neoplastic cells display "partial wreath-like" nuclear configuration, coarse chromatin clumping and multiple nucleoli (TP, Pap stain, ×1000). B. Clusters of smaller tumor cells showing variation in size, high nuclear to cytoplasm ratio, scant homogeneous cytoplasm, with even chromatin distribution and occasional clefting of the nuclear membrane (TP, Pap stain, ×1000).
A repeat CSF specimen was processed as two Cytospin slides, one stained with DQ and one with Pap. It showed a cytomorphology similar to that of the peritoneal mass FNA (Figures 5A and 5B).
Figure 5 CSF cytology. A. Nests of anaplastic tumor cells are seen with round to oval eccentric nuclei, coarse chromatin, prominent nucleoli and moderate homogeneous cytoplasm (Diff-Quik, ×1000). B. Single giant malignant cells with similar morphology are also seen (Diff-Quik, ×1000).
The patient continued to deteriorate clinically. In addition to the non-improving neurologic status, he developed bowel obstruction which was unresponsive to palliative pelvic radiation and peritoneal infusion of phosphorus (P32) radioisotope. He expired in early February 2005, 13 months after initial presentation. An autopsy was declined.
Discussion
Primary diffuse leptomeningeal gliomatosis (PDLG) is a rare neoplastic disorder limited to the meninges, in the absence of primary tumor within the brain or spinal cord parenchyma [1]. The diagnosis is most often made at autopsy, mainly because of lack of specific clinical, radiologic and laboratory diagnostic criteria, and the progressive natural history of the disease [1,5,8,10]. As such, appropriate management of affected patients is often delayed [1,5]. Despite rare encouraging reports of clinical improvement after aggressive radiation and chemotherapy [6,7], the disease is fatal in the majority of cases [1,5,6,8]. The survival depends, in part upon the World Health Organization (WHO) differentiation grade of tumor, complicating lesions (e.g. infarcts), and site of tumor [1].
The case presented in this report is unique in several respects. It describes an exceedingly uncommon disorder, PDLG, with peritoneal seeding as a result of symptomatic treatment (VP shunt). While this complication is known to occur in several different types of primary CNS tumors [10,11], such a clinical course has not been previously described in PDLG. This patient's tumor appeared poorly differentiated, befitting WHO grade III or IV. Such high tumor grade was likely a key factor determining aggressive tumor behavior, and also the relative ease with which it was recognized in various cytologic specimen preparations described above. A confirmatory piece of evidence in support of the final diagnosis was demonstration of GFAP-staining cells in ascitic fluid and core biopsy material of the peritoneal mass.
Immunocytochemistry has been advocated as a useful adjunct test in addition to routine CSF cytology in differentiating chronic aseptic meningitis from leptomeningeal carcinomatosis or gliomatosis especially in clinically suggestive cytologically negative cases [12]. It seems far fetched to suggest an evidence-based approach of indiscriminate GFAP immunocytochemistry in addition to the routine cytologic CSF examination. Nevertheless, the case presented here serves to remind us of the need to remain vigilant in complete review of patient information, forthcoming in communication with clinicians and other specialists, and diligent in formulation of differential diagnosis in each case. Only such an approach will offer patients standard-of-care treatments, and allow for improvements in those standards.
==== Refs
Baborie A Dunn EM Bridges LR Bamford JM Primary diffuse leptomeningeal gliomatosis predominantly affecting the spinal cord: case report and review of the literature J Neurol Neurosurg Psychiatry 2001 70 256 258 11160482 10.1136/jnnp.70.2.256
Cooper IS Kernohan JW Heterotopic glial nests in the subarachnoid space: histopathologic characteristics, mode of origin and relation to meningeal gliomas J Neuropathol Exp Neurol 1951 10 16 21 14804124
Hirano S Houdou S Hasegawa M Kamei A Takashima S Clinicopathologic studies on leptomeningeal glioneuronal heterotopia in congenital anomalies Pediatr Neurol 1992 8 441 444 1476573 10.1016/0887-8994(92)90006-K
Bailey OT Relation of glioma of the leptomeninges to neuroglia nests; report of a case of astrocytoma of the leptomeninges Arch Path 1936 21 584 600
Rees JH Balakas N Agathonikou A Hain SF Giovanonni G Panayiotopoulos CP Luxsuwong M Revesz T Primary diffuse leptomeningeal gliomatosis simulating tuberculous meningitis J Neurol Neurosurg Psychiatry 2001 70 120 122 11118261 10.1136/jnnp.70.1.120
Beauchesne P Pialat J Duthel R Barral FG Clavreul G Schmitt T Laurent B Aggressive treatment with complete remission in primary diffuse leptomeningeal gliomatosis. A case report J Neur-Oncol 1998 37 161 167 10.1023/A:1005888319228
Paulino AC Chinnamma T Slomiany DJ Suarez CR Diffuse malignant leptomeningeal gliomatosis in a child: a case report and review of the literature Am J Clin Oncol 1999 22 243 246 10362329 10.1097/00000421-199906000-00006
Dietrich PY Aapro MS Rieder A Pizzolato GP Primary diffuse leptomeningeal gliomatosis (PDLG): a neoplastic cause of chronic meningitis J Neur-Oncol 1993 15 275 283 10.1007/BF01050075
Yung WA Horten BC Shapiro WR Meningeal gliomatosis: a review of 12 cases Ann Neurol 1980 8 605 608 6260012 10.1002/ana.410080610
Rickert CH Reznik M Lenelle J Rinaldi P Shunt-related abdominal metastasis of cerebral teratocarcinoma: report of an unusual case and review of the literature Neurosurgery 1998 42 1378 1382 9632200 10.1097/00006123-199806000-00118
Kumar R Sahay S Gaur B Singh V Ascites in ventriculoperitoneal shunt Indian Journal of Pediatrics 2003 70 859 864 14703222
Thomas JE Falls E Velasco ME Zaher A Diagnostic value of immunocytochemistry in leptomeningeal tumor dissemination. A report of 2 cases Arch Pathol Lab Med 2000 124 759 761 10782164
|
16174291
|
PMC1239923
|
CC BY
|
2021-01-04 16:36:22
|
no
|
Cytojournal. 2005 Sep 20; 2:16
|
utf-8
|
Cytojournal
| 2,005 |
10.1186/1742-6413-2-16
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 1618779410.1371/journal.pbio.0030343Research ArticleBioengineeringBioinformatics/Computational BiologyBiophysicsEvolutionSystems BiologyBiochemistrySaccharomycesEubacteriaDynamic Properties of Network Motifs Contribute to Biological Network Organization Dynamic MotifsPrill Robert J
1
Iglesias Pablo A
1
2
Levchenko Andre [email protected]
1
1Department of Biomedical Engineering, Johns Hopkins University, Baltimore, Maryland, United States of America,2Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, Maryland, United States of AmericaLander Arthur Academic EditorUniversity of California, IrvineUnited States of America11 2005 4 10 2005 4 10 2005 3 11 e34322 3 2005 4 8 2005 Copyright: © 2005 Prill et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Charting the Interplay between Structure and Dynamics in Complex Networks
Motifs, control, and stability
Biological networks, such as those describing gene regulation, signal transduction, and neural synapses, are representations of large-scale dynamic systems. Discovery of organizing principles of biological networks can be enhanced by embracing the notion that there is a deep interplay between network structure and system dynamics. Recently, many structural characteristics of these non-random networks have been identified, but dynamical implications of the features have not been explored comprehensively. We demonstrate by exhaustive computational analysis that a dynamical property—stability or robustness to small perturbations—is highly correlated with the relative abundance of small subnetworks (network motifs) in several previously determined biological networks. We propose that robust dynamical stability is an influential property that can determine the non-random structure of biological networks.
The authors model how network motifs respond to small-scale perturbations and find a strong correlation between motif stability and abundance in a network, suggesting that dynamic properties of network motifs may play a role in overall network structure.
==== Body
Introduction
Life is a dynamical process. As the intricate connectedness of biological systems is revealed, it is essential to keep in mind that network maps are graphical representations of dynamic systems. A network with fixed structure is dynamic in the sense that the nodes take on values corresponding to activities that change in time. In a specific context, these activities may represent the concentration of a molecule, phosphorylation state of an enzyme, depolarization state of a neuron, etc. For example, the neural connection map of the nematode, Caenorhabditis elegans, is fixed in the adult worm, and invariant from individual to individual [1]. Although the network structure is static, the behavior of the neural network is dynamic. At the lowest level, the dynamic nature of the system is exhibited as depolarizations of individual neurons. At an intermediate level of organization, each neuron potentially influences the behavior of its nearest neighbors via synaptic connections.
Example biological networks, such as those analyzed here, are artificially separated from the highly integrated and complex whole, which includes interconnected metabolic, signal transduction, transcriptional, cytoskeletal, and other types of interlocked networks and pathways. Even with this limitation, consideration of the structure of isolated networks has profoundly influenced contemporary biology, which is increasingly focused on systems-level concepts. The topological structures of the networks studied here: the transcriptional regulatory networks of Escherichia coli, Saccharomyces cerevisiae, the developmental transcriptional network of Drosophila melanogaster, the signal transduction knowledge environment (STKE) network, and the neural connection map of C. elegans, have been analyzed from various perspectives, leading to provocative ideas relating structure to function. These and other natural networks have non-Poisson degree distribution (often power-law) [2]. In addition, recent studies have indicated that certain patterns of local connectivity (network motifs) are statistically over- and underrepresented in various networks, including those regulating development and function of living organisms [3,4]. The term “network motif” refers to a directed subgraph, consisting of a few nodes, that is embedded in a larger directed graph. Some motifs, such as the three-node feed-forward loop, may perform specific regulatory functions [5–7], although in general it is not presumed that motif instances are necessarily functional modules. Used as a structural analysis technique, enumeration of all subgraphs consisting of three or four nodes summarizes the local connectivity patterns that compose a complex network [8–10].
Presently, it is not clear what determines the particular frequencies of all possible network motifs in a specific network. At least two alternative explanations can be offered. One can hypothesize that certain constraints on development of a network as a whole determine which motifs become abundant. Conversely, some network motifs may possess properties important enough to become overrepresented and thereby drive the network evolution and its ultimate structure. The latter explanation is consistent with the frequently made assumption that the function of a small biological network or pathway is largely determined by the connections among the constituting genes, proteins, metabolites, or cells [11–16]. Many of the genetic and biochemical systems extensively studied to date as functional units, such as the Lac and Che systems in E. coli (responsible for lactose utilization and chemotaxis, respectively) are of this sort [17–19]. In these systems, connectivity is typically translated into a dynamic property, which, in turn, takes on a functional meaning. For instance, a positive feedback loop in the Lac system (a structural characteristic) is used to implement a biochemical switch (dynamic behavior) that acts as a primitive memory mechanism (functional role). We hypothesize that the dynamic behavior displayed by a network motif is an important criterion in determination of functional significance of the motif and, potentially, its abundance in biological networks. Further, we propose that it is computationally tractable to determine certain dynamic properties for every possible network motif. This strategy can explain why certain network motifs are overrepresented in real biological networks while others are not.
A comprehensive analysis of the dynamics of networks, large or small, is considerably more complicated than the corresponding analysis of their structure. For instance, structures of all possible three- or four-node network motifs can be generated and enumerated easily. However, their dynamics may not be completely determined owing to unknown and potentially complex functional dependencies between nodes, the lack of knowledge of parameters defining specific instances of motifs in real networks, and “unmodeled” interactions that may be absent from the network representation yet relevant to dynamics. In addition, motifs with the same topology may give rise to different dynamic behaviors, as was recently demonstrated experimentally using small synthetic genetic circuits [20]. Clearly, there is not a direct one-to-one relationship between network structure and possibly complex system dynamics. However, one can try to address these fundamental challenges by taking advantage of the widely acknowledged notion that biological systems perform various functions robustly, i.e., under wide ranges of their parameters [21]. Thus, instead of considering properties of a particular instance of a motif, we analyze generic properties that arise from the topology of the network motif.
Although the particular parametric values can confer unique properties onto individual instances of motifs, we argue that all instances of a particular motif display characteristics that can be studied comprehensively. Here we present the analysis of a specific robust property that can be displayed by network motifs: robust stability to small-scale perturbations of the activities of the biological entities. Small perturbations include intrinsic stochastic fluctuations (noise) and transient up- or down-regulation of activity. “Small” implies that a linear approximation of the potentially nonlinear relations is still valid (see Materials and Methods).
Intuitively, one expects that for complex biological networks to function robustly, it is necessary that they display stability and resist both noise and stressful small-scale perturbations. These basic homeostatic properties, however, are not inherent to any biological or biochemical system, as small stimuli can, in principle, trigger large-scale sustained responses, especially if feedback interactions between members of a biological network are involved. For example, relatively short stimulations by tumor necrosis factor α (5 min or less) can trigger sustained (60 min) activation of the NF-κB pathway leading to expression of a battery of genes [22]. Here we show that stability to small perturbations displayed robustly by network motifs can be characterized comprehensively. In addition, and more significantly, we show that this property can be a driving force defining the structure of several biological networks.
Results
The dynamic behavior of a circuit is determined by the direction, sign, and strength of the connections. To develop some intuition, consider the case of a two-node feedback loop (Figure 1). The sign and strength of a connection from node j to node i is labeled aij
. Additionally, we assume that a self-interaction, denoted aii, represents the commonly observed mechanisms of constitutive degradation or inactivation of the biological entities. Without proper mathematical intuition, one could incorrectly assume that negative feedback ensures stability and positive feedback destroys stability. As we demonstrate next, the dynamics of this simple circuit are more complicated.
Figure 1 Dynamic Behaviors of a Two-Node Feedback Loop in Response to a Small Perturbation from Steady-State
(Top) Schematic of a simple feedback circuit with the nodes and edges labeled.
(Middle) We illustrate the dependence of stability on the values of a
12 and a
21 as they are varied from −1 to 1, with constant self-degradation terms a
11 = −0.3 and a
22 = −0.7. The system can be stable (green), oscillatory (blue), or unstable (red). These regions are determined by gain k = a
12
a
21. Shown in white are contours of constant gain k. The quadrants labeled (+) correspond to positive feedback (a
12 and a
11 have the same sign), while the quadrants labeled (−) correspond to negative feedback (a
11 and a
21 have opposite signs).
(Bottom) The stability regions vary as the values of self-degradation terms a
11 and a
12 change. The more stable the open-loop nodes (more negative a
11 and a
22), the greater the regions of closed-loop stability. However, if a
11 and a
22 are close in sign and magnitude, the size of the oscillatory regions increases.
We analyze the response of this circuit to a small perturbation from a steady-state under different assumptions on the parameters (Figure 1). For a particular set of constant values for the self-degradation terms (a
11
, a
22
), the system can be stable (green), unstable (red), or oscillatory (blue), depending on the sign and strength of the feedback gain (k = a
12
a
21
). However, these regions occupy different relative areas of parameter space depending on the particular values of the self-degradation terms. The more stable the individual nodes (more negative a
11
, a
22), the greater the regions of closed-loop stability. Consequently, positive feedback produces stability if constitutive degradation dominates the feedback gain. Both positive and negative feedback can produce oscillations in general (only negative feedback can produce oscillations in the two-node case). Finally, upstream input nodes and downstream output nodes can be added to the system without altering the stability, provided that the number or size of feedback loops is unchanged.
Similar analysis can be extended to all possible three- or four-node network motifs by defining a metric, the structural stability score (SSS), as the probability that the dynamical system corresponding to a given motif relaxes monotonically to steady-state following a small perturbation. An SSS value of 1 indicates that non-oscillatory relaxation to a steady-state (henceforth simply termed stability) is guaranteed by connectivity and does not depend on the specific parameter values that define the functional interactions, thus making the system structurally stable. SSS scores less than unity indicate the extent to which parameter values influence stability: the lower the score, the more precise the balance of connection signs and strengths necessary to achieve stability. For example, the stability of a small transcriptional network with a high SSS would be robust to fluctuations of protein expression levels, seemingly the most important source of variation of gene expression across bacterial cell populations [23]. This compact description of the relationship between network structure and system dynamics enables a comprehensive characterization of the stability of small circuits in real networks.
Determination of the SSS values (fully described in Materials and Methods) reveals that all topologically distinct three-node (13 total) and four-node (199 total) network motifs partitioned into three classes that display distinct stability regimes. In particular, the first class (I) comprised all the robustly stable motifs (SSS = 1) devoid of feedback loops, i.e., motifs that are directed acyclic graphs. The second class (II) consisted of moderately stable circuits (SSS ≈ 0.4) containing a single two-node feedback loop. These motifs can be guaranteed not to be unstable (but may have damped oscillations) provided that the feedback present is negative (see Figure 1). The third class (III) contained motifs with much lower SSS values (< 0.2), which were a mixture of more complicated circuits: multiple two-node loops, three- and four-node loops, nested multi-node loops, etc. Their stability cannot be guaranteed by specifying the sign of the feedback loops present. Generally, as the number or length of loops increases, the SSS decreases. Thus, highly connected motifs are generally the least stable.
Next, we contrasted the abundance of motifs in several real biological networks with the corresponding motif SSS values (Figure 2). Remarkably, in all networks analyzed, the stability scores showed excellent correlation with the motif abundance. The networks are mostly composed of the motifs in the highest stability class, suggesting that the overall network behavior is stable to small perturbations. This finding further suggests that the overall network structure may be driven by the requirement of stability to small perturbations, such as noise.
Figure 2 Abundance of a Motif Is Correlated with Its SSS
The number of instances of all three-node (left) and four-node (right) motifs in biological networks (log scale), and SSS (black dashed line). The motifs are sorted on the x-axis from high to low SSS.
Above, we chose a conservative definition of stability, classifying any system with oscillatory behavior as unstable, even if the oscillations are damped. However, our results still hold if we adopt the more traditional, and less conservative, definition of stability in which damped oscillations are classified as stable dynamics, irrespective of the potential presence of oscillation (see Figures S1 and S2, damped oscillations in the SSS metric). Thus, our conclusions do not have a strong dependence on the particular definition of stability used.
We also noted that, for the same SSS values, the motif abundance was correlated with the number of edges in the motif, as can be seen from the particular ordering of the three-node motifs used (Figure 2, left panel). Since all the networks considered here are very sparse (compared to fully connected graphs of the same number of nodes), it stands to reason that motifs containing a large number of edges might be encountered less frequently than motifs with a low number of edges, independently of the stability properties. In subsequent analyses we focused on a comparison of relative abundances of network motifs within motif groups having the same number of edges (a “density group”).
Is the non-random character of network organization driven, at least to some extent, by the structural stability of network motifs? If this is the case, one would expect that motifs of relatively higher stability would be overrepresented compared to their relatively less stable counterparts when compared to random networks of the same size. To test this hypothesis, we generated 100 Erdös-Renyi (ER)–type random graphs with the same number of nodes and edges as each of the real networks. Lacking any organizing principle, the distribution of motifs in this type of random graph is determined by the density of edges [24]. Although there is some controversy as to whether the ER-type or scale-free random network model is a more natural representation of a network devoid of organization, we chose the former because it is the historically accepted model of a network produced by a random process of linking nodes. Also, it agrees with our intuition that power-law degree distribution is itself a level of organization of a network. However, we also performed simulations using randomized networks (which preserve degree distribution) as the null model (Figures S3 and S4). The particular choice of null model does not affect the main conclusions of the paper. We compared motifs in real and random networks by using Z scores as a metric of statistical over- or underrepresentation, as previously proposed [3,25]. A positive Z score indicates that a motif is overrepresented in the real network, whereas a negative score indicates underrepresentation in the real network compared to random graphs. The Z score profiles were normalized to unit vectors to enable comparisons of scores across different networks. Normalized Z score profiles are represented as bar graphs in Figure 3 (left panel) and Figure 4.
Figure 3 Distribution of Normalized Z Scores of Three-Node Network Motifs (Left), and Z Scores Classified by Stability within Specific Density Groups (Right)
(Top panel, left) All 13 network motifs are sorted on the x-axis first, according to increasing number of edges (solid red line). For a given number of edges, they are then sorted from high to low SSS (black bars).
(Bottom five panels, left) Normalized Z scores (green bars) for all 13 motifs of the transcriptional networks of E. coli and S. cerevisiae, STKE network, D. melanogaster developmental transcriptional network, and C. elegans neuronal network, shown with outlines of the SSS from the top panel (dotted black outline provided as a guide to the eye). Each vertical red line indicates a change in the number of motif edges, indicating boundaries of density groups.
(Right) Z scores of network motifs in the three-edge and four-edge density groups are plotted in columns labeled I, II, and III, corresponding to the three stability classes, SSS = 1, SSS ≈ 0.4, and SSS < 0.2, respectively. Note that the four-edge density group does not contain any motifs with SSS = 1 since four edges among three nodes dictates at least one feedback loop.
Figure 4 Distribution of Normalized Z Scores of Four-Node Network Motifs
(Top panel) All 199 motifs are sorted on the x-axis first, according to increasing number of edges (solid red line). For a given number of edges, they are then sorted from high to low SSS (black bars).
(Bottom five panels) Normalized Z scores (green bars) for all 199 network motifs of the indicated biological networks, shown with outlines of the SSS from the top panel (dotted black outline provided as a guide to the eye). Each vertical red line indicates a change in the number of motif edges, indicating boundaries of density groups. The composition of the four-node density groups is specified in Protocol S1.
The top panels of Figures 3 and 4 illustrate the computational model of stability expressed as the distribution of SSS values, while the lower panels display the data for each real network. Network motifs are ordered, left to right, from low edge-density to high edge-density. Then, motifs with a given number of edges are ordered, left to right, from high to low SSS. For example, in Figure 3, motifs 1, 2, and 3 have two links and SSS = 1. They comprise a “density group” consisting of all the motifs in which three nodes are connected by two links. The next density group is comprised of motifs 4, 5, 7, and 8, which are the motifs in which three nodes are connected by three links. Within this density group, the SSS are diverse. Motif 7 has SSS = 1, motifs 4 and 5 have SSS ≈ 0.4, and motif 8 has SSS ≈ 0.2. All the real networks have some overrepresentation of motif 7, which has a higher SSS than those of motifs 4, 5, and 8. Next, some networks have a lesser overrepresentation for motifs 4 and 5, members of the moderately stable class. The least structurally stable circuit, motif 8, is not overrepresented in any real network. Similarly, the four-edge (three-node) density group also presents an opportunity to investigate the stability trend since more than one stability class is represented. The four-edge density group consists of two moderately stable motifs (9 and 10) and two minimally stable motifs (6 and 11). Z scores classified by stability are plotted in Figure 3 (right panel). Overrepresentation of structurally stable motifs (within each density group) argues for importance of motif stability as a dynamic property affecting network organization.
Four-node network motifs are combinatoric elaborations of three-node motifs [26]. The four-node significance profiles (Figure 4) capture a richer representation of the local connectivity patterns than the three-node profiles. At this resolution both similarities and differences between the networks are immediately apparent, yet the general stability trend is the same as in the three-node analysis. The networks differ in precisely which motifs are overrepresented, but the dynamic properties of overrepresented motifs are conserved across all networks we analyzed. As with the three-node profiles, the network motifs with the highest Z scores also have higher SSS than the other motifs with the same number of edges.
On average, the most structurally stable motifs have higher Z scores than those with lower SSS within each density group (Figure 5). For example, the Drosophila developmental transcription network (Figure 5D) has overrepresentation of some four, five, and six-edge network motifs. Box and whisker plots for each density group indicate that high-stability motifs (class I) have high Z scores, intermediate-stability motifs (class II) have intermediate Z scores, and low-stability motifs (class III) have low Z scores. The p-value attached to each plot was calculated using the Kruskal-Wallis test (one-way analysis of variance on ranks), which expresses the probability that the observed differences in Z scores between stability classes is expected by chance. In most cases, the difference in Z scores between stability classes is significant at the usual criterion of 95% confidence or better.
Figure 5 Z Scores of Four-Node Network Motifs, Classified by Stability
Density groups consisting of four, five, and six-edge network motifs contain at least one motif in each stability class. Box and whisker plots indicate a difference in the average Z score between stability classes (I, II, and III correspond to SSS = 1, SSS ≈ 0.4, and SSS < 0.2, respectively). The networks examined here are the transcriptional networks of (A) E. coli, (B) S. cerevisiae, (C) STKE network, (D) D. melanogaster developmental transcriptional network, and (E) C. elegans neuronal network. A box and whisker plot is interpreted as follows: The box indicates the inner quartile range. The median Z score is indicated by a horizontal red line within the box. The whiskers extend to cover the upper and lower quartiles up to a distance of 1.5 times the inner quartile range. Red crosses indicate extreme scores (beyond the whisker length). P-values attached to each plot express the probability that the observed difference in Z scores between stability classes is expected by chance.
In addition to statistically significant differences in average Z scores among stability classes, there is similarity in stability properties among highly overrepresented network motifs. Table 1 demonstrates the preference for structural stability among the highly overrepresented network motifs in density groups consisting of four, five, and six edges (four nodes). We selected the most significant network motifs in each density group by picking Z scores in excess of one standard deviation above the mean Z score for the group. These motifs dominate the non-random organization of the network within their density group. The stability classification (I, II, or III) is indicated in parentheses next to the Z score. In all cases, the highly significant network motifs belong to classes I or II. Finally, Table 1 reports the size of each stability class, which indicates, to some extent, avoidance of low-stability motifs. For example, the six-edge density group is comprised of 47 network motifs: one class I, nine class II, and 37 class III. The Drosophila developmental network contains a highly overrepresented class I motif, which has the highest Z score in the density group. The remaining high Z scores correspond to class II motifs. None of the 37 low-stability motifs have high Z scores. Overall, high Z scores correspond to high SSS, but an SSS equal to unity does not always result in a high Z score, implying that structural stability may be necessary, but not sufficient, for network motif overrepresentation.
Table 1 High Z Scores of Four-Node Network Motifs
In the preceding investigation, we extended a structural analysis technique (decomposition of a network into subgraphs) to a dynamical analysis. Our characterization of motif dynamics implicitly assumed that subsystems consisting of a few nodes could behave relatively autonomously despite being embedded in a larger network. Conceptually, the assumption is valid when the activity of a node does not propagate a long distance through the network, creating the situation in which small subgraphs are functionally relevant. This situation can arise when (i) only a small fraction of the nodes are activated, or (ii) only a small fraction of the potential regulatory interactions are activated. Under these conditions, the “active network” is a fragmented version of the full potential network. Note that the preceding dynamical analysis was appropriate for only a small perturbation of the activities of the nodes. Now we utilize the concept of a small perturbation to demonstrate that the active network can be a fragmented version of the full potential network, with true functional regulatory units consisting of three or four nodes.
Regulation of gene expression is dependent on the particular demands of a cell with respect to its environment. For example, many transcription factors are active only in the presence of additional signaling molecules. cAMP receptor protein, the most connected node in the bacterial transcription regulation network, is active only when bound to co-regulator cAMP, which, in turn, is present under only special circumstances, such as absence of glucose in the medium. In the absence of a cAMP signal, cAMP receptor protein is inactive, and its links are inactive. We define a context-dependent network to be the subset of the nodes and links that are activated in a certain context, such as glucose deprivation of a bacterial culture. With this principle in mind, we used high-throughput gene expression datasets to infer the context-dependent gene regulation networks, with specific contexts defined by mild environmental stress perturbations. These contexts are relevant to the preceding dynamical analysis.
We considered five different context-dependent gene regulation networks based on the genomic expression pattern of the yeast S. cerevisiae subjected to relatively small environmental perturbations including mild heat shock, osmotic shock, hydrogen peroxide treatment, etc. [27]. In the context of a specific environmental shock, we mapped the activated nodes (greater than 2-fold expression change) to their locations in the network. We retained all inbound links to regulated nodes, irrespective of whether or not altered mRNA expression of the upstream transcription factor was observed. We disregarded inactive links (the links that terminate on inactive nodes). Then we obtained the distribution of the size of active functional regulatory motifs by a recursive algorithm that “walked backward” through the network to discover all the upstream regulators that affect each regulated node, similar to a method previously employed [28]. In five different contexts, the mean number of nodes in a regulatory structure ranges from 3 to 3.5 (Table 2). Furthermore, all of the active regulatory motifs (three nodes, four nodes, and larger) had SSS = 1 (see Figure S5 for examples of context-dependent regulatory motifs.) Thus, in principle, the non-random character of the yeast transcriptional network could have resulted from selection acting on small network motifs that are functionally relevant in specific environmental contexts, particularly with regard to robustness to small-scale perturbations. Given appropriate small perturbation data, similar approaches could be employed to analyze the fragmentation of other complex biological networks in various functional contexts.
Table 2 Size of Active Regulatory Motifs in the Yeast Gene Regulation Network, for Various Experimental Contexts
Discussion
Do common “driving forces” underlie the organization of biological networks? It seems fantastic to suggest that such forces could exist, considering that the biological entities involved are as diverse as genes, enzymes, and whole cells. Nevertheless, even functionally unrelated systems may have evolved under fundamental constraints. The analysis presented here suggests that the dynamic properties of small network motifs contribute to the structural organization of biological networks. In particular, robustness of small regulatory motifs to small perturbations is highly correlated with the non-random organization of these networks. Inspection of highly overrepresented motifs in the four-node significance profile (see Figure 4 and Table 1) and statistical analysis of Z scores within and between stability classes (Figure 5) demonstrates a correlation between overrepresentation of a network motif and its SSS.
There are two separate trends in the motif significance profiles (see Figures 3 and 4). The similarity among all the examined networks is the overrepresentation of stable motifs compared to the other motifs with the same number of edges (i.e., within a density group). A separate phenomenon is that the networks differ in precisely which density groups are overrepresented, as is clearly evident in Figure 4. Previously, multiple networks have been grouped into a few families displaying remarkable similarity in the three-node motif distribution profiles [25]. At the resolution of four-node motifs (see Figure 4), bacteria and yeast transcription networks have a preference for the same network motifs. It is likely that these networks, which perform gene regulation in single-celled organisms, have evolved under similar constraints and can be considered a true “family.” However, according to the four-node motif significance profile, STKE, Drosophila transcription, and C. elegans neuron networks do not seem to form a family (as mentioned in [25] based on a three-node analysis), since they independently have preferences for different four-node network motifs. Intuitively, it makes sense that these networks do not share the same global constraints since they are built from vastly different components and have vastly different functional purposes. Initially, we hypothesized that either (i) motifs are a consequence of global constraints, as suggested by motif families, or (ii) global structure is a consequence of motifs, as suggested by constraints on robust stability of subsystems. The four-node motif significance profile (see Figure 4) demonstrates evidence for both mechanisms. Again, although there are differences in precisely which motifs are enriched in various networks, the dynamic properties of the overrepresented motifs are conserved across all networks we analyzed. Compared to network motifs with the same number of edges, the more structurally stable motifs are overrepresented. We conclude that robust stability of subsystems contributes to the global structure of the large-scale biological networks studied here.
An evolutionary argument may help explain the overrepresentation of structurally stable motifs in real networks compared to random graphs. Evolutionary pressure may select for network innovations that are structurally stable because these configurations are robust to variations in the strength of the connections. A high SSS indicates that it is likely that randomly assigned connection strengths and signs will result in a stable equilibrium, while a low SSS indicates that stability is possible although it requires precisely weighted connection strengths. Easily parameterized network designs that are predisposed to dynamical stability can be advantageous considering the evolutionary mechanisms of random mutation and natural selection. Of course, stability to small perturbations is by no means the only functional constraint on network performance and structure. For instance, in the developmental transcriptional regulation network in Drosophila considered here, irreversible switching of transcriptional circuits involving feedback regulation is an important determinant of irreversibility of the developmental progress, which might lead to selection of relatively unstable network motifs with feedbacks. The C. elegans neuron network, which strays the furthest from structural stability in our analysis, may also have functional constraints leading to overrepresentation of oscillators and memory switches. Nevertheless, the correlation between network motif overrepresentation and the SSS suggests that stability of small functional circuits may be a basic constraint common to all networks, which along with other functional requirements can significantly bias the likelihood that a given motif is selected for. Thus, one would expect to find with higher probability that some (though not all) network motifs with a high SSS would be strongly overrepresented, whereas the probability of finding overrepresented motifs with low SSS would be relatively much smaller, as suggested by Table 1.
The relationship between structural stability and overrepresentation of motifs can be illustrated using the example of three-node loops. The three-node feed-forward loop (motif 7) is an acyclic topology. Like all feed-forward architectures, it has an SSS of 1, implying that its relative abundance in a network might be high, as is indeed the case. As suggested above, structural stability is usually necessary but not sufficient for the emergence and preservation of network motifs. In the case of the feed-forward loop, previous modeling has demonstrated that this motif can produce a wealth of computational functions [5], providing functional explanations for its overrepresentation. Moreover, although all eight possible sign combinations (describing whether interactions between nodes are positive or negative) have the same SSS, only two are abundant in the transcriptional networks of E. coli and S. cerevisiae [5], providing further evidence of a particular role for this motif in these networks. On the opposite side of the SSS spectrum, the corresponding three-node feedback loop (motif 8) has a low SSS score implying low abundance, which is the case. The low SSS suggests that any possible implementation of feedback loop will be more prone to instabilities, stochastic fluctuations, and oscillations (also proposed in [25], which has indeed been demonstrated experimentally for a synthetic feedback loop network, the “repressilator” [29]). Our results suggest that low structural stability of this network motif is prohibitive for its evolutionary selection, whatever its potential benefits might be.
An important caveat of our work is that we have assumed that the network operates in isolation of other components of the overall system that may influence stability. Clearly, these example networks are a significant simplification of a considerably more complicated reality of biological regulation. For example, in assigning SSS values in transcriptional regulation networks, protein–protein interactions and other modifications of transcription factors and their targets are ignored. In this case, it is likely that these unmodeled components represent dynamics with faster time scales, and that the transcriptional network represents the slower, dominant dynamic behavior. Moreover, it is reasonable to conjecture that a network with higher SSS would be more tolerant of perturbations in the faster scale components than one with a lower SSS value. However, in some cases, these unmodeled components may exert considerable influence on the system. For example, in a recently published model of the heat-shock response in E. coli, considerable control is achieved by feedback loops involving mRNA and interactions with chaperones and proteases [30], none of which are found in the description of the transcriptional motifs. Nevertheless, despite the simplification that is inherent in our analysis, it is all the more interesting and surprising that we find the correlation between the structural stability of network motifs and their occurrence to be so strong. It may suggest that external control, though clearly present and significant in many cases, such as the heat-shock response, may still not be of overriding importance in others. As we increase our knowledge of the real biological networks, as well as our ability to include interactions at different levels, the analysis described here will still be applicable to their fuller description and may result in further insights about network structure and evolution.
Our characterization of motif dynamics implicitly assumed that subsystems consisting of a few nodes could behave relatively autonomously despite being embedded in a larger network. We demonstrated how the yeast gene expression network is only partially activated by mild environmental perturbations (heat shock, osmotic shock, etc.), concluding that nodes and links in large-scale networks can be interpreted as potentially active, rather than “always on.” A context-dependent network, composed of the subset of the nodes and links that are activated at a given moment, can be a fragmented version of the full potential network, where network motifs are on the scale of three or four nodes. We demonstrated context-dependence in yeast gene regulation, but we expect that, in general, it is appropriate to assume that all elements in a network are not “always on.” The more fragmented a network in various functional contexts, the more relevant our analysis of dynamic properties of network motifs.
In summary, our results suggest that robust stability of network motifs is an important determinant of biological network structure. This conclusion favors the intuitively appealing claim that biological networks need to be resistant to small-scale perturbation, including noise, and that this resistance may be structurally embedded in the network organization. While our analysis shows a strong correlation between network motif abundance and structural stability, it leaves us to speculate as to why motifs with various numbers of edges are overrepresented in different networks. Since the bacteria and yeast networks regulate gene expression in relatively simple organisms, many features provided by the presence of less stable, feedback-containing motifs may not be necessary or could potentially be detrimental. In contrast, transcriptional networks of higher organisms or non-transcriptional regulatory networks may benefit from the increased occurrence of feedback-containing motifs and more complex functions potentially provided by them. Stability to small perturbations can be important for robust network performance. Thus, it is reasonable to expect that motif distribution among diverse networks represents a balance of abundant, stable motifs and relatively rare, potentially unstable motifs, allowing greater functional flexibility coupled with predominant dynamic stability.
Materials and Methods
Computational model of motif stability
Given complete knowledge of the functional dependencies of the nodes, a dynamical system corresponding to a particular motif consisting of n interconnected nodes (Figure 6, panel I) can be represented by a system of differential equations:
Figure 6 Scheme for Modeling the Local Stability of a Network Motif
(I) An example network motif (10) depicted with nodes and edges labeled. The sign and weight of an edge from node j to node i is denoted aij (solid line) consistent with the notation in the rest of the figure. Additionally, we assume that every node may have a self-interaction aii (dotted line).
(II) A motif is a graphical depiction of a dynamical system that can be modeled using (possibly nonlinear) ordinary differential equations, which are functions of the state of the nodes in the motif.
(III) Assuming that the motif can achieve steady-state dynamics, the stability of the equilibrium may be analyzed by making a linear approximation of the (possibly nonlinear) functions. Linearization involves computation of the Jacobian, a matrix of partial derivatives of the functional dependencies expressed in the equations with respect to the variables. The linearized system is evaluated at the equilibrium point(s), resulting in specific values of aij.
(IV) The general form of the Jacobian matrix corresponding to this motif contains zero entries wherever one node exerts no influence on another node and non-zero elements wherever it does. An instance of a Jacobian consistent with the corresponding motif is created by sampling the non-zero terms from some distribution. Eigenvalues of the Jacobian indicate the local stability properties of the equilibrium point. Sampling many instances of structurally equivalent Jacobians allows computation of the probability of stability of an equilibrium point.
where the variable xi represents the state of the ith node, fi represents the combined influence of all nodes having connections to the ith node, and is the rate of change of xi (Figure 6, panel II). Most frequently, we do not have enough information to construct specific functions fi, which may exhibit complicated nonlinear dependencies. However, if we restrict our focus to local stability, we can alleviate the lack of functional relationships by assuming that one or more steady-states (denoted x*) exist, and examine the system behavior in a small neighborhood of these steady-states. Under these conditions, a linear approximation of the dynamics can be used (Figure 6, panel III). This approximation is accomplished by considering the evolution of small deviations of the system from the steady-state. It involves the computation of the Jacobian, a matrix of partial derivatives of the functional dependencies expressed in the equations with respect to the variables, evaluated at the equilibrium of interest.
The linearized system in a small neighborhood (δx) of the steady-state is:
Although the precise values of the elements of the Jacobian matrix might not be known, it is clear that the matrix has zero-valued elements whenever one node exerts no influence on another node, and non-zero elements whenever it does. The sign of these elements depends on whether the influence is activating (positive) or inhibitory (negative). We note that the Jacobian can be reduced to the corresponding adjacency matrix by normalizing the entries to ones or zeros. Thus, the adjacency matrix is a particular example of a Jacobian consistent with the potentially nonlinear differential equations. Thus, in our analysis, the general form of the Jacobian J = {aij} corresponding to each motif was obtained by inspection of the adjacency matrix. We also assumed that the self-connections, reflected in the diagonal terms of the Jacobian, are always negative. This assumption represents the mechanisms of constitutive degradation or inactivation of the biological entities, including gene products, phosphorylation states of signaling molecules, or depolarization states of neurons.
We analyzed the stability properties of the 13 three-node and 199 four-node motifs using root locus analysis (see Protocol S2). This technique, developed to analyze engineering control systems, allows determination of the stability of the system as a parameter is allowed to vary. Root locus and graphical methods become unwieldy for motifs containing multiple feedbacks.
Rather than calculate an exact solution for all possible three- and four-node motifs, we employed a numerical method to estimate stability under a particular probability distribution for the aij terms. For each motif, we created 10,000 instances of the Jacobian matrix where the diagonal terms (self-interactions) were sampled from a uniform (−1,0) distribution, and the off-diagonal terms were sampled from a uniform (−1,1) distribution similar to the method described in [31]. These ranges were used since we consider linearized systems where only relative values of the aij terms are significant. Eigenvalues were determined for each random instance of the Jacobian (Figure 6, panel IV). We defined a metric, SSS, as the probability that the dynamical system corresponding to a given motif displays a non-oscillatory, damped response to a small perturbation from a steady-state. This is equivalent to the fraction of instances in which all eigenvalues are complex numbers with negative real part and zero imaginary part. For instance, SSS = 1 indicates that non-oscillatory relaxation to a steady-state (henceforth simply termed stability) is guaranteed by the motif connectivity and does not depend on the specific parameter values that define the interactions between the variables in the corresponding dynamical system, thus making the system structurally stable.
Abundance of network subgraphs
We used the Mfinder1.1 software program distributed by U. Alon's group (http://www.weizmann.ac.il/mcb/UriAlon/) to count the number of three- and four-node subgraphs in the real and random networks.
Motif Z scores and null model
Motif abundances in the real networks were compared to those in ER- type random graphs with the same number of nodes and edges as the corresponding real network. This is a different null model than previously described [3], which utilized “randomized” networks that preserved the in- and out-degree of each node. The randomized null model is appropriate for investigating local phenomena while preserving global degree distribution. We take a completely unbiased approach, using motif abundances to study both local and global network structure simultaneously. ER random graphs were generated by first placing n nodes on a canvas, then adding e directed edges. The statistical significance of the motif abundance was evaluated by calculation of the Z score as utilized previously [25],
where Nreali is the abundance of the ith motif, and <Nrandi> and std(Nrandi) are the mean and standard deviation of its abundance in the corresponding 100 random graphs with the same number of nodes and edges as the real network. We normalized the vector of Z scores to unit length:
Data sources
The E. coli transcription network, STKE network, Drosophila transcription network, and C. elegans neuron network described in [3,25] were obtained from U. Alon. The S. cerevisae transcription network was obtained from a high-throughput dataset described in [32]. We utilized the p < 0.001 binding interactions observed under YPD culture conditions.
Supporting Information
Figure S1 Distribution of Normalized Z Scores of Three-Node Network Motifs Using a More Expansive Definition of SSS That Includes Damped Oscillations
The SSS is represented as a stacked bar graph (top panel) in which the black portion of the bars represents the component of SSS due to monotonic decay, and the grey portion of the bars represents the component of SSS due to damped oscillations. The ordering of three-node motifs by the more expansive definition (traditional in the field of control systems engineering) produces the identical ordering of three-node motifs as conservative definitions of SSS that exclude oscillations (compare to Figure 3). Incorporating damped oscillations into the definition of stability creates a gradual continuum of stability scores. Exclusion of oscillations from the definition of stability creates a jagged landscape of SSS values, separable into classes that can also be discriminated by structural features: the number and size of feedback loops. The particular definition of SSS (including or excluding damped oscillations) does not affect the stability rank of the three-node motifs, only the raw SSS value.
(19 KB PDF).
Click here for additional data file.
Figure S2 Distribution of Normalized Z Scores of Four-Node Network Motifs Using a More Expansive Definition of SSS That Includes Damped Oscillations
The SSS is represented as a stacked bar graph (top panel) in which the black portion of the bars represents the component of SSS due to monotonic decay, and the grey portion of the bars represents the component of SSS due to damped oscillations. The ordering of four-node motifs by the more expansive definition (traditional in the field of control systems engineering) produces a substantially similar ordering of four-node motifs as the conservative definition of SSS that excludes oscillations (compare to Figure 4). The Spearman rank correlation coefficient (r), which expresses the correlation between the rank of the SSS and the rank of the Z score, is indicated for selected density groups, as well as a p-value indicating the probability that such a correlation coefficient occurs by chance.
(168 KB PDF).
Click here for additional data file.
Figure S3 Distribution of Normalized Z Scores of Three-Node Network Motifs Computed Using a Degree-Constrained Null Model, Identical to the Method Previously Proposed [3,25]
Overrepresented motifs (positive Z scores) tend to have higher SSS scores than other motifs with the same number of edges (compare to Figure 3). SSS does not provide insight into the occurrence of all underrepresented motifs (negative Z scores).
(19 KB PDF).
Click here for additional data file.
Figure S4 Distribution of Normalized Z Scores of Four-Node Network Motifs Computed Using a Degree-Constrained Null Model Identical to the Method Previously Proposed [3,25]
Overrepresented motifs (positive Z scores) tend to have higher SSS scores than other motifs with the same number of edges (compare to Figure 4). SSS does not provide insight into the occurrence of all underrepresented motifs (negative Z scores).
(165 KB PDF).
Click here for additional data file.
Figure S5 The Regulation of Putative Yeast Gene YHR087W in Various Experimental Contexts
(A) The full potential network affecting the expression of YHR087W (bottom node) was topologically sorted to reveal the causal structure. Green nodes indicate that the gene was regulated (> 2-fold change in expression) in at least one of the five environmental perturbation experiments examined here (heat shock, H2O2, dithiothrietol, diamide, hyper-osmotic shock). Green links are regulatory interactions that were inferred from the regulated genes, irrespective of whether or not the mRNA expression of the upstream transcription factor was observed. (All inbound links to a regulated node are inferred to be potentially active.)
(B) The context-dependent regulatory network in response to a mild heat shock was obtained by traversing the green links backwards from YHR087W, stopping when no more green links were encountered. The regulatory unit is larger than the average motif size of three nodes, yet dynamical stability is a robust property of the structure.
(C) The context-dependent regulatory networks for hyper-osmotic shock and diamide perturbations are a subgraph of the heat-shock regulatory structure.
(B and C) The context-dependent motifs have SSS = 1, indicating that dynamical stability is a robust property of the structure of the regulatory unit. Images generated with Cytoscape1.1 software [33].
(56 KB PDF).
Click here for additional data file.
Protocol S1 Four-Node Motif Dictionary
ID labels match the output from mfinder1.1 motif-finding software provided by U. Alon.
(215 KB PDF).
Click here for additional data file.
Protocol S2 Theoretical Analysis
Stability analysis using root locus.
(45 KB PDF).
Click here for additional data file.
We thank Uri Alon for providing datasets and the mfinder software. We are grateful to Elias Issa, Jef Boeke, members of the Levchenko lab, and three anonymous reviewers for discussion and critical review. This work was supported by the National Institutes of Health (1-U54-RR020839 to AL, and 71920 to PAI) and the National Science Foundation (083500 to PAI).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. RJP, PAI, and AL conceived and designed the experiments, performed the experiments, and analyzed the data. RJP, PAI and AL contributed reagents/materials/analysis tools. RJP, PAI, and AL wrote the paper.
Citation: Prill RJ, Iglesias PA, Levchenko A (2005) Dynamic properties of network motifs contribute to biological network organization. PLoS Biol 3(11): e343.
Abbreviations
ERErdös-Renyi
SSSstructural stability score
STKEsignal transduction knowledge environment
==== Refs
References
White JG Southgate E Thomson JN Brenner S The structure of the nervous system of the nematode C. elegans
Philos Trans R Soc Lond B Biol Sci 1986 275 1 340
Barabasi AL Oltvai ZN Network biology: Understanding the cell's functional organization Nat Rev Genet 2004 5 101 113 14735121
Milo R Shen-Orr S Itzkovitz S Kashtan N Chklovskii D Network motifs: Simple building blocks of complex networks Science 2002 298 824 827 12399590
Shen-Orr SS Milo R Mangan S Alon U Network motifs in the transcriptional regulation network of Escherichia coli
Nat Genet 2002 31 64 68 11967538
Mangan S Alon U Structure and function of the feed-forward loop network motif Proc Natl Acad Sci U S A 2003 100 11980 11985 14530388
Mangan S Zaslaver A Alon U The coherent feed-forward loop serves as a sign-sensitive delay element in transcription networks J Mol Biol 2003 334 197 204 14607112
Kalir S Mangan S Alon U A coherent feed-forward loop with a SUM input function prolongs flagella expression in Escherichia coli
Mol Syst Biol 2005 10.1038/msb4100010
Sporns O Kotter R Motifs in brain networks PLoS Biol 2004 2 e369 15510229
Lee TI Rinaldi NJ Robert F Odom DT Bar-Joseph Z Transcriptional regulatory networks in Saccharomyces cerevisiae
Science 2002 298 799 804 12399584
Song S Sjostrom PJ Reigl M Nelson S Chklovskii DB Highly nonrandom features of synaptic connectivity in local cortical circuits PLoS Biol 2005 3 e68 15737062
Levchenko A Bruck J Sternberg PW Regulatory modules that generate biphasic response in biological systems IEE Proc Sys Biol 2004 1 139 148
Wolf DM Arkin AP Motifs, modules and games in bacteria Curr Opin Microbiol 2003 6 125 134 12732301
Savageau MA Design principles for elementary gene circuits: Elements, methods, and examples Chaos 2001 11 142 159 12779449
Spirin V Mirny LA Protein complexes and functional modules in molecular networks Proc Natl Acad Sci U S A 2003 100 12123 12128 14517352
Rao CV Arkin AP Control motifs for intracellular regulatory networks Annu Rev Biomed Eng 2001 3 391 419 11447069
Francois P Hakim V Design of genetic networks with specified functions by evolution in silico Proc Natl Acad Sci U S A 2004 101 580 585 14704282
Ozbudak EM Thattai M Lim HN Shraiman BI Van Oudenaarden A Multistability in the lactose utilization network of Escherichia coli
Nature 2004 427 737 740 14973486
Alon U Surette MG Barkai N Leibler S Robustness in bacterial chemotaxis Nature 1999 397 168 171 9923680
Barkai N Leibler S Robustness in simple biochemical networks Nature 1997 387 913 917 9202124
Guet CC Elowitz MB Hsing W Leibler S Combinatorial synthesis of genetic networks Science 2002 296 1466 1470 12029133
Stelling J Sauer U Szallasi Z Doyle FJ Doyle J Robustness of cellular functions Cell 2004 118 675 685 15369668
Hoffmann A Levchenko A Scott ML Baltimore D The IkappaB-NF-kappaB signaling module: Temporal control and selective gene activation Science 2002 298 1241 1245 12424381
Elowitz MB Levine AJ Siggia ED Swain PS Stochastic gene expression in a single cell Science 2002 297 1183 1186 12183631
Itzkovitz S Milo R Kashtan N Ziv G Alon U Subgraphs in random networks Phys Rev E Stat Nonlin Soft Matter Phys 2003 68 26127
Milo R Itzkovitz S Kashtan N Levitt R Shen-Orr S Superfamilies of evolved and designed networks Science 2004 303 1538 1542 15001784
Kashtan N Itzkovitz S Milo R Alon U Topological generalizations of network motifs Phys Rev E Stat Nonlin Soft Matter Phys 2004 70 31909
Gasch AP Spellman PT Kao CM Carmel-Harel O Eisen MB Genomic expression programs in the response of yeast cells to environmental changes Mol Biol Cell 2000 11 4241 4257 11102521
Luscombe NM Babu MM Yu H Snyder M Teichmann SA Genomic analysis of regulatory network dynamics reveals large topological changes Nature 2004 431 308 312 15372033
Elowitz MB Leibler S A synthetic oscillatory network of transcriptional regulators Nature 2000 403 335 338 10659856
El-Samad H Kurata H Doyle JC Gross CA Khammash M Surviving heat shock: Control strategies for robustness and performance Proc Natl Acad Sci U S A 2005 102 2736 2741 15668395
Variano EA McCoy JH Lipson H Networks, dynamics, and modularity Phys Rev Lett 2004 92 188701 15169539
Harbison CT Gordon DB Lee TI Rinaldi NJ Macisaac KD Transcriptional regulatory code of a eukaryotic genome Nature 2004 431 99 104 15343339
Shannon P Markiel A Ozier O Baliga NS Wang JT Cytoscape: A software environment for integrated models of biomolecular interaction networks Genome Res 2003 13 2498 2504 14597658
|
16187794
|
PMC1239925
|
CC BY
|
2021-01-05 08:21:51
|
no
|
PLoS Biol. 2005 Nov 4; 3(11):e343
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030343
|
oa_comm
|
==== Front
PLoS BiolPLoS BiolpbioplosbiolPLoS Biology1544-91731545-7885Public Library of Science San Francisco, USA 10.1371/journal.pbio.0030369SynopsisBioengineeringBioinformatics/Computational BiologyBiophysicsEvolutionSystems BiologyBiochemistrySaccharomycesEubacteriaCharting the Interplay between Structure and Dynamics in Complex Networks Synopsis11 2005 4 10 2005 4 10 2005 3 11 e369Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Dynamic Properties of Network Motifs Contribute to Biological Network Organization
Motifs, Control, and Stability
==== Body
While intelligent-design proponents enjoy their 15 minutes of fame denying the role of evolutionary forces in generating complex networks in nature, scientists are probing the organizing principles that govern these networks. Traditional models of complex networks assumed that connections between units—such as genes, proteins, neurons, or species—occur randomly. These notions changed as studies of protein interaction networks and other biological systems revealed “small world” features—characterized by short paths between nodes and highly clustered connections—and varying levels of organization, with certain patterns of local connections occurring more frequently in complex networks than in random networks. What determines the abundance of these so-called network motifs in specific networks is not known.
To study complex, dynamic systems, researchers create graphical representations with network maps. Though the network structure is static, the nodes represent values that change with time. For example, the neural map of the worm Caenorhabditis elegans contains 302 neurons (represented as nodes) and roughly 7,000 synaptic connections that appear fixed even though they represent transient behavior, such as activation states of individual neurons and their probable interactions. Discerning the global dynamics of these network structures has proved a major challenge.
In a new study, Robert Prill, Pablo Iglesias, and Andre Levchenko use the power of computational analysis to tackle the problem of identifying the dynamic features of a large-scale network. They found a high correlation between a dynamic property of a network motif—ability to withstand small fluctuations in the system—and its relative abundance in well-characterized biological networks. Their results suggest that just as connections between individual components of a biological network—be they genes, proteins, or cells—influence function, the dynamic properties of a network motif relate to the motif's function and could determine its prevalence in biological networks.
In this schematic of a transcriptional network, nodes represent operons (a set of bacterial structural genes and their regulatory elements) and links represent transcription factor DNA binding interactions. (Image generated with Cytoscape v1.1)
For a network motif to qualify as stable, it must return to steady state after small-scale perturbations, defined as intrinsic random fluctuations, or noise, and transient oscillations in activity. The behavior of a motif is determined by the direction, sign (presence of positive or negative feedback loops), and strength of the connections. The authors varied these parameters to simulate motif response to small perturbations. To measure stability, the authors assigned a structural stability score (SSS) as the probability that a particular motif returns to a postperturbation steady state. They use this metric to analyze the dynamics of all possible three- or four-node networks (noting that even two-node networks exhibit complex behavior). Based on the SSS scores, all the structurally distinct three- and four-node network motifs fell into three distinct categories: robustly stable circuits with no feedback loops, moderately stable circuits with a single two-node feedback loop (assuming a negative feedback loop), and, least stable, a mixture of complicated, highly connected motifs.
Comparing motif abundance in known biological networks with the SSS scores of simulated motifs revealed an “excellent correlation” between stability and motif abundance. Higher stability motifs were more abundant in the real networks, while low-stability motifs were absent, suggesting that the nonrandom character of network organization is driven by the structural stability of network motifs. To see how these motifs might operate as functional units, the authors used microarray data from yeast subjected to five different environmental stresses, including mild heat shock and hydrogen peroxide treatment, and mapped activated genes (represented as nodes) to their locations in the network. All the active regulatory motifs had a high stability score, suggesting that the nonrandom nature of the yeast transcriptional network may have arisen from selection acting on small motifs that respond robustly to specific environmental stresses. Expanding their analysis to other biological networks, the authors found that yeast and the pathogen Escherichia coli have similar motif profiles, likely reflecting similar environmental pressures, while the fruit fly transcription program and worm neuron network contain different motifs, reflecting both different environmental and functional demands.
These results suggest that both global constraints on the network and properties of network motifs themselves influence the abundance of motifs and the overall structure of a given network. While the authors caution that their networks are stripped-down versions of those found in biological systems, they point out that their approach can incorporate more complicated interactions as understanding of living networks increases. And with this new understanding, scientists can test the hypothesis that selective pressures favor motifs with particular dynamic properties. For more information on structural stability and networks, please see the accompanying Primer by Doyle (DOI: 10.1371/journal.pbio.0030392). —Liza Gross
|
0
|
PMC1239926
|
CC BY
|
2021-01-05 08:21:28
|
no
|
PLoS Biol. 2005 Nov 4; 3(11):e369
|
utf-8
|
PLoS Biol
| 2,005 |
10.1371/journal.pbio.0030369
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1647093710.1371/journal.pgen.001003905-PLGE-RA-0062R3plge-01-03-11Research ArticleBioinformatics - Computational BiologyEvolutionGenetics/Gene ExpressionYeast and FungiSaccharomycesComparative Gene Expression Analysis by a Differential Clustering Approach: Application to the Candida albicans Transcription Program Analysis of
Candida Transcription
Ihmels Jan 1Bergmann Sven 12Berman Judith 3Barkai Naama 1*1 Departments of Molecular Genetics and Physics of Complex Systems, Weizmann Institute of Science, Rehovot, Israel
2 Department of Medical Genetics, University of Lausanne, Switzerland
3 Departments of Genetics, Cell Biology & Development, and Microbiology, University of Minnesota, Minneapolis, Minnesota, United States of America
Kruglyak Leonid EditorPrinceton University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 3 e3930 3 2005 12 8 2005 Copyright: © 2005 Ihmels et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Differences in gene expression underlie many of the phenotypic variations between related organisms, yet approaches to characterize such differences on a genome-wide scale are not well developed. Here, we introduce the “differential clustering algorithm” for revealing conserved and diverged co-expression patterns. Our approach is applied at different levels of organization, ranging from pair-wise correlations within specific groups of functionally linked genes, to higher-order correlations between such groups. Using the differential clustering algorithm, we systematically compared the transcription program of the fungal pathogen Candida albicans with that of the model organism Saccharomyces cerevisiae. Many of the identified differences are related to the differential requirement for mitochondrial function in the two yeasts. Distinct regulation patterns of cell cycle genes and of amino acid metabolic genes were also revealed and, in some cases, could be linked to the differential appearance of cis-regulatory elements in the gene promoter regions. Our study provides a comprehensive framework for comparative gene expression analysis and a rich source of hypotheses for uncharacterized open reading frames and putative cis-regulatory elements in C.
albicans.
Synopsis
Candida albicans is a fungal inhabitant of the intestinal tract of most healthy humans. It becomes a serious and often lethal pathogen in people with a weak immune system. C. albicans is a distant relative of the well-studied baker's yeast, Saccharomyces cerevisiae. It is now possible to determine the degree to which these two fungi have similar or different patterns of transcription.
Here, methods were developed that comprehensively compare the expression patterns of S. cerevisiae and C. albicans. A novel algorithm was used to determine if the expression of groups of genes in one organism are fully, partially, or not at all similar in the other organism. This algorithm was first applied to pre-defined groups of genes predicted to have similar functions and was then used to compare the global organization of the transcription programs between the two organisms.
The analysis revealed that the expression patterns reflect the different metabolic preferences of the two yeasts. The authors also found that amino acid metabolism regulation is more differentiated in C.
albicans. Furthermore, the different expression patterns can be traced down to the use of different regulatory sequences. This study provides a comprehensive framework for comparative gene expression analysis, as well as a Web site with interactive analysis tools, which allow the development of hypotheses concerning uncharacterized genes and the sequences that regulate them.
Citation:Ihmels J, Bergmann S, Berman J, Barkai N (2005) Comparative gene expression analysis by a differential clustering approach: Application to the Candida albicans transcription program. PLoS Genet 1(3): e39.
==== Body
Introduction
Phenotypic diversity can often be traced to the differential expression of specific regulatory genes [1–5]. Recently, microarray experiments revealed large-scale differences in the genome-wide transcription response of related organisms to equivalent environmental conditions. For example, the transcription program underlying insect metamorphosis differs considerably between related species of the Drosophila melanogaster subgroup [6]. Similarly, both the meiotic and the mitotic cell cycle transcription program have diverged significantly between the budding and the fission yeasts [7]. The impact of such large-scale variations in gene expression on the phenotypes of the organisms is not yet understood.
Existing computational approaches for the comparative analysis of large-scale gene expression data have focused primarily on evolutionarily distant model organisms, for which large sets of expression data are available [8–11]. Such studies demonstrated that conservation of co-expression can improve functional gene annotation [9,10]. Common expression programs are invoked by related perturbations, such as adult onset in the nematode Caenorhabditis elegans, and the fruit fly D. melanogaster [11]. A generalization of the singular value decomposition approach that is applicable for such a comparative study was applied to cell cycle datasets from Saccharomyces cerevisiae and human [8]. Yet, the challenge of systematically comparing the gene expression program in related organisms is only starting to be addressed.
Candida albicans is an opportunistic pathogen that causes mucosal as well as systemic infections, especially in immune-compromised human hosts. This budding ascomycetous yeast diverged from the S. cerevisiae lineage between 140 and 800 million years ago [12,13]. Recently, the C. albicans genome was sequenced [14], revealing that almost two-thirds of its ~6,000 open reading frames are orthologous to S. cerevisiae genes. Microarray studies were performed by several groups characterizing the C. albicans genome-wide expression program under a range of conditions [15–24]. The availability of large sets of expression data in both S. cerevisiae and C. albicans, which are related organisms that span a significant evolutionary distance, provides a useful framework to develop and test computational tools for comparative gene expression analysis.
Here we present a novel approach for comparative gene expression analysis. We demonstrate the utility of our methods by systematically comparing the C. albicans and S. cerevisiae transcription programs at different levels of organization, ranging from the co-expression patterns between genes, to higher-order relationships between functional attributes. We describe large-scale differences in the transcription programs of the two organisms and use promoter analysis to link some of these differences to distinct cis-regulatory elements. All our results, as well as interactive analysis tools, are accessible through our Web server at http://barkai-serv.weizmann.ac.il/candida.
Results
C. albicans Expression Data
We assembled a dataset describing the genome-wide transcriptional responses of C. albicans to diverse perturbations, including drug resistance [15–17], stress [18], expression of only one mating type locus [19], and response to mating pheromone [20]. Also included were transcription profiles of cells growing as yeast or hyphal cells [25], in biofilms [21], exposed to blood components [22,23], altered pH [24], or signaling molecules [26,27]. The studies were performed primarily with laboratory strains, but also with some clinical isolates [15,21,24]. Altogether, the dataset consists of 244 expression profiles, generated by seven different laboratories, using four independently designed microarrays. All data were put into a unified format (orf19), which included a total of 6,167 open reading frames (ORFs) (see Materials and Methods).
Previous studies demonstrated that genes with similar functions are often co-expressed (see [28–31]). To determine if this relationship is observed in the C. albicans expression data, we examined the similarity of the expression patterns of genes assigned to the same biological process within the Gene Ontology (GO) categories [32]. The significance of co-expression within a specific GO category was quantified by calculating the distribution of pair-wise correlations between genes within the category, and by comparing it to the distribution of random gene assemblies of the same size (see Materials and Methods and Figure S1). Indeed, a large fraction of predicted GO categories received a highly significant score, indicating that, also in the C. albicans data, functionally linked genes tend to be co-expressed (Figure 1A).
Figure 1 Functionally Linked Genes Tend to Be Co-Expressed
(A) The extent of correlations between genes assigned to a particular GO category was quantified by the t-value associated with the distribution of pair-wise correlations between genes within the category (given in units of standard deviation (σ of the control distribution; see Materials and Methods). Shown is the fraction of GO categories whose t-value exceeds a threshold value T, as a function of T. As a control, we repeated the analysis for random assignment of genes into the GO categories (red). A similar analysis using genes assigned to a particular KEGG category is show in Figure S2.
(B) The significance of GO term co-expression in C. albicans versus S. cerevisiae. Each dot corresponds to a specific GO category. GO terms that are significantly correlated in both organisms (t-value > 4σ) are colored in black, whereas those that are significantly correlated in only one organism are colored in blue or green.
(C) PCMs of genes assigned to the indicated GO categories. Only genes defined as orthologous between C. albicans and S. cerevisiae were considered (Materials and Methods). Orthologs are arranged in the same order in both organisms. Mitochondrial and cytoplasmic genes are colored blue and magenta, respectively.
For comparison, we performed an analogous analysis of S. cerevisiae, using a dataset of ~1,000 publicly available genome-wide expression profiles [33]. Overall, the significance of co-expression within the C. albicans GO terms was lower than that of the S. cerevisiae counterparts (Figures 1A and S1). This lower significance may reflect the smaller size of the dataset available for C. albicans, its quality, or the fact that the GO terms were originally defined for S. cerevisiae. Alternatively, transcriptional regulation may play a less prominent role in C. albicans. The mitochondrial-targeting and protein-folding GO categories, which were co-expressed more tightly in C. albicans, provided an interesting exception, although the significance of this difference was marginal (Figure 1B). Despite the quantitative difference, we observed a strong correlation between the significance of the co-expression in the two organisms (correlation coefficient 0.92, Figure 1B). For example, in both organisms, functional groups involved in aspects of protein synthesis and sugar metabolism were most stringently co-expressed.
Differential Clustering Algorithm for Comparative Analysis of Gene Expression Data
While providing a useful means for systematic analysis, GO categories do not necessarily correspond to transcriptional units. In fact, in most GO categories, only a subset of the genes is co-expressed (e.g., Figure 1C). Moreover, in certain cases, a single GO category can be separated into subsets that display independent or even inversely correlated expression patterns. For example, the C. albicans genes attributed to gluconeogenesis were split into two autonomously co-expressed subgroups, one associated with the glycolysis pathway itself, the other involved in other aspects of gluconeogenesis. Interestingly, in this case, this split was conserved between S. cerevisiae and C. albicans (Figure 1C). However, in general, the fine structures in regulatory patterns differed between the two organisms (e.g., tRNA aminoacetylation, Figure 1C).
Differences in the pattern of gene regulation within individual GO categories are likely to reflect differences in the physiology, or in the adaptation to different environments, of the two organisms. Existing approaches for comparative gene expression analyses emphasize mostly conserved co-regulation patterns, rather than differences in expression patterns [8,9,11]. To better capture differential expression patterns, we developed a novel approach, termed the differential clustering algorithm (DCA), for systematically characterizing both similarities and differences in the fine structure of co-regulation patterns (Figure 2).
Figure 2 The Differential Clustering Algorithim (DCA)
(A) PCMs are calculated (see Materials and Methods for details).
(B) The PCMs are combined into a single matrix, where each triangle corresponds to one of the PCMs (1). The genes are then ordered in two steps: First, genes are clustered and the PCMs are re-arranged according to the correlations in the reference organism (“B”) (2). Second, the genes assigned to each of the resulting primary clusters are re-clustered according to their correlations in the “target” organism “A” (secondary clustering) (3). Note that, at each step of the clustering, orthologous genes are arranged in the same order in both organisms. The procedure is then repeated reciprocally, i.e., this time using organism “A” as the reference and organism “B” as the target. Finally, the conservation patterns of each cluster are classified automatically into one of the four conservation classes (4) (see also Figure 3A).
Figure 3 The DCA Method Automatically Classifies Clusters to Different Conservation Classes
(A) Classification flowchart: Each primary cluster is subdivided into two secondary clusters, a and b. The cluster is then characterized by three correlation values, corresponding to the average correlations of genes within (Ca, Cb < Ca) and between (Cab) these clusters. These correlations determine its assignment to one of four basic conservation patterns as depicted in the flowchart. The cutoff parameter T is chosen heuristically.
(B) Classification values for clusters derived from functional GO categories. A list of clusters was obtained by applying the DCA method to sets of orthologous genes assigned to all functional GO categories containing between five and 200 orthologs. Sets of genes containing a specific sequence element in their 600-basepair promoter region were also considered (Materials and Methods). Shown are examples of clusters classified to each of the four basic types of conservation. Importantly, this assignment of clusters to the different conservation categories is robust to sub-sampling of the available conditions (Figure S3).
(C) PCMs of the clusters shown in (B). Color code is as in Figure 1. Additional clusters related to these categories and gene names associated with all of the clusters are provided at http://barkai-serv.weizmann.ac.il/candida. Mitochondrial and cytoplasmic genes are colored blue and magenta, respectively. The category above each cluster refers to the GO term or sequence from which it was obtained. Note that each PCM represents only one cluster derived from the corresponding category, such that in general only a subset of the genes assigned to each category is shown.
The DCA is applied to a set of orthologous genes that are present in both organisms. As a first step, the pair-wise correlations between these genes are measured in each organism separately, defining two pair-wise correlation matrices (PCMs) of the same dimension (i.e., the number of orthologous genes) (Figure 2A). Next, the PCM of the primary (“reference”) organism is clustered, assigning genes into subsets that are co-expressed in this organism, but not necessarily in the second (“target”) organism. Finally, the genes within each co-expressed subgroup are re-ordered, by clustering according to the PCM of the target organism. This procedure is performed twice, reciprocally, such that each PCM is used once for the primary and once for secondary clustering, yielding two distinct orderings of the genes.
The results of the DCA are presented in terms of the rearranged PCMs. Since these matrices are symmetric and refer to the same set of orthologous genes, they can be combined into a single matrix without losing information. Specifically, we join the two PCMs into one composite matrix such that the lower-left triangle depicts the pair-wise correlations in the reference organism, while the upper-right triangle depicts the correlations in the target organism (Figure 2B). Inspection of the rearranged composite PCM allows for an intuitive extraction of the differences and similarities in the co-expression pattern of the two organisms (Figure 3). An automatic scoring method is then applied to classify clusters into one of the four conservation categories: full, partial, split, or no conservation of co-expression (Figure 3A and 3B).
Functionally Related Genes Exhibit Different Degrees of Co-Expression Conservation
To systematically characterize the conservation or divergence of co-expression between genes with a related function, we applied the DCA to gene groups defined by membership in the same biological process GO categories [32]. We also applied it to groups of genes that have a common DNA sequence motif of length 6 or 7 base-pairs in their promoter region (within 600 base-pairs upstream of the predicted start codon). The DCA procedure identifies co-expressed clusters embedded within these gene sets, and assigns each of these clusters to one of the four above-mentioned conservation categories (full, partial, split, or no conservation, Figure 3).
Examples of clusters assigned to each category are shown in Figure 3C. Clusters associated with growth, including genes encoding ribosomal components (Figure 3C, 14) and genes containing the PAC motif (Figure 3C, 13, primarily genes encoding rRNA processing proteins), were strongly co-regulated in both organisms, and were thus assigned to the full conservation class. Also assigned to this class were clusters of genes involved in oxidative phosphorylation (Figure 3C, 15) and monosaccharide catabolism (Figure 3C, 16).
Of particular interest are clusters that are differentially expressed between the two organisms. The most noticeable differences were found for clusters whose genes are involved in both cytoplasmic and mitochondrial translation. This included, for example, the GO terms “protein synthesis” (Figure 3C, 9), “tRNA metabolism” (Figure 3C, 5), and “tRNA amino-acetylation” (Figure 1C). These clusters were uniformly co-expressed in C. albicans. In contrast, in S. cerevisiae they were split into two distinct subclusters, associated with cytoplasmic or mitochondrial functions, respectively, which displayed independent or even inversely correlated expression patterns. This differential expression pattern of mitochondrial genes reflects a major phenotypic difference between the two organisms: rapidly growing S. cerevisiae cells utilize fermentation and do not require oxygen. In contrast, rapid growth in C. albicans relies on aerobic respiration and requires mitochondrial functions.
Flexible Regulatory Patterns of Cell Cycle Genes
Among the clusters assigned to the no conservation class was a group of cell cycle genes that are involved in the transition from S-phase to mitosis (Figure 3C, 3). These genes were tightly co-expressed in C. albicans, but not in S. cerevisiae, suggesting that the cell cycle transcription program differs between the two organisms.
To better characterize the differences in regulation of cell cycle genes, we examined the “cell cycle” GO category in more detail. We included in this analysis also expression data from Schizosaccharomyces pombe [7,34], which is evolutionarily more distant to S. cerevisiae and C. albicans [13]. For S. cerevisiae and S. pombe, we also restricted the expression data to cell cycle experiments. No such cell cycle–dedicated conditions were available for C. albicans. We note, however, that many experiments in the C. albicans dataset used cells emerging from stationary phase with some degree of synchrony, which likely captured some features of cell cycle–specific regulation. It should be noted that the gene set is based on the S. cerevisiae GO term, and therefore does not include genes that are cell cycle–related only in the other two organisms.
We applied the DCA to the above-mentioned data, with each of the three yeasts serving once as a reference and once as a target organism (off-diagonal in Figure 4, green background). As a control, we considered the same organism as both the reference and target organism, but used only 25% of the expression data for the secondary clustering (diagonal in Figure 4, gray background). Moreover, for S. cerevisiae and S. pombe, we tested complementary expression data containing no cell cycle experiments as another control. In this case the cluster conservation was weaker, yet some aspects of cell cycle regulation remained (unpublished data).
Figure 4 DCA Analysis of Cell Cycle Genes
(A) The DCA algorithm was applied to a restricted gene set, consisting of all genes common to S. cerevisiae, C. albicans, and S. pombe, with GO annotation “cell cycle.” The reference organism is indicated on the left, whereas the target organism is indicated on the top. Most of the primary clusters (white boxes) are, at most, partially co-expressed in the target organism (green background). The diagonal (gray background) represents the control, where the secondary clustering is performed for the same species as in the primary clustering, but using a reduced set (25%) of the expression data.
(B) Primary clusters from (A) that contain CDC28. Note that CLB2 and CDC5 are tightly co-expressed in S. cerevisiae and C. albicans (but not in S. pombe), but that CDC28 is co-expressed with these genes only in C. albicans. Details of all other clusters are available in Figures S4–S13.
Essentially all clusters identified as co-expressed in the reference organism were, at most, partially co-expressed in the other two organisms (Figures 4 and S4–S13). As an example, we highlight here the regulation of the major cyclin-dependent kinase (encoded by CDC28 in S. cerevisiae) and the associated mitotic B-cyclin (encoded by CLB2) (Figure 4B).
In S. cerevisiae, there are six B-cyclins, several with redundant functions [35–38], and their expression is cell cycle–regulated. CDC28 expression is not correlated with any of them. Accordingly, CDC28 and CLB2 were associated with two distinct clusters: CDC28 was assigned to a cluster composed of genes involved in the early cell cycle functions (e.g., budneck formation, DNA replication, and repair [Figure S4]), whereas CLB2 was assigned to a cluster composed of genes with functions in mitosis (Figure S12). Neither of these clusters was co-expressed in C. albicans or in S. pombe.
S. pombe has one major, essential B-cyclin, cdc13 (the CLB2 ortholog), which is required for mitosis. In the S. pombe cell cycle data, expression of cdc13 was inversely correlated with expression of cdc2 (the CDC28 ortholog). cdc2 was co-expressed with a cluster of genes, many of whose S. cerevisiae orthologs are involved in replication and DNA repair (Figure S9), whereas cdc13 was co-regulated with genes involved primarily in mitosis and general cell cycle control (Figure S11).
C. albicans has two B-cyclins, and one of them, CLB2, is essential [39]. Interestingly, in C. albicans the CDC28 and CLB2 orthologs were co-expressed. Both genes were assigned to a cluster associated with anaphase and mitotic exit (Figures 4B and S11). Northern blot analysis of CDC28 and CLB2 transcripts in C. albicans cells emerging synchronously from stationary phase confirmed that the mRNA levels of CDC28 and CLB2 correlate, peaking with the presence of large budded cells (S/G2 phase) (JB and M. McClellan, unpublished data).
We conclude that transcriptional regulation of cell cycle genes is highly flexible and has diverged significantly between the three yeast species. Our results expand on previous reports that have shown that only a small set of genes are subject to similar cell cycle regulation in both S. cerevisiae and S. pombe [7,40]. Each of these fungi has a distinctive repertoire of morphologies: S. cerevisiae and C. albicans undergo budding to form yeast or pseudohyphae; C. albicans also forms true hyphae by a non-budding mechanism involving different organellar structures [41]; S. pombe is a fission yeast with a distinct, non-budding mechanism of morphogenesis. In all three fungi, cell cycle regulation and morphogenesis are clearly linked [39,42,43]. Further analysis is needed to establish how these distinct morphologies are connected to the differential pattern of gene expression found in each organism.
C. albicans Transcription Modularity
The analysis above focused on pre-defined sets of genes that are known to be related and thus are suspected to be, at least partially, co-expressed. In order to identify novel regulatory relationships that are not confined to specific function-related genes, we conducted a complementary, unsupervised analysis of the C. albicans expression data. To this end, we used the iterative signature algorithm (ISA) [31,44] to determine the modular organization of the C. albicans transcription program. The ISA segregates the data into overlapping transcription modules, each consisting of a subset of co-expressed genes together with the subset of experimental conditions inducing this co-expression.
The ISA assigned 2,770 C. albicans genes into transcription modules with varying stringencies of correlated expression. Modules were classified as core modules (15%), composed primarily of genes possessing an S. cerevisiae ortholog; as C. albicans–specific modules (37%), consisting primarily of genes without S. cerevisiae orthologs; or as modules with a mixture of both types of genes (48%) (Figure 5A–5C).
Figure 5
C. albicans Module Tree
(A) Transcription modules were identified using the ISA [31,44]. Modules were annotated manually, and are colored according to their enrichment for S. cerevisiae orthologs or C. albicans–specific genes. An interactive version of the tree with details of the genes and conditions of each module is provided at http://barkai-serv.weizmann.ac.il/candida.
(B) Proportion of genes without S. cerevisiae orthologs in C. albicans transcription modules (orange), compared to a control distribution obtained from random sets of genes of the same size. Note the over-representation of C. albicans–specific modules.
(C) Distribution of overlaps between transcription modules of C. albicans and S. cerevisiae. For each representative module in C. albicans, the S. cerevisiae module with the highest overlap in terms of orthologous genes was identified and the proportion of overlap was plotted (Materials and Methods).
(D) Transcription modules are significantly enriched in common GO terms and upstream sequence elements. For each transcription module in C. albicans, enrichment p-values were calculated for all GO terms or sequence elements (6-, 7-mers) in the 5′ UTR (Materials and Methods), and the n smallest p-values were recorded for each module. The results are shown for n = 5, but are robust to the precise choice of n. The fraction of categories and sequence elements exceeding a threshold p-value, as a function of the threshold, is shown and compared to a control distribution obtained from random gene sets of the same sizes.
(E) PCMs of genes involved in rRNA processing. Shown are the gene–gene correlation matrices of the top-scoring 25 genes assigned to the rRNA processing module in C. albicans (left panel) and their S. cerevisiae orthologs (right panel). Genes are ordered by their gene score in the C. albicans transcription module.
(F) orf19.5850-YFP, assigned to the rRNA processing module, co-localizes with Nop1-CFP to the nucleolus.
Modules were annotated manually by examining their gene and condition contents (Figure 5A; see also http://barkai-serv.weizmann.ac.il/candida). In addition, we systematically checked each module for over-representation of GO categories and of DNA sequence motifs in the 5′-UTR. This analysis clearly established the biological relevance of the C. albicans transcription modules. First, many modules contained one or several over-represented GO terms, indicating their functional coherence (Figure 5D). Second, most modules were associated with sequence motifs that were significantly enriched in the promoter regions of genes within the module (Figure 5D).
Module association provides numerous functional links for C. albicans genes (see http://barkai-serv.weizmann.ac.il/candida). We experimentally tested one of these links, namely orf19.5850. Previous studies reported that a strain heterozygous for a transposon disruption allele of this gene exhibits reduced filamentous growth [45]. Our analysis assigned orf19.5850 to the rRNA processing module (Figure 5E). Indeed, tagging this predicted protein product with yellow fluorescent protein (YFP) revealed its localization to the nucleolus (Figure 5F), as expected for a gene involved in rRNA processing. After this experiment was initiated, the localization of the S. cerevisiae ortholog was shown to be both nucleolar and nuclear [46].
The C. albicans versus S. cerevisiae Transcription Modularity
The hierarchical organization of a transcription program is captured by its module tree, which connects related modules identified at different stringencies of correlated expression [10,31,44] (Figure 5A). The C. albicans module tree was composed of three main branches. One of these branches was associated with Candida-specific cell types: they were induced in opaque cells and/or repressed in white cells. This module included genes important for fatty acid metabolism, mating, and arginine and glutamine biosynthesis, as well as genes repressed under conditions of biofilm production. The second main branch was composed primarily of modules pertaining to core functions, including genes required for rapid growth (e.g., ribosomal proteins and rRNA processing genes). Finally, the third main branch was associated with carbohydrate metabolism and the response to stress, as well as with genes involved in C. albicans–specific processes such as hyphal or white-opaque growth.
This global organization is similar to that found in the S. cerevisiae module tree, in which two of the major branches were associated with rapid growth and stress-response, respectively [31,44]. In contrast, in higher eukaryotes, including D. melanogaster, C. elegans, Arabidopsis thaliana, and human, these two core functions did not correspond to main branches of the module trees [10].
Apart from this global similarity, the module trees of C. albicans and S. cerevisiae displayed some notable differences. First, in C. albicans, amino acid biosynthesis was associated with the protein synthesis branch, whereas no such association was seen in S. cerevisiae [31,33,44]. This indicates that in C. albicans, but not in S. cerevisiae, amino acid biosynthesis is induced under conditions that also increase protein synthesis (e.g., rapid growth). To test if these differences arose from the distinct types of conditions available in the two datasets, we removed from the S. cerevisiae data all environmental perturbations relevant for amino acid metabolism (such as amino acid or nitrogen starvation). We also removed other subsets of conditions, such as the set of 300 profiles of deletion mutants [47], or the set of general environmental perturbations [48]. In all cases, the amino acid and the protein synthesis modules appeared on separate branches (unpublished data). This indicates that the observed distinctions in the module trees of the two yeasts reflect differences in the underlying organization of their transcriptional programs, rather than differences in the set of available conditions.
In C. albicans, the core protein synthesis branch also included specific modules, which contained members of the major repeat sequence family [49] along with genes important for cell wall synthesis and several genes involved in cell cycle progression, such as CLB2, CDC5, and CDC28. The reason for this association of cell wall proteins, the major repeat sequence family, and cell cycle genes is not clear. Examining the conditions associated with this module, we noted that this module is induced primarily in white cells and is repressed primarily in opaque cells [19], and thus may reflect a common regulation associated with the conditions used to study the white-opaque transition.
An intriguing feature of the C. albicans–specific branch of the transcription program is that genes related to arginine biosynthesis were separated from the main amino acid biosynthesis module. These genes were co-expressed with genes required for biotin synthesis, most likely because biotin is required for the activity of ornithine transcarbamylase (encoded by ARG3) [50]. In addition, these genes were co-expressed with genes associated with the mating response [19,20] and were up-regulated in C. albicans cells interacting with macrophages [23]. Because methylated arginines are inhibitors of nitric oxide [51], which is produced by macrophages, it is tempting to speculate that the expression of genes required for arginine synthesis elicits a protective response of C. albicans cells to macrophage attack.
Furthermore, in C. albicans, the mitochondrial ribosomal protein module and the ergosterol biosynthesis module both appear on the protein synthesis branch associated with rapid growth. In contrast, the S. cerevisiae mitochondrial ribosomal protein module is associated with stress responses. Again, this pattern of co-regulation likely reflects the fact that rapid growth requires mitochondria-mediated respiration in C. albicans but not in S. cerevisiae.
Higher-Order Regulatory Relationships between GO Terms Provide Complementary Views of Transcription Programs
The above direct comparison of the two module trees is useful for distinguishing broad features of the respective organizations, yet it is limited by the lack of a one-to-one relationship between the two module sets. For example, the average overlap between S. cerevisiae modules and their best matching C. albicans counterparts is only 19% (Figure 5C). Furthermore, although many modules are significantly enriched with genes belonging to a specific GO category, typically several distinct GO categories are represented in each module. Thus, associating each module with one summarizing annotation is a simplification that does not capture the full complexity of the transcriptional organization.
To overcome these difficulties, we developed a new approach, termed “higher-order connectivity analysis” (HOCA), in which we analyze the modular components of the transcription program through their association with functional categories. Specifically, we define a GO connectivity network, where two GO terms are connected if they are both over-represented in at least one common transcription module (Figure 6A, and Materials and Methods). Applying HOCA to the S. cerevisiae and C. albicans expression data yielded two independent “GO networks,” corresponding to the regulatory relationships between the GO terms in S. cerevisiae and C. albicans, respectively. The two networks were composed of a corresponding set of nodes (GO terms), connected by organism-specific links. We quantified the strength of each link using the topological overlap [52], which weights each edge by the similarity in the overall connectivity of the two nodes (Figure 6A, and Materials and Methods). The C. albicans GO connectivity diagram is displayed in Figure 6B.
Figure 6 Connectivity Analysis between Gene Attributes Reveals Different Patterns of Co-Expression in C. albicans and S. cerevisiae
(A) Generalized attributes (GO terms, sequence motifs, etc.) are connected if they are significantly over-represented in the same transcription module. To analyze the resulting enrichment networks, we first define correlations between attributes based on the topological overlap measure ([52]; see Materials and Methods).
(B) Clustering of the PCM of hierarchical overlaps in C. albicans. Shown is the clustered PCM (left) and a matrix of average correlation/topological overlap values for each cluster (right).
(C) To compare networks between organisms, the DCA method was applied to PCMs of topological overlaps.
(D) Shown are examples of clusters obtained from DCA analysis of the GO networks of C. albicans and S. cerevisiae.
(E) Same as in (D), but using the occurrence of hexa- and heptamer binding motifs in the promoter as gene attributes. (Interactive figures with the list of the GO terms or binding motifs assigned to each cluster, are provided at http://barkai-serv.weizmann.ac.il/candida.)DOI: 10.1371/journal.pgen.0010039.g006
Differential Connectivity in the C. albicans versus S. cerevisiae GO Networks
To compare the GO networks of C. albicans and S. cerevisiae, we restricted the set of nodes to the GO terms that are common to both organisms. In this case, we have two matrices of the same dimension (i.e., the number of common GO terms), describing the topological overlaps between all pairs of GO terms in each organism (Figure 6C). The two matrices were analyzed using the DCA method to automatically classify the resulting clusters of GO terms into the full, split, partial, and no conservation classes of co-expression.
Figure 6D depicts some of the GO term associations assigned to the different conservation classes. Notably, GO terms concerning carbohydrate metabolism (c.f. cluster 3) were correlated with the stress response in C. albicans but not in S. cerevisiae. This may be related to the fact that C. albicans requires mitochondrial function during rapid (aerobic) growth, producing high levels of reactive oxygen species that, in turn, would induce oxidative stress–related genes. In contrast, rapid (fermentive) growth in S. cerevisiae does not generate such high levels of reactive oxygen species and therefore would not induce these genes.
Sequence Motifs Associated with the Differential Regulation of C. albicans Amino Acid Biosynthesis Genes
Consistent with the modular analysis described above, we detected an interesting difference in the regulation of amino acid biosynthesis genes in C. albicans relative to S. cerevisiae. Cluster 5 (Figure 6D) includes GO terms involved in the biosynthesis of several amino acids. All these GO terms are connected in S. cerevisiae, presumably reflecting their common regulation by the transcription factor Gcn4p. In contrast, only one subset of these GO terms (arginine, glutamine, and sulfur amino acid metabolism) was connected in C. albicans. This suggests a differential, and more refined regulation of amino acid biosynthesis by C. albicans.
To better characterize this differential co-regulation pattern, we applied the DCA to the genes of the amino acid biosynthesis transcription module in S. cerevisiae (Materials and Methods). In S. cerevisiae, these genes are uniformly co-expressed. In contrast, in C. albicans this group was split into four clusters that displayed distinct regulatory patterns (Figure 7). These clusters were associated with arginine, methionine, aromatic, and general amino acid biosynthesis.
Figure 7 DCA Analysis of Amino Acid Biosynthesis Genes
(A) Gene–gene correlation matrix for genes assigned to the S. cerevisiae amino acid biosynthesis module. Lower triangle corresponds to the S. cerevisiae data, while the upper triangle depicts the C. albicans correlations.
(B) Sequences motifs over-represented in the different DCA clusters.
To address the mechanism underlying this differential regulatory pattern, we asked whether these clusters are linked to differential appearance of cis-regulatory elements. To this end, we examined the promoter sequences of the genes in each cluster, searching for an over-represented DNA sequence of length 6–8 nucleotides. First, we analyzed the S. cerevisiae promoters and found that, as expected, all clusters were significantly enriched with the TGACTC motif, which is the known binding site for Gcn4p, the transcriptional activator of amino acid biosynthetic genes. Furthermore, the cluster that includes genes required for methionine biosynthesis was associated with an additional motif (CACGTG), which is bound by the Cbf1 transcription factor, a known regulator of methionine biosynthesis genes [53].
Next, we searched for over-represented DNA sequences in the promoters of genes in the C. albicans clusters. The TGACTC motif was significantly enriched in three of the four clusters (numbers 1–3), consistent with previous reports showing its conservation across different yeast species [54,55]. Notably, the cluster associated with methionine biosynthesis genes, which is not co-regulated in our dataset, appears to have lost both the TGACTC (Gcn4-binding) and the CACGTG (Cbf1-binding) motifs (Figure 7).
Strikingly, the three C. albicans clusters that maintained the TGACTC motif were all associated with additional over-represented motifs that were not found in the promoters of the corresponding S. cerevisiae genes (Figure 7). Specifically, the arginine and general amino acid clusters were each associated with a distinct novel motif (TAACCGC and TTCCTG, respectively), whereas all three clusters were associated with the AATTTT [56] motif. These results suggest that combinatorial regulation by different transcription factors underlies the distinct pattern of amino acid biosynthesis genes in C. albicans. Interestingly, the AATTTT motif (or its reverse complement; see Figure 3C, 11) is also enriched in genes involved in ribosome biogenesis and rRNA processing, providing a possible explanation for the observed correlation between amino acid biosynthesis and the protein synthesis branch in the C. albicans module tree.
Differential Connectivity between Cis-Regulatory Elements
The above analysis described the higher-order organization of the C. albicans transcription program based on gene sets sharing functional attributes (i.e., GO categories). A complementary approach is to define putative regulatory units based on common sequence motifs in the 5′-UTRs of its genes.
In a given transcription module, more than one sequence element is typically over-represented. Multiple associations of binding motifs that differ by a single nucleotide likely reflect flexibility in the binding specificity of a single transcription factor. These sequences can be summarized by a consensus motif. Indeed, several clusters of motifs assigned to the “split” conservation pattern correspond to consensus motifs that are partially conserved, but exhibit some organism-specific modifications. Interestingly, many single nucleotide sequence variations of a motif were connected only in S. cerevisiae, suggesting that S. cerevisiae transcription factors tend to have a higher degree of DNA binding flexibility as compared to their C. albicans counterparts. Moreover, the consensus sequences in S. cerevisiae were usually slightly different from those in C. albicans.
Over-representation of several distinct sequence motifs in a given transcription module most likely indicates combinatorial regulation of these genes by different transcription factors. For example, in both organisms, the known consensus motifs PAC [57] and the sequence AAAATT were linked in a single cluster (Figure 6E) pointing to combinatorial action of the associated transcription factors. Moreover, the sequence TGAAAAT was connected to this cluster, but only in S. cerevisiae. This indicates that in S. cerevisiae, the common sequence AAAAAT almost always appears with the prefix TG. In contrast, this TG prefix is not seen in C. albicans. Additional results are summarized at http://barkai-serv.weizmann.ac.il/candida.
Discussion
We present a novel computational approach for the comparative analysis of large-scale gene expression data. Expression data in two organisms were compared at three different levels. First, the DCA was used to analyze co-regulation within specific groups of genes. These groups were assembled based on a priori biological knowledge and are likely to include a subset of co-regulated genes. Focusing on specific functional groups of interest allows the direct analysis of co-expression patterns without interference from genes of unrelated function. Second, the ISA [31,44] was used to identify modules of co-regulated genes. Modular decomposition was performed independently for the two organisms, leading to two module trees that can be compared directly. This unsupervised analysis enables the identification of novel regulatory relationships, which may not be captured by our first analysis based on a priori functional classification. Third, the HOCA was used to rigorously compare the connectivity between different functional units. This analysis relies on the segregation of the expression data into condition-specific transcription modules. Importantly, the HOCA approach can be applied to characterize and compare the connectivity between different types of functional attributes, such as GO terms or cis-regulatory motifs.
A common approach for comparative analysis of gene expression is to consider the transcriptional responses to sets of perturbations that are assumed to be equivalent in both organisms. Yet, robust analysis of gene expression data requires a large number of profiles, and restricting the data to a subset of experiments with common conditions severely restricts the number of available profiles. Moreover, obtaining precisely the same experimental conditions is difficult, if not impossible, when analyzing public datasets. In particular, even when equivalent conditions can be identified, different responses in gene expression could reflect differences in signal transduction mechanisms rather than in the underlying transcriptional network.
The present approach circumvents the need for equivalent experiments because it compares the patterns of gene–gene correlation between the two organisms. The input to the DCA consists of two matrices of the same dimensions, describing the pair-wise similarities between orthologous genes, or groups of genes, measured separately in each dataset. The DCA approach performs clustering sequentially and reciprocally, each time using one set of expression data for primary partitioning and the other dataset to identify the secondary patterns of co-expression within these partitions. Thus, the DCA allows for the identification of diverged, partially conserved, and well-conserved patterns of co-expression between the two datasets. Compared to previous studies that focused primarily on conserved co-regulation [9], this provides an important advantage, especially when more closely related species are analyzed.
It is important to note that, in a heterogeneous compendium of expression profiles, condition-specific co-expression can be obscured when using a simple correlation metric over all conditions. In our initial application of the DCA to the PCMs of pre-defined gene sets, we neglected this issue for simplicity, although this limitation could, in principal, be alleviated using different distance matrices (such as “mutual information” [58]). However, condition-specific co-regulation is taken into account in our global modular analysis using the ISA, as well in our HOCA approach, which is based on module association.
To illustrate the utility of our approaches, we applied them systematically to compare the transcription program of C. albicans with the well-characterized S. cerevisiae program. While the co-expression of many functionally related groups was conserved between C. albicans and S. cerevisiae, our analysis also revealed major distinctions between the two transcription programs. For some of these differences, such as the distinct regulation of genes involved in mitochondrial versus cytoplasmic protein synthesis, the association with distinct phenotypes (e.g., aerobic versus anaerobic rapid growth) is apparent. Other differences, such as those related to cell cycle or amino acid biosynthesis, remain to be elucidated. The former may be connected to different mechanisms of cell cycle regulation pertaining to morphology and/or to different points of cell cycle control exhibited by the two organisms. The latter may reflect the fact that C. albicans lives primarily within a human host, and thus may grow in an environment that readily provides specific subsets of amino acids.
It should be noted that although the number of C. albicans transcription profiles used in this analysis (~250 different arrays) is large, this dataset is probably far from being saturated. Additional differences are likely to be revealed once more data become available. Our comprehensive account of co-regulation in C. albicans provides numerous functional links, as well as important regulatory information, about individual C. albicans ORFs. All the results are available in an interactive format on our Web page at http://barkai-serv.weizmann.ac.il/candida.
Understanding the principles underlying the evolution of gene expression requires systematic comparison of expression data between related organisms. The methods presented in this paper will assist in this challenge. Furthermore, our approach is not limited to the analysis of two sets of expression data, but can be adapted to compare large-scale data of different types, e.g., expression data with protein–protein interaction data or with phenotypic data.
Materials and Methods
Expression data.
Individual experimental datasets were all put into a standardized orf19 gene name format using conversion information provided by A. Nantel, C. D'Enfert, and A. Tsong. Expression data were stored as log2 ratios. Initial analysis identified a significant number of modules that reflected genes with a strong bias for Cy3 versus Cy5 dye labeling. To minimize this effect, dye swap data for the same experimental conditions were averaged whenever possible, resulting in a total of 244 conditions.
Definition of orthologous genes.
We used the Inparanoid software to determine orthologous pairs of genes [59]. Sequence information for C. albicans was based on the orf19 assembly. In the case of multiple genes in a cluster (~5%), we used the one with the highest score, resulting in 3,619 one-to-one ortholog pairs.
Definition of gene sets.
Functional GO categories were downloaded from http://www.geneontology.org. The assignment of genes to the original GO categories was extended to include parent terms, i.e., a gene assigned to a given category was automatically assigned to all the parent categories as well. Only genes classified as orthologous between C. albicans and S. cerevisiae were considered, and C. albicans categorization was inferred from S. cerevisiae orthologs. All GO terms containing at least five orthologs were considered. In the HOCA of GO terms in C. albicans (Figure 6B), this categorization was supplemented with C. albicans–specific GO annotations obtained from the Candida Genome Database (http://www.candidagenome.org). For the analysis shown in Figure 3, we also added gene sets based on promoter sequence elements. For each sequence element (of length 6 and 7), the genes containing the element in their 600-basepair upstream regions were identified for both S. cerevisiae and C. albicans. The Signature Algorithm [33] was applied to distinguish those genes that are mutually co-expressed in each set [10]. The final set associated with each sequence consisted of the union of co-expressed orthologs from each organism.
Co-expression of GO terms.
The extent of co-expression of genes assigned to each GO category was quantified by a normalized t-value. For each organism, pair-wise Pearson correlation coefficients were evaluated for all gene pairs within the category, using all conditions in the dataset. The resulting distribution was compared to a background distribution of 10,000 randomly chosen gene pairs, and a t-statistic was calculated for the two distributions. t-Statistics were calculated for all GO categories, as well as for randomly composed control gene sets of the same size distribution. The t-values shown in the figure are given in terms of the standard deviation of t-values obtained from the random control sets.
DCA clustering.
The algorithm was implemented in Matlab using its standard routine for hierarchical clustering with average linkage. The similarity Sij between genes i and j was defined by the Euclidean distance between the vectors Cik and Cjk containing the Pearson correlations (over all experiments) to all the other genes k, i.e,
For the HOCA, the Pearson correlations were replaced by the topological overlap, defined below. The cluster definition cutoff was given in terms of the fraction of the maximum linkage value. Cutoff values were chosen heuristically: 0.6 for the gene correlation analysis, 0.4 for the GO term connectivity analysis, and 0.3 for the sequence connectivity analysis.
Topological overlap.
Following Ravasz et al. [52], the topological overlap between two nodes i and j in the network was defined as OT (i, j) = Jn(i, j)/[min (ki, kj)], where Jn(i, j) denotes the number of nodes to which both i and j are linked (plus 1 if there is a direct link between i and j), and ki and kj are the total number of links of nodes i and j, respectively.
Enrichment p-values.
Enrichment p-values were calculated using the hypergeometric probability density function. The significance p-value of observing z genes assigned to a given category in a gene set of size N is given by
, where K is the total number of genes assigned to the category and M is the number of genes in the genome. The probability of making a connection between two attributes (GO terms, 6-mers, or 7-mers) is given by
, where n is the number of attributes and nm is the number of representative modules in the dataset (a list of which is given on http://barkai-serv.weizmann.ac.il/candida). Note that this also accounts for multiple hypothesis testing. We imposed a p-value of 0.05 for a network connection corresponding to the following significance cutoff for p0 (in units of −log10): C. albicans: 6-mers: 5.0; 7-mers: 5.6; GO terms: 4.6; S. cerevisiae: 6-mers: 4.8, 7-mers: 5.4; GO terms: 4.5.
Strain construction.
Yeast strain YJB9073 (Figure 5F) was constructed by transforming strain YJB8911 (BWP17 Nop1-CFP) with the PCR amplification product of plasmid pYFP-URA3 [60] and primers F1776 (CAAAAGAAAAAAGAAGAAGAAGAGGATGAGCAAGAAGATGAAGATATTGTAATGGAGGAGGAAGATGATGAGTCTAAAGGTGAAGAATTATT) and R1777 (ATTTAGTCTTGTAT-AACACTATCATATATGTAATATTATTATCGTGTATTAACACAACTGTAAATTATTTGTCTAGAAGGACCACCTTTGATTG), which was designed to insert a C. albicans codon-optimized version of YFP at the C-terminus of orf19.5850. The correct integration product was confirmed by PCR with primers F1791 (TTGCAAGCTGTTGATTTCGAACAC) from the middle of orf19.5850 and R658 (TTTGTACAATTCATCCATACCATG) from the 3′ end of the YFP coding sequence.
Supporting Information
Figure S1 Illustration of the Use of t-Statistics to Evaluate the Extent of Co-Expression of Genes Assigned to a Given Functional Category
From left to right: (1) Based on prior functional annotation (as given by the GO or KEGG database), the corresponding subsets of orthologous genes in S. cerevisiae and C. albicans are selected. (2) Pairwise correlations between these genes are computed in both organisms using the respective set of expression data. (3) The distribution of these correlations are compared to the background distribution corresponding to random subsets of the same size. The significance of co-expression among the functionally associated genes is determined using the t-statistics for the two distributions.
(13 KB PDF)
Click here for additional data file.
Figure S2 Extent of Co-Expression of Genes Assigned to KEGG Pathways in the Two Organisms
Analysis as described for GO terms (c.f. Figure 1A), but using KEGG pathways instead.
(22 KB PDF)
Click here for additional data file.
Figure S3 Robustness of Analysis with Respect to Sub-Sampling of Conditions
The analysis leading to Figure 3A (left panel) was repeated using only a fraction of the expression data (as indicated above each plot). Note that although the average correlations vary slightly (the error bars denote the standard deviations resulting from different sub-samples), they give rise to the same distinct classifications, even when using only 10% of the available expression data.
(15 KB PDF)
Click here for additional data file.
Figure S4 DCA Analysis of Cell-Cycle Genes (Cluster 1)
(2.7 MB JPEG)
Click here for additional data file.
Figure S5 DCA Analysis of Cell-Cycle Genes (Cluster 2)
(2.4 MB JPEG)
Click here for additional data file.
Figure S6 DCA Analysis of Cell-Cycle Genes (Cluster 3)
(2.0 MB JPEG)
Click here for additional data file.
Figure S7 DCA Analysis of Cell-Cycle Genes (Cluster 4)
(2.2 MB JPEG)
Click here for additional data file.
Figure S8 DCA Analysis of Cell-Cycle Genes (Cluster 5)
(2.1 MB JPEG)
Click here for additional data file.
Figure S9 DCA Analysis of Cell-Cycle Genes (Cluster 6)
(2.0 MB JPEG)
Click here for additional data file.
Figure S10 DCA Analysis of Cell-Cycle Genes (Cluster 7)
(2.2 MB JPEG)
Click here for additional data file.
Figure S11 DCA Analysis of Cell-Cycle Genes (Cluster 8)
(2.2 MB JPEG)
Click here for additional data file.
Figure S12 DCA Analysis of Cell-Cycle Genes (Cluster 9)
(1.7 MB JPEG)
Click here for additional data file.
Figure S13 DCA Analysis of Cell-Cycle Genes (Cluster 10)
(2.0 MB JPEG)
Click here for additional data file.
Interactive versions of Figures S4–S13 are available at http://barkai-serv.weizmann.ac.il/candida/html/cc_analysis.html.
We thank Maryam Gerami-Nejad for construction of YFP-tagged strains. We thank the following for providing transcription profiling datasets: R. Bennett, A. Tsong, A. Johnson, M. Lorenz, C. D'enfert, G. Fink, M. Whiteway, A. Nantel, P.D. Rogers, and especially P.D. Rogers, C. Bachewich, U. Oberholzer, E. Bensen, M. McClellan, P. Sudbery, P. Amornrattanapan, D. Davis, D. Harcus, B. Hube, and D. Sanglard, for providing transcription profile datasets prior to publication. We also thank A. Nantel, C. D'enfert, and A. Tsong for providing gene name information that allowed assignment of genes on different arrays to their orf19 gene identities. We thank M. Lapidot and Y. Pilpel for helpful discussions and O. Reiner for comments on the manuscript. This work was supported by National Institutes of Health grants AI50562 (NB) and DE/AI 14666 (JB) and a grant from the Kahn Fund for Systems Biology at the Weizmann Institute of Science (NB).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. JB and NB conceived and designed the experiments. JB collected the experimental data. JI, SB, and NB contributed analysis tools. JI, SB, JB, and NB analyzed the data, and wrote the paper.
Abbreviations
DCAdifferential clustering algorithm
GOGene Ontology
HOCAhigher-order connectivity analysis
ISAiterative sequence algorithm
ORFopen reading frame
PCMpair-wise correlation matrix
YFPyellow fluorescent protein
==== Refs
References
Carroll SB 2000 Endless forms: The evolution of gene regulation and morphological diversity Cell 101 577 580 10892643
True JR Carroll SB 2002 Gene co-option in physiological and morphological evolution Annu Rev Cell Dev Biol 18 53 80 12142278
Wray GA Hahn MW Abouheif E Balhoff JP Pizer M 2003 The evolution of transcriptional regulation in eukaryotes Mol Biol Evol 20 1377 1419 12777501
Abouheif E Wray GA 2002 Evolution of the gene network underlying wing polyphenism in ants Science 297 249 252 12114626
Gompel N Prud'homme B Wittkopp PH Kassner VA Carroll SB 2005 Chance caught on the wing: Cis-regulatory evolution and the origin of pigment patterns in Drosophila
Nature 433 481 487 15690032
Rifkin SA Kim J White KP 2003 Evolution of gene expression in the Drosophila melanogaster subgroup Nat Genet 33 138 144 12548287
Rustici G Mata J Kivinen K Lio P Penkett CJ 2004 Periodic gene expression program of the fission yeast cell cycle Nat Genet 36 809 817 15195092
Alter O Brown PO Botstein D 2003 Generalized singular value decomposition for comparative analysis of genome-scale expression data sets of two different organisms Proc Natl Acad Sci U S A 100 3351 3356 12631705
Stuart JM Segal E Koller K Kim SK 2003 A gene-coexpression network for global discovery of conserved genetic modules Science 302 249 255 12934013
Bergmann S Ihmels J Barkai N 2004 Similarities and differences in genome-wide expression data of six organisms PLoS Biol 2 e9. DOI: 10.1371/journal.pbio.0020009 . 14737187
McCarroll SA Murphy CT Zou S Pletcher SD Chin CS 2004 Comparing genomic expression patterns across species identifies shared transcriptional profile in aging Nat Genet 36 197 204 14730301
Heckman DS Geiser DM Eidell BR Staffer RL Kardos NL 2001 Molecular evidence for the early colonization of land by fungi and plants Science 293 1129 1133 11498589
Berbee ML Taylor JW 2000 Fungal molecular evolution: Gene trees and geologic time McLaughlin DJ McLaughlin E The mycota New York Springer-Verlag 229 246
Jones T Federspiel NA Chibana H Dungan J Kalman S 2004 The diploid genome sequence of Candida albicans
Proc Natl Acad Sci U S A 101 7329 7334 15123810
Cowen LE Nantel A Whiteway MS Thomas DY Tessier DC 2002 Population genomics of drug resistance in experimental populations of Candida albicans
Proc Natl Acad Sci USA 99 9284 9289 12089321
Rogers PD Barker KS 2003 Genome-wide expression profile analysis reveals coordinately regulated genes associated with stepwise acquisition of azole resistance in Candida albicans clinical isolates Antimicrob Agents Chemother 47 1220 1227 12654650
Karababa M Coste AT Rognon B Bille J Sanglard D 2004 Comparison of gene expression profiling between Candida albicans azole-resistant clinical isolates and laboratory strains exposed to drugs inducing multidrug transporters Antimicrob Agents Chemother 48 3064 3079 15273122
Enjalbert B Nantel A Whiteway M 2003 Stress-induced gene expression in Candida albicans: Absence of a general stress response Mol Biol Cell 14 1460 1467 12686601
Tsong AE Miller MG Raisner RM Johnson AD 2003 Evolution of a combinatorial transcriptional circuit: A case study in yeasts Cell 115 389 399 14622594
Bennett RJ Uhl MA Miller MG Johnson AD 2003 Identification and characterization of a Candida albicans mating pheromone Mol Cell Biol 23 8189 8201 14585977
Garcia-Sanchez S Aubert S Iraqui I Janbon G Ghigo JM 2004
Candida albicans biofilms: A developmental state associated with specific and stable gene expression patterns Eukaryot Cell 3 536 545 15075282
Fradin C De Groot P MacCallum D Schaller M Klis F 2005 Granulocytes govern the transcriptional response, morphology, and proliferation of Candida albicans in human blood Mol Microbiol 56 397 415 15813733
Lorenz MC Bender JA Fink GR 2004 Transcriptional response of Candida albicans upon internalization by macrophages Eukaryot Cell 3 1076 1087 15470236
Bensen ES Martin SJ Li M Berman J Davis DA 2004 Transcriptional profiling in C. albicans reveals new adaptive responses to extracellular pH and functions for Rim101p Mol Microbiol 54 1335 1351 15554973
Nantel A Dignard D Bachewich C Harcus D Marcil A 2002 Transcription profiling of C. albicans cells undergoing the yeast to hyphal transition Mol Biol Cell 13 3452 3465 12388749
Lee CM Nantel A Jiang L Whiteway M Shen SH 2004 The serine/threonine protein phosphatase SIT4 modulates yeast-to-hypha morphogenesis and virulence in Candida albicans
Mol Microbiol 51 691 709 14731272
Harcus D Nantel A Marcil A Rigby T Whiteway M 2004 Transcription profiling of cyclic AMP signaling in Candida albicans
Mol Biol Cell 15 4490 4499 15269278
Eisen MB Spellman PT Brown PO Botstein D 1998 Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 95 14863 14868 9843981
Alon U Barkai N Notterman DA Gish K Ybarra S 1999 Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays Proc Natl Acad Sci U S A 96 6745 6750 10359783
Kim SK Lund J Kiraly M Duke K Jiang M 2001 A gene expression map for Caenorhabditis elegans
Science 293 2087 2092 11557892
Ihmels J Bergmann S Barkai N 2004 Defining transcription modules using large-scale gene expression data Bioinformatics 20 1993 2003 15044247
Ashburner M Ball CA Blake JA Botstein D Butler H 2000 Gene ontology: Tool for the unification of biology. The Gene Ontology Consortium Nat Genet 25 25 29 10802651
Ihmels J Friedlander G Bergmann S Sarig O Ziv Y 2002 Revealing modular organization in the yeast transcriptional network Nat Genet 31 370 377 12134151
Chen D Toone WM Mata J Lyne R Burns G 2003 Global transcriptional responses of fission yeast to environmental stress Mol Biol Cell 14 214 229 12529438
Surana U Robitsch H Price C Schuster T Fitch I 1991 The role of CDC28 and cyclins during mitosis in the budding yeast S. cerevisiae
Cell 65 145 161 1849457
Fitch I Dahmann C Surana U Amon A Nasmyth K 1992 Characterization of four B-type cyclin genes of the budding yeast Saccharomyces cerevisiae
Mol Biol Cell 3 805 818 1387566
Richardson H Lew DJ Henze M Sugimoto K Reed SI 1992 Cyclin-B homologs in Saccharomyces cerevisiae function in S phase and in G2 Genes Dev 6 2021 2034 1427070
Schwob E Nasmyth K 1993 CLB5 and CLB6, a new pair of B cyclins involved in DNA replication in Saccharomyces cerevisiae
Genes Dev 7 1160 1175 8319908
Bensen ES Clemente-Blanco A Finley KR Correa-Bordes J Berman J 2005 The mitotic cyclins Clb2p and Clb4p affect morphogenesis in Candida albicans
Mol Biol Cell 16 3387 3400 15888543
Peng X Krauturi RK Miller LD Lin K Jia Y 2005 Identification of cell cycle-regulated genes in fission yeast Mol Biol Cell 16 1026 1042 15616197
Crampin H Finley K Gerami-Nejad M Court H Gale C 2005
Candida albicans hyphae have a Spitzenkörper that is distinct from the polarisome found in yeast and pseudohyphae J Cell Sci In press.
Lew DJ Reed SI 1995 A cell cycle checkpoint monitors cell morphogenesis in budding yeast J Cell Biol 129 739 749 7730408
Verde F Mata J Nurse P 1995 Fission yeast cell morphogenesis: Identification of new genes and analysis of their role during the cell cycle J Cell Biol 131 1529 1538 8522609
Bergmann S Ihmels J Barkai N 2003 Iterative signature algorithm for the analysis of large-scale gene expression data Phys Rev E Stat Nonlin Soft Matter Phys 67 031902 12689096
Uhl MA Biery M Craig N Johnson AD 2003 Haploinsufficiency-based large-scale forward genetic analysis of filamentous growth in the diploid human fungal pathogen C. albicans
Embo J 22 2668 2678 12773383
Huh WK Falvo JV Gerke LC Carroll AS Howson RW 2003 Global analysis of protein localization in budding yeast Nature 425 686 691 14562095
Hughes TR Marton MJ Jones AR Roberts CJ Stoughton R 2000 Functional discovery via a compendium of expression profiles Cell 102 109 126 10929718
Gasch AP Spellman PT Kao CM Carmel-Harel O Eisen MB 2000 Genomic expression programs in the response of yeast cells to environmental changes Mol Biol Cell 11 4241 4257 11102521
Chibana H Magee BB Grindle S Ran Y Scherer S 1998 A physical map of chromosome 7 of Candida albicans
Genetics 149 1739 1752 9691033
Dixon B Rose AH 1966 A specific requirement for biotin in the synthesis of ornithine carbamoyltransferase by yeast Biochem J 99 513 520 5964954
Leiper J Vallance P 1999 Biological significance of endogenous methylarginines that inhibit nitric oxide synthases Cardiovasc Res 43 542 548 10690326
Ravasz E Somera AL Mongru DA Oltvai ZN Barabasi AL 2002 Hierarchical organization of modularity in metabolic networks Science 297 1551 1555 12202830
Kent NA Eibert SM Mellor J 2004 Cbf1p is required for chromatin remodeling at promoter-proximal CACGTG motifs in yeast J Biol Chem 279 27116 27123 15111622
Gasch AP Moses AM Chiang DY Fraser HB Berardini M 2004 Conservation and evolution of cis-regulatory systems in ascomycete fungi PLoS Biol 2 e398. DOI: 10.1371/journal.pbio.0020398 . 15534694
Tanay A Regev A Shamir R 2005 Conservation and evolvability in regulatory networks: The evolution of ribosomal regulation in yeast Proc Natl Acad Sci U S A 102 7203 7208 15883364
Ihmels J Bergmann S Gerami-Nejad M Yanai I McClellan M 2005 Rewiring of the yeast transcriptional network through the evolution of motif usage Science 309 938 940 16081737
Tavazoie S Hughes JD Campbell MJ Cho RJ Church GM 1999 Systematic determination of genetic network architecture Nat Genet 22 281 285 10391217
Daub CO Steuer R Selbig J Kloska S 2004 Estimating mutual information using B-spline functions—An improved similarity measure for analysing gene expression data BMC Bioinformatics 5 118 15339346
Remm M Storm CE Sonnhammer EL 2001 Automatic clustering of orthologs and in-paralogs from pairwise species comparisons J Mol Biol 314 1041 1052 11743721
Gerami-Nejad M Berman J Gale CA 2001 Cassettes for PCR-mediated construction of green, yellow and cyan fluorescent protein fusions in Candida albicans
Yeast 18 859 864 11427968
|
16470937
|
PMC1239936
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Sep 30; 1(3):e39
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010039
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1620578910.1371/journal.pgen.001004105-PLGE-RA-0028R3plge-01-03-14Research ArticleBioinformatics - Computational BiologyEvolutionGenetics/GenomicsGenetics/Population GeneticsGenetics/Functional GenomicsGenetics/Genetics of DiseaseHomo (Human)Positive Selection of a Pre-Expansion CAG Repeat of the Human SCA2 Gene Selection of a Pre-Expansion CAG RepeatYu Fuli 1Sabeti Pardis C 2Hardenbol Paul 3Fu Qing 1Fry Ben 2Lu Xiuhua 1Ghose Sy 1Vega Richard 1Perez Ag 1Pasternak Shiran 1Leal Suzanne M 1Willis Thomas D 3Nelson David L 1Belmont John 1Gibbs Richard A 1*1 Human Genome Sequencing Center, Baylor College of Medicine, Houston, Texas, United States of America
2 Broad Institute of the Massachusetts Institute of Technology, and Harvard University, Cambridge, Massachusetts, United States of America
3 ParAllele Bioscience, South San Francisco, California, United States of America
Abecasis Goncalo EditorUniversity of Michigan, United States of America* To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 26 8 2005 1 3 e4123 2 2005 26 8 2005 Copyright: © 2005 Yu et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.A region of approximately one megabase of human Chromosome 12 shows extensive linkage disequilibrium in Utah residents with ancestry from northern and western Europe. This strikingly large linkage disequilibrium block was analyzed with statistical and experimental methods to determine whether natural selection could be implicated in shaping the current genome structure. Extended Haplotype Homozygosity and Relative Extended Haplotype Homozygosity analyses on this region mapped a core region of the strongest conserved haplotype to the exon 1 of the Spinocerebellar ataxia type 2 gene (SCA2). Direct DNA sequencing of this region of the SCA2 gene revealed a significant association between a pre-expanded allele [(CAG)8CAA(CAG)4CAA(CAG)8] of CAG repeats within exon 1 and the selected haplotype of the SCA2 gene. A significantly negative Tajima's D value (−2.20, p < 0.01) on this site consistently suggested selection on the CAG repeat. This region was also investigated in the three other populations, none of which showed signs of selection. These results suggest that a recent positive selection of the pre-expansion SCA2 CAG repeat has occurred in Utah residents with European ancestry.
Synopsis
Natural selection ultimately acts on the genetic variants existing among human populations. Therefore, there are “footprints” that the selective force has left behind in the human genome. In this study, Yu et al. identified an extremely large region on Chromosome 12 that is under positive selection in Utah residents with European ancestry by characterizing the correlation patterns of genomic variants. Further analyses on this interval suggested that selection centered on one of the many forms of Spinocerebellar ataxia type-2 (SCA2) gene. The selected form was next demonstrated to associate with one short version of the disease-causing CAG repeat in the SCA2 gene. These results suggest that the CAG repeat was positively selected. An abnormally long version of CAGs can cause SCA2, a neurodegenerative disease that severely impairs the abilities of body movement. The authors showed how they unraveled natural selection acting on the SCA2 gene. Their findings might lead to the discovery of the biological functions of this gene and its CAG repeat. This kind of study holds potential to facilitate the finding of common disease genes.
Citation:Yu F, Sabeti PC, Hardenbol P, Fu Q, Fry B, et al (2005) Positive selection of a pre-expansion CAG repeat of the human SCA2 gene. PLoS Genet 1(3): e41.
==== Body
Introduction
The International Haplotype Mapping (HapMap) Project [1,2] has generated a large set of evenly spaced human genomic variation data on samples from four different populations [Utah residents with ancestry from northern and western Europe (CEU); Han Chinese in Beijing, China (CHB); Japanese in Tokyo, Japan (JPT); and Yoruba in Ibadan, Nigera (YRI)]. The marker distribution on the chromosomal scale enables the identification of regions that represent locus-specific statistical deviations from overall genomic patterns. When effects from technical or sampling factors are properly considered, then regions of biological interest are revealed. With a genome-wide dense marker set, effects from ascertainment bias in marker selection are minimized, and other driving forces (e.g. drift, population expansion, migration, and non-random mating) are controlled, because they act upon all loci across the entire genome in a similar and predictable fashion [3–5]. In this study, we have demonstrated the effects of selection on a region surrounding a single allele. First, we noted a region of extensive linkage disequilibrium (LD) containing multiple single nucleotide polymorphisms (SNPs) suggesting a candidate region for natural selection on Chromosome 12 in CEU. We next applied more rigorous Extended Haplotype Homozygosity (EHH) analyses [6] to measure allelic-specific LD in this region of interest and to identify the selected genes and alleles. In order to control for recombination rate variation across the genome, the Relative EHH (REHH) test was applied, in which the different alleles in the same region serve as internal controls to normalize recombination rate variation.
A principal allele was identified within one particular haplotype spanning exon 1 of the spinocerebellar ataxia type 2 (SCA2) gene. The trinucleotide repeat expansion in this exon causes spinocerebellar ataxia [7], a neurodegenerative disorder typically with severe olivo-ponto-cerebellar atrophy [8,9], which affects the normal sensory/motor controlling functions. Previous studies of the triplet allele frequency distribution had shown that SCA2 was unusual, in that the pre-expansion allele accounted for more than 90% of sampled chromosomes [10], whereas much higher triplet polymorphism rates are commonly seen in other genes that underlie neurodegenerative diseases [11]. Our results provide evidence that positive selection has favored this pre-expansion CAG repeat in the human SCA2 gene, and that it is responsible for its overall predominance in CEU in pre-disease versions of the gene.
Results
Identification of One Positively Selected Region on Human Chromosome 12
We carried out large scale genotyping on human Chromosome 12 using Molecular Inversion Probe technology [12,13]. Across the four populations, on average 47,452 SNP markers were successfully genotyped and deposited in the HapMap database with a completeness, repeatability, and trio accuracy of 98.9%, 99.4%, and 99.5%, respectively. Among these SNPs, ~70% of markers have a minor allele frequency (MAF) greater than 0.05 in at least one population. The marker density on average is one SNP per ~2.8 kb, which is sufficient to enable the understanding of detailed haplotype structure of human genome.
One effect of a recent selective sweep is a significantly large interval with strong LD around the selected site [5,14,15]. We used this effect as one criterion to detect selection across the entire Chromosome 12. In order to identify the large region with increased LD, we first constructed haplotype blocks using the LD-based empirical block definition proposed by Gabriel et al. [16]. The haplotype block size distribution of Chromosome 12 in CEU has a mean size of 26 kb, standard deviation of 43 kb, and median size of 13 kb (Figure 1). The largest block spans 987 kb however, and has 138 markers with MAF > 0.05. This is a striking outlier in the block size distribution. All studied markers in this block are in very strong LD with each other, as well as with markers in two adjacent centromeric blocks (Figure 2A), with averages of pair-wise |D′| and r2 of 0.91 and 0.51, respectively, that together extend the size of the overall region (110,230,654–111,393,524) in high LD to ~1.2 Mb. There are 168 SNP markers with MAF > 0.05 in this interval. The common haplotypes (> 1%, more than 2 chromosomes observed) that were inferred using Haploview [17] in the large block account for 75% of all the haplotypes, with the most common present at 30% (Figure 2A).
Figure 1 Haplotype Block Size Distribution of Human Chromosome 12 in CEU
The haplotype blocks were defined by D′ confidence interval [16] and binned according to their sizes. The inset provides the ungrouped distribution of the haplotype blocks larger than 100 kb.
Figure 2 LD Patterns and Haplotypes of the Largest Block on Human Chromosome 12 in CEU (A), CHB (B), JPT (C), and YRI (D)
Pair-wise LD of common SNPs (MAF > 0.05) is expressed as D′, with red indicating strong LD. The figure also shows the haplotype blocks of 1 Mb intervals in both upstream and downstream of the largest block. Schematic position of the SCA2 gene is shown.
The global LD patterns of this region in three additional HapMap data sets (CHB, JPT, and YRI) were found to be similar, but at fine-scale, different to CEU (Figures 2B, 2C, and 2D). Specifically, there are more blocks that are shorter in length, and the LD measured by |D′| and r2 is weaker in the non-CEU samples (Table 1), reflecting the expected existence of more haplotype block breakpoints and historical recombination events in CHB, JPT, and YRI in this interval. This implies that the underlying haplotype structures or the evolutionary forces in this region are different in CEU when compared with the other three populations.
Table 1 Fine-Scale Haplotype Structures and LD of the ~1.2 Mb Region in Four Populations
A possible explanation for this strong LD pattern was the presence of a large, frequent chromosomal inversion specific to CEU, as a similar structure involving a 900 kb inverted haplotype at 17q21.31 has been found in 20% of Europeans [18]. We eliminated this possibility on Chromosome 12 using PCR with primers that spanned the LD block boundary regions (111,393,230–111,397,947 and 111,428,321–111,437,985) with the possible breakpoints in 30 CEU trios (father-mother-child combinations, Table S1). We were able to predict the regions with the potential breakpoints because of the sharpness of the LD block boundaries. Because both the entire ~4 kb and the ~10 kb regions were investigated with tiled primer pairs that were designed to amplify the reference genomic sequences, at least one PCR reaction would fail to amplify any fragment in a large number of the CEU samples if an inversion breakpoint existed within this region,. In contrast, the expected fragment would be amplified in the other CEU samples without the hypothetical inversion. The amplification results revealed the expected normal fragments but no inversion breakpoints (unpublished data).
Application of EHH and Relative EHH Analyses to Map the Causative Locus
We next sought to identify which of the multiple genes within this 1.2 Mb region (Table S2), and their positively selected alleles, might be responsible for this large region of high LD. We applied the EHH approach proposed by Sabeti et al. [6] to compare the rates of allelic specific LD decay. The EHH test exploits the assumption that under neutral evolution theory, LD of common alleles tends to be less extensive than for rare ones, due to the increased frequency of recombination and mutation as a function of the time needed for these alleles to become enriched in the population. In contrast, regions under positive selection have frequent alleles, existing on long range LD backgrounds. We also applied the REHH test, which corrects for local variation in recombination rate by comparing the EHHs of different core haplotypes present at a locus.
We used a sliding window approach to scan the entire 1.2 Mb interval and identify a “causative core region” (Materials and Methods). This was defined as the common core haplotype (frequency > 0.3), for which the “haplotype-specific homozogosity” was elevated when tested against distant markers, and after normalization with the recombination rate variation. Specifically, the cutoff used is that the REHH values must be greater than 2 with long-range markers, radiating to distances greater than 200 kb from the core site. This cutoff has previously been demonstrated to establish a 95th percentile of significance among 5,000 simulated data sets [6]. This test identified a core region spanning seven common SNPs (Figures 3A and 3B). Among the seven SNPs, three (rs3809274, rs1544396, and rs9300319) are in the 5′ upstream of the SCA2 gene, one (rs695871) in the 5′ coding region, while the other three (rs593226, rs616513, and rs653178) are within intron 1 (Figure 4). These data therefore led to the hypothesis that evolutionary pressure acting on the SCA2 gene is primarily responsible for the extensive LD in the 1.2 Mb region.
Figure 3 Mapping the Locus that is Under Positive Selection to the SCA2 Gene Using the EHH Approach
The mapped core region consists of 7 markers (rs593226, rs616513, rs653178, rs695871, rs3809274, rs1544396, and rs9300319). (A) The relative gene positions, EHH × distance and REHH × distance plots. (B) Four core haplotypes [TCGGGAT (39%), TCAGGAT (22%), CAACCGC (20%), and CCAGGAT (7%)]. The common SNPs are highlighted in yellow. The ancestral alleles predicted as the chimp alleles were shown on the top. (C) Haplotype bifurcation plots of the four core haplotypes.
Figure 4 Schematic of the SCA2 gene, (CAG)n, and the 7 SNPs (rs593226, rs616513, rs653178, rs695871, rs3809274, rs1544396, and rs9300319) in the Core Region
The REHH test identified one core haplotype (CH-1: TCGGGAT, frequency 39%) with elevated values over long distances (Figure 3A). CH-1 had clearly shown strikingly slower EHH decay, when compared to other haplotypes, even at a very long range. We expanded the interval by 1 Mb both upstream and downstream in an effort to detect the extended boundaries. The high EHH diminished abruptly at ~1 Mb and ~600 kb distances from the core on the proximal and distal ends, respectively (Figure 3A). At further distances from the core region, the difference of EHH values between different core haplotypes was not significant. We next used the bifurcation patterns illustrated in the diagrams to view the preserved long range homozogosity specific to each core haplotype. Each diagram is for one core haplotype, with the black dot representing the core region and each node representing one marker. The thickness of the branched lines reflects the number of samples carrying a specific haplotype. The visualized bifurcation diagram illustrates an extended predominance of one marker lineage for CH-1 (Figure 3C). To assess the significance of CH-1′s REHH values, we calculated REHH at ~0.25 centiMorgans (cM) distance on both sides from a core, for all the possible cores on Chromosome 12. We chose 0.25 cM distances for testing because it has been argued that 0.25 cM has sufficient power to detect recent (~10,000 years) selection marks [6]. The REHH values of CH-1 exceeded the 95th percentile when REHH was plotted against allele frequency (Figure 5). The p-values for REHH calculated at CH-1′s telomeric and centromeric boundaries are 0.003 and 0.0002, respectively, when estimated by comparing with the entire Chromosome 12 distribution. Similar results were obtained when evaluated against 1,000 simulated loci [p = 0.0009 and p = 0.001 at 1 Mb telomeric and 400 kb centromeric distances from the core region. (Figure S2)]. Moreover, this core region has been recently identified as a highly significant outlier in the whole-genome REHH distribution (P. C. Sabeti, unpublished data).
Figure 5 REHH × Frequency Distribution of CEU Chromosome 12
REHH was calculated at 0.25 cM distances on both sides for all possible core haplotypes from CEU Chromosome 12 and represented with blue dots and given with 50th, 75th, 90th, and 95th percentiles. REHH values of CH-1 were represented with brown dots.
The REHH analyses centered on this core region were also applied to CHB, JPT, and YRI. None of the frequent core haplotypes showed high enough REHH to satisfy our criteria (Figure S1). The results suggested that the selection associated with CH-1 is specific to CEU. Furthermore, the observation that the allele frequencies of the common SNPs in this region in CEU are quite different from the other three populations (unpublished data) also supported the existence of CEU-specific selection pressure on this region.
Positive Selection of a Pre-Expansion CAG Repeat of the SCA2 Gene
The reason that the SCA2 gene would be under strong positive selection in CEU is not obvious. An expanded trinucleotide repeat expansion that codes for polyglutamine in exon 1 has been characterized as the causative mutation of progressive cerebellar ataxia. This is the most common autosomal dominant cerebellar ataxia in diverse ethnic and geographical populations [19]. The normal alleles of the CAG repeat vary in length from 14–31 triplets, and frequently include one or more CAA interruptions, whereas the disease alleles have more than 31 CAG triplets without CAA interruption. The severity of the disease and the age of onset are negatively correlated with the length of CAG repeat. Because exon 1 is within the “core” of positive selection for the 1.2 Mb region, and because it contains the causative mutation site for this locus, we hypothesized that a specific allele of the CAG trinucleotide repeat was associated with the core haplotype and potentially played an important role in selection.
In order to test this hypothesis, we both genotyped and sequenced the CAG repeat in samples from CEU, CHB, JPT, and YRI. There are 15 different SCA2 CAG repeat alleles found in these samples (Table 2). More alleles were detected in YRI than in CEU, CHB, and JPT, and the common alleles (a-4 and a-5) found in non-Africans represent a subset of the common ones (a-4, a-5, and a-10) in YRI. Specifically, a-4 and a-5 are present at relatively high frequency in all four populations, and together account for 49%, 90%, and 100% of chromosomes in YRI, CEU, and CHB / JPT, respectively. The allele a-10 was a unique common (37%) allele found only in YRI. This observation is consistent with the known historical bottleneck for out-of-Africa populations. Another important observation is that except for two common alleles (a-4 and a-5), none of the rare alleles found in CEU are the same as the rare alleles in YRI, and almost all the rare alleles present in either CEU or YRI can be derived from the common ones with a point mutation or a slight CAG slippage. These results suggested that the rare alleles in CEU and YRI emerged after the separation of the major continental populations.
Table 2 Sequencing Results of (CAG)n Repeat in the SCA2 Gene in Four Different Populations
The sequenced alleles were then phased with the 7 SNPs in the core region in CEU. CH-1 is completely associated with a-5 of the CAG repeat. The a-5 allele is associated with CH-1 about 40% of the time. A chi-square test confirmed the significant correlation between CH-1 and a-5 of CAG repeat (df = 1, chi-square = 20, p < 0.001) in CEU. We performed a pair-wise LD test between these 7 SNP markers and the CAG repeat. The results illustrated that a-5 of the CAG repeat had a strong association with each of the other 7 SNPs (Figure 6).
Figure 6 LD of CAG repeat in SCA2 gene with the 7 Adjacent SNPs (rs593226, rs616513, rs653178, rs695871, rs3809274, rs1544396, and rs9300319)
Different alleles of CAG repeat were recoded as such: 1 = (CAG)8CAA(CAG)4CAA(CAG)8 allele; 2 = all the other alleles; 0 = failed.
The more traditional approaches, including Tajima's D-test, and Fu and Li's D*-test and F*-test, were performed to detect the selection on CAG repeat (Materials and Methods, Table S3). These test polymorphic sites and the frequency spectrum of the alleles to examine deviations from the expectations of neutral evolutionary theory. They each emphasize different characteristics: Tajima's D is sensitive to the presence of rare alleles, whereas Fu and Li's D* and F* are sensitive to singletons. These approaches are informative about the evolutionary forces in that positive selection is implied by negative values. The significantly negative Tajima's D value (p < 0.01) was observed for CEU, which agreed with our hypothesis that this locus has undergone a positive selection.
Discussion
This study shows that the (CAG)8CAA(CAG)4CAA(CAG)8 allele (a-5) of the CAG repeat in exon 1 of the SCA2 is significantly associated with a haplotype (CH-1) that has been detected to be under recent positive selection in CEU. As a result, a region of nearly one megabase of Chromosome 12 around this locus shows extensive LD. This is a dramatic, recent evolutionary pattern which appears to be restricted to Europeans.
Other alleles tested by the EHH approach served as internal controls to eliminate the possibilities of reduction of either the mutation or recombination rates in this physical region of chromatin or as a function of different alleles being accountable for the long LD. For example, the TCAGGAT allele is only one base different from CH-1 in the core region and only a few bases different across the entire ~1 Mb region (Figure 2A). However, its EHH decayed very quickly (Figure 3A), and showed no significance when plotted REHH values against their allele frequencies (Figure S2). These results suggest that no complication of local recombination rate variations has led to the predominance of the CH-1 haplotype. Additionally, the recombination rates (sex-average = 0.52, female = 0.87, and male = −0.11) estimated for this region using the natural logarithm of the ratio of the map distances cM × 106/Mb based on deCODE and Marshfield markers do not show unusual statistical distributions, and would not account for the extremity of LD (J. Belmont et al., unpublished data).
The SCA2 gene has an unusually low repeat variance relative to the other disease-associated coding triplet repeats [11]. The allele distribution is highly skewed towards a-4 and a-5 in CEU. Although the mutation rate has been suggested to be relatively low in this locus due to being stabilized by CAA interruptions [10], it cannot fully explain this low level of variance, because a comparable number of rare alleles were found in SCA2 as in SCA1 [10]. In addition, the rare alleles could easily arise from the common alleles, which implies that selective pressure could act to maintain the predominance of only a small number of alleles.
The population-specific allele spectra imply the action of other driving forces. In our study, we found that a-4 and a-5 accounted for 100% of chromosomes in CHB and JPT samples. This suggests an historical population bottleneck that could also have contributed to the formation of the 1 Mb LD observed in CEU. The chromosome-wide distribution pattern demonstrated, however, that the LD is an “outlier”, and thus it is unlikely that such factors can solely account for it. For CEU, we propose that a-4 and a-5 migrated from Africa to Europe with a-5 at much lower frequency than that found in modern Europeans, whereas the recent selective advantage on a-5 enriched it in the population. The adjacent region hitchhiked with this allele and reached high frequency quickly, therefore, the long-range LD was preserved in CEU.
The possible functional mechanisms whereby positive selection acted on the a-5 allele of the SCA2 gene in the recent human population history are unclear. It seems unlikely that the total number of glutamine residues plays a role (a-4 and a-5 each encode 22 Gln residues), however, differences in the number of uninterrupted CAG repeats at the mRNA level could alter normal function through changes in mRNA folding and stability [20] or association with RNA binding factors. The allele a-5 shows a very low likelihood of expansion to the disease state [21], but given the late age of onset and low prevalence of the disease, it seems unlikely that disease predisposition could be directly related to the selection at this locus.
The SCA2 gene product's normal function is unknown, although it may play a role in regulated cell death [22,23], and changes in this function could clearly be under selective constraint. Recent analysis of a C. elegans homologue suggests a role in translational control in the germline, another potential function under strong selective constraints [24]. Recognition of the role of selection at this locus will stimulate further investigation of the mechanism through functional studies.
It remains possible that other linked functional alterations in SCA2 or in nearby genes on the CH-1 haplotype background were necessary for selection on a-5 or even the primary target of natural selection, with the coding triplet instead hitchhiking to high frequency.
Based on our phased results, a-5 also associates with core haplotypes other than CH-1, which do not show significantly high REHH. Other polymorphisms specific to the CH-1 allele are possibly important elements for selection. One model is that both a-5 and other unknown genetic variants on the CH-1 background each contribute modestly to the unidentified biological function, and are necessary to form a specific combination in order to confer a selective advantage. Indeed, two coding SNPs (rs695871 and rs695872) are found within 200 bp of the CAG repeat in exon 1. Our data demonstrated that a-5 is significantly associated with the rs695871 G allele, which codes for Val versus Leu. In addition, as CH-1 spans a large region (~70 kb), including the intergenic sequences between SCA2 and BRAP, and the 5′-UTR and the intronic sequences of SCA2, the polymorphisms that might induce alternative splice sites and that might regulate differentiated expression levels of SCA2 or the adjacent BRAP gene are potential candidates as well and need to be further investigated. Other genes in this ~1.2 Mb interval cannot be completely excluded as being the targets of selection even though they were not detected by the REHH analysis. For example, the ALDH2 gene has been suspected to be selected for its hypothetical functions in resistance to endemic disease in east Asia [25]. Nevertheless, the SCA2 is in the center of the mapped window and remains the strongest candidate gene for selection.
The uncertainty of the precise biochemical mechanism for the selection illustrates the power of the statistical genetic methods used for the identification of a “biological signal” from this locus. We expect other genomic regions to be identified in this way, and eventually to correlate the results of this kind of study with our growing understanding of biological processes.
Materials and Methods
Large scale SNP genotyping using Molecular Inversion Probe technology.
Our genotyping effort was carried out with Molecular Inversion Probe chemistry [12,13]. Both 2-dye and 4-dye labeling protocols for microarray based detection were used. The SNPs were allocated by the International HapMap Consortium. Assays were designed with even marker spacing and genotyped 30 trios (consisted of parents and a child) of CEU, 30 trios of YRI, 45 unrelated individuals of CHB and 45 unrelated individuals of JPT designated for the HapMap project. The data have been submitted to www.hapmap.org.
Haplotype block definition.
NCBI build 34 and HapMap public release #16 were used as references throughout this study. The pair-wise D′ was calculated, and blocks were defined using the D′ confidence interval approach [16]. The largest block on human Chromosome 12 maps to 12q24.12 −13, and the physical coordinates are from 110,405,839–111,393.524. The detailed blocks and their underlying haplotype structure were visualized using Haploview 3.0 [17].
CAG repeat genotyping and sequencing.
For genotyping assay, PCR amplification was performed using a pair of primers, SCA-A and SCA-B, (F, 5′- GGGCCCCTCACCATGTCG-3′; R, 5′- CGGGCTTGCGGACATTGG 3′) as previously described [7], in which SCA-A was 5′ end labeled with either FAM or TET. Twenty pmole each of primers were added to 25 ng of human DNA with Invitrogen's 2× multiplex mix. After an initial denaturation at 95 °C for 5 min, 36 cycles were repeated with a denaturation at 96 °C for 1.5 min, an annealing temperature of 62 °C for 30 s, an extension at 72 °C for 1.5 min, and a final extension of 5 min at 72 °C. Microsatellite genotyping was carried out by using Applied Biosystems 3730 sequencer, and the data was analyzed by using Genemapper software version 3.5 (Applied Biosystems, Foster City, California, USA). The size standard used to analyze the data was GS500 (-250LIZ). Multiple runs were analyzed for each patient and microsatellite polymorphisms were confirmed by pedigree checking.
In the sequencing assay, SCA-A and -B were tailed with the universal sequencing primers (Forward, 5′-CTCGTGTAAAACGACGGCCAGT-3′; Reverse, 5′-CTGCTCAGGAAACAGCTATGAC-3′). After PCR products were purified with Exo-SAP, sequencing reactions were carried out with both SCA-A/B and the universal primers using standard BigDye V3.1 protocol. Traces were manually analyzed using the Sequencher program.
The samples from CEU were both genotyped and sequenced. The genotyping peak lengths completely agreed with their corresponding (CAG)n sequences (unpublished data).
Tajima's D, Fu and Li's D* and F* tests.
We sequenced 130 bases (according to the reference genomic sequence) centered on CAG repeat in SCA2 as described above. All the polymorphisms identified were in the CAG repeats, including the CAG copy number changes that were recoded as nucleotide polymorphisms. The numbers of polymorphic sites discovered are 31, 1, 1, and 27 in CEU, CHB, JPT and YRI, respectively. We performed Tajima's D [26], and Fu and Li's D* and F* tests [27] on (CAG)n using DnaSP 4.0 program [28]. The statistical significance was obtained by testing the confidence limits of the statistics (two tailed test).
Haplotype reconstruction using HAPLORE and PHASE 2.0 programs and EHH analyses.
For the 30 CEU trios and 30 YRI trios, haplotypes were first constructed from their SNP data using the logic rules implemented in HAPLORE program [29]. PHASE 2.0 [30,31] was next used to infer the haplotypes on some markers that cannot construct unambiguous haplotypes based on the offsprings' information. For the unrelated samples from CHB and JPT, PHASE 2.0 was directly applied to infer the haplotypes.
The EHH is defined as “the probability that two randomly chosen chromosomes carrying a tested core haplotype are homozygous at all SNPs for the entire interval from the core region to the distance x.” The REHH is “the ratio of the EHH on the tested core haplotype compared with the EHH of the grouped set of core haplotypes at the region not including the core haplotype tested.”
By definition of EHH analysis, the core region needs to have almost no recombination event. Since most of the nearby markers in the 1.2 Mb interval are in strong LD with high |D′| in CEU, it allowed us to use a 4-marker sliding window as our core region with 2-marker overlap between adjacent windows to scan the entire region. The “Sweep” program (P. C. Sabeti et al., in preparation) was used for detailed EHH and REHH analysis on the identified core region.
REHH significance estimation.
We tested the significance of REHH using two comparison datasets, empirical data from Chromosome 12 of the HapMap project Release 16 and simulations. For simulations, we generated 1000 loci of 1 MB length, calibrated to provide data consistent with a variety of measures of empirical data (i.e., FST, heterozygosity, minor allele frequency distribution), and using a set of model parameters (i.e., demography, recombination rate) in accordance with current estimates (S. Schaffner et al., unpublished data).
For both comparison data sets, we placed haplotypes into 20 bins based on their frequency. We compared the REHH for each common haplotype at SCA2 to all equally frequent haplotypes from the simulations. We obtained p-values by log-transforming the REHH in the bin to achieve normality, and calculating the mean and standard deviation. We carried out analysis using the Sweep software program (P. C. Sabeti et al., unpublished data).
Supporting Information
Figure S1 REHH × Distance Plots in the Other Three Populations
(A) CHB.
(B) JPT.
(C) YRI.
(34 KB PDF)
Click here for additional data file.
Figure S2 REHH by Frequency Plot
The REHH is plotted against the core haplotype frequency at ~1 Mb telomeric (A) and ~400 kb centromeric (B) to the core region.
(1.2 MB PDF)
Click here for additional data file.
Table S1 PCR to Test Inversion
(134 KB DOC)
Click here for additional data file.
Table S2 Gene List in the Largest Haplotype Block in Human Chromosome 12
(33 KB DOC)
Click here for additional data file.
Table S3 Tests of Selection on (CAG)n of SCA2
(31 KB DOC)
Click here for additional data file.
Accession Numbers
The Entrez Gene (http://www.ncbi.nlm.nih.gov/gquery/gquery.fcgi) accession numbers are ALDH2 (217), BRAP (8315), and SCA2 (6311).
We thank Steve Schaffner for use of simulations program prior to publication. We thank David Reich for critical reading of this manuscript. We also greatly appreciate the two anonymous reviewers for their constructive comments on the manuscript. This work was funded by the HapMap grant from National Human Genome Research Institute (1U01 HG2755).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. FY, PCS, JB, and RAG conceived and designed the experiments. FY, QF, XL, SG, RV, and AP performed the experiments. FY, PCS, PH, SP, and SML analyzed the data. PCS, PH, BF, and TDW contributed reagents/materials/analysis tools. FY, DLN, JB, and RAG wrote the paper.
A previous version of this article appeared as an Early Online Release on August 4, 2005 (DOI: 10.1371/journal.pgen.0010041.eor).
Abbreviations
CEUUtah residents with ancestry from northern and western Europe
CHBHan Chinese in Beijing, China
cMcentiMorgans
EHHextended haplotype homozygosity
HapMapThe International Haplotype Mapping Project
JPTJapanese in Tokyo, Japan
LDlinkage disequilibrium
MAFminor allele frequency
REHHrelative extended haplotype homozygosity
SNPsingle nucleotide polymorphism
SCA2Spinocerebellar ataxia type 2
YRIYoruba in Ibadan, Nigeria
==== Refs
References
The International HapMap Consortium 2004 Integrating ethics and science in the International HapMap Project Nat Rev Genet 5 467 475 15153999
The International HapMap Consortium 2003 The International HapMap Project Nature 426 789 796 14685227
Akey JM Zhang G Zhang K Jin L Shriver MD 2002 Interrogating a high-density SNP map for signatures of natural selection Genome research 12 1805 1814 12466284
Akey JM Eberle MA Rieder MJ Carlson CS Shriver MD 2004 Population history and natural selection shape patterns of genetic variation in 132 genes PLoS Biology 2 1591 1599
Bamshad M Wooding SP 2003 Signatures of natural selection in the human genome Nat Rev Genet 4 99 111 12560807
Sabeti PC Reich DE Higgins JM Levine HZ Richter DJ 2002 Detecting recent positive selection in the human genome from haplotype structure Nature 419 832 837 12397357
Pulst SM Nechiporuk A Nechiporuk T Gispert S Chen X-N 1996 Moderate expansion of a normally biallelic trinucleotide repeat in spinocerebella ataxia type 2 Nature Genetics 14 269 276 8896555
Giuffrida S Saponara R Restivo DA Trovato Salinaro A Tomarchio L 1999 Supratentorial atrophy in spinocerebellar ataxia type 2: MRI study of 20 patients J Neurol 246 383 388 10399871
Ihara T Sasaki H Wakisaka A Takada A Yoshiki T 1994 Genetic heterogeneity of dominantly inherited olivopontocerebellar atrophy (OPCA) in the Japanese: Linkage study of two pedigrees and evidence for the disease locus on chromosome 12q (SCA2) Jpn J Hum Genet 39 305 313 7841441
Sobczak K Krzyzosiak WJ 2004 Patterns of CAG repeat interruptions in SCA1 and SCA2 genes in relation to repeat instability Human Mutation 24 236 247 15300851
Andres AM Lao O Soldevila M Calafell F Bertranpetit J 2002 Dynamics of CAG repeat loci revealed by the analysis of their variability Human Mutation 21 61 70
Hardenbol P Baner J Jain M Nilsson M Namsaraev EA 2003 Multiplexed genotyping with sequence-tagged molecular inversion probes Nat Biotechnol 21 673 678 12730666
Hardenbol P Yu F Belmont J Mackenzie J Bruckner C 2005 Highly multiplexed molecular inversion probe genotyping: Over 10,000 targeted SNPs genotyped in a single tube assay Genome Res 15 269 275 15687290
Olson S 2002 Population genetics. Seeking the signs of selection Science 298 1324 1325 12434033
Shlotterer C 2003 Hitchhiking mapping - functional genomics from the population genetics perspective Trends in Genetics 19 32 38 12493246
Gabriel SB Schaffner SF Nguyen H Moore JM Roy J 2002 The structure of haplotype blocks in the human genome Science 296 2225 2229 12029063
Barrett JC Fry B Maller J Daly MJ 2005 Haploview: Analysis and visualization of LD and haplotype maps Bioinformatics 21 263 265 15297300
Stefansson H Helgason A Thorleifsson G Steinthorsdottir V Masson G 2005 A common inversion under selection in Europeans Nat Genet 37 129 137 15654335
Klockgether T 2000 Recent advances in degenerative ataxias Current Opinion in Neurology 13 451 455 10970064
Sobczak K Krzyzosiak WJ 2005 CAG repeats containing CAA interruptions form branched hairpin structures in spinocerebellar ataxia type 2 transcripts J Biol Chem 280 3898 3910 15533937
Choudhry S Mukerji M Srivastava AK Jain S Brahmachari SK 2001 CAG repeat instability at SCA2 locus: Anchoring CAA interruptions and linked single nucleotide polymorphisms Human Molecular Genetics 10 2437 2446 11689490
Huynh DP Yang H-T Vakharia H Nguyen D Pulst SM 2003 Expansion of the polyQ repeat in ataxin-2 alters its Golgi localization, disrupts the Golgi complex and causes cell death Human Molecular Genetics 12 1485 1496 12812977
Wiedemeyer R Westermann F Wittke I Nowock J Schwab M 2003 Ataxin-2 promotes apoptosis of human neuroblastoma cells Oncogene 22 401 411 12545161
Ciosk R DePalma M Priess JR 2004 ATX-2, the C. elegans ortholog of ataxin-2, functions in translational regulation in the germline Development 131 4831 4841 15342467
Oota H Pakstis AJ Bonne-Tamir B Goldman D Grigorenko E 2004 The evolution and population genetics of the ALDH2 locus: Random genetic drift, selection, and low levels of recombination Ann Hum Genet 68 93 109 15008789
Tajima F 1989 Statistical method for testing the neutral mutation hypothesis by DNA polymorphism Genetics 123 585 595 2513255
Fu YX Li WH 1993 Statistical tests of neutrality of mutations Genetics 133 693 709 8454210
Rozas J Sanchez-DelBarrio JC Messeguer X Rozas R 2003 DnaSP, DNA polymorphism analyses by the coalescent and other methods Bioinformatics 19 2496 2497 14668244
Zhang K Sun F Zhao H 2005 HAPLORE: A program for haplotype reconstruction in general pedigrees without recombination Bioinformatics 21 90 103 15231536
Stephens M Smith NJ Donnelly P 2001 A new statistical method for haplotype reconstruction from population data Am J Hum Genet 68 978 989 11254454
Stephens M Donnelly P 2003 A comparison of bayesian methods for haplotype reconstruction from population genotype data Am J Hum Genet 73 1162 1169 14574645
|
16205789
|
PMC1239938
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Sep 30; 1(3):e41
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010041
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1620579010.1371/journal.pgen.001004205-PLGE-RA-0142R2plge-01-03-13Research ArticleEvolutionNeurosciencePsychologyGenetics/Complex TraitsHomo (Human)
AVPR1a and SLC6A4 Gene Polymorphisms Are Associated with Creative Dance Performance AVPR1a and
SLC6A4 and Creative Performance
Bachner-Melman Rachel 1Dina Christian 2Zohar Ada H 3Constantini Naama 4Lerer Elad 5Hoch Sarah 5Sella Sarah 5Nemanov Lubov 5Gritsenko Inga 5Lichtenberg Pesach 5Granot Roni 6Ebstein Richard P 15*1 Department of Psychology, Mount Scopus, Hebrew University, Jerusalem, Israel
2 Génétique Maladies Multifactorielles—Institut de Biologie de Lille, Lille, France
3 Psychology, Behavioral Sciences, Ruppin Academic Center, Emek Hefer, Israel
4 Israeli Olympic Medical Committee and Medical Faculty, Tel Aviv University, Te Aviv, Israel
5 Sarah Herzog Memorial Hospital and Hebrew University, Jerusalem, Israel
6 Musicology Department, Hebrew University, Jerusalem, Israel
Flint Jonathan EditorUniversity of Oxford, United Kingdom* To whom correspondence should be addressed. E-mail: [email protected] 2005 30 9 2005 1 3 e4227 6 2005 26 8 2005 Copyright: © 2005 Bachner-Melman et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Dancing, which is integrally related to music, likely has its origins close to the birth of Homo sapiens, and throughout our history, dancing has been universally practiced in all societies. We hypothesized that there are differences among individuals in aptitude, propensity, and need for dancing that may partially be based on differences in common genetic polymorphisms. Identifying such differences may lead to an understanding of the neurobiological basis of one of mankind's most universal and appealing behavioral traits—dancing. In the current study, 85 current performing dancers and their parents were genotyped for the serotonin transporter (SLC6A4: promoter region HTTLPR and intron 2 VNTR) and the arginine vasopressin receptor 1a (AVPR1a: promoter microsatellites RS1 and RS3). We also genotyped 91 competitive athletes and a group of nondancers/nonathletes (n = 872 subjects from 414 families). Dancers scored higher on the Tellegen Absorption Scale, a questionnaire that correlates positively with spirituality and altered states of consciousness, as well as the Reward Dependence factor in Cloninger's Tridimensional Personality Questionnaire, a measure of need for social contact and openness to communication. Highly significant differences in AVPR1a haplotype frequencies (RS1 and RS3), especially when conditional on both SLC6A4 polymorphisms (HTTLPR and VNTR), were observed between dancers and athletes using the UNPHASED program package (Cocaphase: likelihood ratio test [LRS] = 89.23, p = 0.000044). Similar results were obtained when dancers were compared to nondancers/nonathletes (Cocaphase: LRS = 92.76, p = 0.000024). These results were confirmed using a robust family-based test (Tdtphase: LRS = 46.64, p = 0.010). Association was also observed between Tellegen Absorption Scale scores and AVPR1a (Qtdtphase: global chi-square = 26.53, p = 0.047), SLC6A4 haplotypes (Qtdtphase: chi-square = 2.363, p = 0.018), and AVPR1a conditional on SCL6A4 (Tdtphase: LRS = 250.44, p = 0.011). Similarly, significant association was observed between Tridimensional Personality Questionnaire Reward Dependence scores and AVPR1a RS1 (chi-square = 20.16, p = 0.01). Two-locus analysis (RS1 and RS3 conditional on HTTLPR and VNTR) was highly significant (LRS = 162.95, p = 0.001). Promoter repeat regions in the AVPR1a gene have been robustly demonstrated to play a role in molding a range of social behaviors in many vertebrates and, more recently, in humans. Additionally, serotonergic neurotransmission in some human studies appears to mediate human religious and spiritual experiences. We therefore hypothesize that the association between AVPR1a and SLC6A4 reflects the social communication, courtship, and spiritual facets of the dancing phenotype rather than other aspects of this complex phenotype, such as sensorimotor integration.
Synopsis
Dancing, integrally related to music, likely has its origins close to the birth of Homo sapiens. The authors hypothesized that there are differences in aptitude, propensity, and need for dancing that may be based on differences in common genetic polymorphisms. Identifying such differences may lead to an understanding of the neurobiological basis of dancing.
Variants of the serotonin transporter and the arginine vasopressin receptor 1a genes were examined in performing dancers, elite athletes, and nonathletes/nondancers. The serotonin transporter regulates the level of serotonin, a brain transmitter that contributes to spiritual experience. The vasopressin receptor has been shown in many animal studies to modulate social communication and affiliative behaviors. Notably, dancers scored high on the Tellegen Absorption Scale, a correlate of spirituality, and the Reward Dependence factor in Cloninger's Tridimensional Personality Questionnaire, a measure of empathy, social communication, and need for social contact. Significant differences were observed in allele frequencies for both genes when dancers were compared to athletes as well as to nondancers/nonathletes. These two genes were also associated with scores on the Tellegen Absorption Scale and Tridimensional Personality Questionnaire Reward Dependence, suggesting that the association between these genes and dance is mediated by personality factors reflecting the social communication, courtship, and spiritual facets of the dancing phenotype.
Citation:Bachner-Melman R, Dina C, Zohar AH, Constantini N, Lerer E, et al (2005) AVPR1a and SLC6A4 gene polymorphisms are associated with creative dance performance. PLoS Genet 1(3): e42.
==== Body
Introduction
“With the creation of the universe, the dance too came into being, which signifies the union of the elements. The round dance of the stars, the constellation of planets in relation to the fixed stars, the beautiful order and harmony in all its movements, is a mirror of the original dance at the time of creation.”
Lucian of Samosata (~125 to 180 A.D.), On Dance (De Saltatione)
Dance, an art form closely allied to music, has been little studied from the neuroscience or genetic perspective, despite its significance in all cultures throughout the ages. Dance, like music, is an activity dating to prehistoric times that is sometimes a sacred ritual, sometimes a form of communication, and sometimes an important social and courtship activity; finally, dance is an art form that exists in every culture and manifests diverse paths [1]. Dance, as an expressive art form, is often considered inherently creative, especially when compared with a “nonartistic” domain. It is also a cultural form that results from creative processes that manipulate human bodies in space and time (“embodiment”). In many ways, dance is also a part of music, to which it is integrally related. Finally, professional dancers possess an exceptional talent, and as noted by Kalbfleisch [2], “Exceptional talent is the result of interactions between goal-directed behavior and nonvolitional perceptual processes in the brain that have yet to be fully characterized and understood by the fields of psychology and cognitive neuroscience.”
Dance may appear to be an unusual phenotype for human molecular genetics studies, but it is no more so than two closely related phenotypes, music [3] and athletic performance [4−6], that have both become subjects of molecular research. There is accumulating contemporary interest in the neuroscience of music [3,7−11] providing “proof of principle” that a widespread pursuit historically considered as part the human art and cultural heritage also has a solid basis in neuroscience, evolution, and genetics. Both music and athletic performance are complex phenotypes, the presentation of which is molded by environment and genes (and their interaction), especially in elite performers. A good example is absolute pitch, a relatively “clean” musical phenotype, of which the occurrence in approximately 20% of professional musicians is dependent not only on intrinsic ability but also on age of onset and intensity of musical training [11]. Similarly for athletic performance, evidence has accumulated over the past three decades for a strong genetic influence on human physical performance, with an emphasis on two sets of physical traits, cardiorespiratory and skeletal muscle function, that are particularly important for performance in a variety of sports [4]. A number of individual genetic variants associated with elite athletes have been provisionally identified, but there is little argument that elite athletes as well as elite musicians likely possess other characteristics related to personality and emotion that also contribute to their performance.
We suggest the notion that the “dance” phenotype is no more difficult to define than other complex human behavioral phenotypes (schizophrenia, attention deficit, personality, violence, and others) that have been shown to be both heritable and amenable to genetic analysis. Dancers fulfill a set of criteria with considerable face validity (similar in principle to the usual Diagnostic and Statistical Manual of Mental Disorders–style “symptom checklist” [12]) that both identifies and distinguishes one disorder from another. For example, the US Department of Labor suggests that the following qualities, inter alia, are required to be a professional dancer: flexibility, agility, coordination, grace, a sense of rhythm, a feeling for music, and a creative ability to express oneself through movement [13].
In our ongoing studies of the genetic basis of human personality [14−16], we have recruited currently performing dancers (n = 85) who train for at least 10 h per week, because we thought that a study of this group would help us understand why some individuals are endowed with creative and artistic abilities or inclinations. Toward this end, dancers were characterized using both psychosocial instruments and common genetic polymorphisms. Of particular interest are the Tridimensional Personality Questionnaire (TPQ) [17] and the Tellegen Absorption Scale (TAS) [18], which, respectively, measure aspects of social communication (TPQ Reward Dependence) and spirituality (TAS), personality facets important in the dance phenotype.
We investigated two polymorphic genes that we hypothesized to add to artistic creativity: the arginine vasopressin 1a receptor (AVPR1a) and the serotonin transporter (SLC6A4). The SLC6A4 long promoter allele is more efficient at the level of transcript, producing more transporter protein that presumably more effectively removes serotonin from the synapse [19]. Both common intron 2 VNTR repeats (10 and 12) enhance transcription [20], although individual repeat elements differ in their activity in embryonic stem cell models [21]. In lower vertebrates, the promoter region repeat elements of the AVPR1a receptor determine brain-specific expression patterns and are responsible for differences in patterns of social communication across species [22]. In humans, the functional significance of the promoter repeats remains to be elucidated, although association between these repeats and social communication in humans was recently suggested [14,23,24].
We considered that AVPR1a might contribute to the dance phenotype, reflecting this gene's role in affiliative, social, and courtship behaviors [25], activities that are vital in many kinds of human dancing. Dancing also taps into human spiritual resources as evidenced by the role of dancing in sacred rituals [1]; it has been shown that serotonin plays a role in human spiritual experiences [26]. Additionally, use of ecstasy, a serotonergic neurotoxin, at rave dances and dance clubs [27] further links serotonin to both dancing and states of altered consciousness, two phenomena also linked in the absence of drugs. Finally, many studies show that serotonin enhances the release of vasopressin in the brain [28], suggesting the notion that these two genes, AVPR1a and SLC6A4, are also likely to exhibit epistasis, or gene−gene interactions, in association studies that reflect their interaction at the level of individual neurons as well as on the plane of neurotransmitter pathways. Interestingly, serotonin and vasopressin interact in the hypothalamus to control communicative behavior [29].
Results
We first examined the arginine vasopressin 1a receptor (AVPR1a) promoter region microsatellites (allele frequencies for RS1 and RS3 are shown in Table 1) and the serotonin transporter gene (SLC6A4), initially using a case-control design by implementing the Cocaphase routine in the UNPHASED package to compare allele and haplotype frequencies between two groups, dancers versus athletes (Table 2). Comparing dancers with athletes is of interest because both groups of subjects demonstrate physical prowess and dedication to a demanding training routine but are expected to differ in musical aptitude and inclination. Indeed, dancers are sometimes considered performing athletes [30]. Single-locus analysis showed significance for the RS3 marker and a two-locus haplotype (RS1 and RS3). When the AVPR1a polymorphisms were analyzed conditional on the SLC6A4 polymorphisms, highly significant differences in allele and haplotype frequencies were observed (Table 2). Similar results were obtained when dancers were compared to the nondancers/nonathletes (AVPR1a RS1 and RS3 conditional on SLC6A4 HTTLPR and VNTR: likelihood ratio test [LRS] = 92.76, DF = 44, p = 0.00002).
Table 1
AVPR1a Allele Frequency for RS1 and RS3
Table 2 Case-Control Design (Cocaphase): Comparing Dancers to Athletes
We also tested preferential transmission of allele and haplotypes using a robust family-based design implemented in the Tdtphase routing of UNPHASED by assigning “dancer” as an “affective” status (Table 3). Preferential transmission of the AVPR1a microsatellite alleles from heterozygous parents to their dancer offspring was observed both for individual genes and haplotypes. The most significant evidence for transmission was observed when RS1 and/or RS3 transmission was conditional on SLC6A4 HTTLPR and/or VNTR polymorphisms.
Table 3 Testing Association between AVPR1a and SLC6A4 and Dancing in a Family-Based Design (Tdtphase)
To better understand the psychobiological mechanism by which the AVPR1a gene contributes to the “dancer” phenotype, we also compared dancers to nondancers on several psychosocial scales (Table 4). Highly significant differences between dancers and athletes are observed for two variables: the TAS [18] (effect size = 0.72 standard deviation [SD] units) and TPQ Reward Dependence [17] (effect size = 0.68 SD units). Less significant differences were observed for TPQ Self-Esteem [31], Fear of Failure [32], and Drive for Success [32]. Absorption is a tendency to experience episodes of “total” attention that fully engage perceptual, enactive, imaginative, and ideational resources [18]. The high scores of the dancers on the TAS and TPQ Reward Dependence suggested it would be worthwhile to examine association between these scores and both genes in the full group (nondancers and nonathletes) of subjects (n = 872 from 414 families) who we recruited in our personality studies. As shown in Table 5 (see Figures 1 and 2 for the distribution of TAS and TPQ Reward Dependence scores in dancers versus nondancers), using the family-based design there is an association between the AVPR1a and SLC6A4 genes and scores on the TAS. The most common (42%) SLC6A4 haplotype, short HTTLPR promoter region-12 repeat VNTR repeat, shows the strongest association with high TAS scores (p = 0.018). The two-locus RS1 and RS3 haplotype is also significantly associated with the TAS (p = 0.047). We also categorized TAS scores (“affected” as top 20%; score > 24, n = 203 affected) and observed significant association when RS1 and RS3 were conditional on the HTTLPR and VNTR polymorphisms (p = 0.01). As shown in Table 6, association was also observed between the AVPR1a RS1 microsatellite and TPQ Reward Dependence (p = 0.009), and the results were highly significant (p = 0.00097) for two-locus analysis when both RS1 and RS3 microsatellites were conditional on both SCL6A4 polymorphisms (n = 181 “affected” top 20% scores > 18). No association was observed between these two genes and TPQ Self-Esteem, Drive for Success, or Fear of Failure (data not shown).
Table 4 Comparison between Dancers and Athletes on Demographic and Psychosocial Scales
Table 5 Family-Based Design (TDT) and Testing Association between AVPR1a, SLC6A4, and TAS
Figure 1 Distribution of TAS in Female Dancers and Nondancers/Nonathletes
Figure 2 Distribution of TPQ Reward Dependence Scores in Female Dancers and Nondancers/Nonathletes
Table 6 Family-Based Design (TDT) and Testing Association between AVPR1a, SLC6A4, and TPQ Reward Dependence
We also examined other polymorphic genes of some interest in neuroscience but failed to show a significant association with creative dancing. These included the dopamine D5 microsatellite marker (DRD5), linked in some studies to attention deficit [33,34], the insulin-like growth factor 2 (IGF2, three single nucleotide polymorphisms), linked to eating disorders and self-report measures of altruism [15,35], catechol-O-methyltransferase (COMT) [36], linked to cognitive function and schizophrenia [37], and monoamine oxidase A (MAOA) [38], linked to violence [39]. Preliminary evidence for an association between the dopamine D4 receptor (DRD4) [35,40] and the dancing phenotype was observed, and further studies of this gene are now in progress.
Discussion
Few, if any, genes have been associated with artistic creativity, and none, to our knowledge, specifically with dancing. The current study considered two genes, AVPR1a and SCL6A4, that are significantly associated with performing dancers. Although the association seems robust and is significant both by case-control and family-based designs, it is nevertheless a challenge to unravel the brain mechanisms by which these genes partially contribute to dancing, a phenotype extending across musical processing, motor coordination, and artistic creativity.
The association between AVPR1a and SLC6A4 polymorphisms and creative dancing does not exclude the presence of the same polymorphisms in nondancing groups of subjects. Almost all of us dance and almost all of us have engaged in sports. What the current study suggests is that the combination of polymorphic variants contributing to creative dancing is overrepresented in the dancers. There is no reason to suggest that the nondancer athletes or the control group of nondancers/nonathletes are devoid of these polymorphisms, but the current study provides evidence that these variants are relatively scarce in other groups not specifically selected for the creative dancing phenotype. Importantly, we not only compared creative dancers to performing athletes but also validated the case-control design using a family-based study that avoids the conundrum of a comparison control group that might be “contaminated” with polymorphisms contributing to creative dancing. As for most complex traits, the effect size of these two genes is small and in Risch's terminology will have small displacement [41]
The AVPR1a gene makes a profound contribution to affiliative and social behavior in lower vertebrates [25]. It is hypothesized that microsatellite promoter-region instability may be a major factor producing diversity in both region-specific gene expression and the resulting phenotypes [42]. In humans, however, the role of the AVPR1a repeat regions has yet to be resolved. Nevertheless, despite the dearth of research regarding the molecular mechanisms involved, the AVPR1a promoter-region microsatellites have recently been associated with autism [23,24], a disorder whose core symptom is a deficit in social communication. Additionally, we have shown that this gene is associated with measures of social behavior in control subjects [14]. We suggest the notion that the association between AVPR1a and dancing may be reflecting the importance of social relations and communication in the dance form and that both dance and its associated gene, AVPR1a, contribute to molding social interactions from the molecular level to the dance floor. As noted by Kaeppler [1], the cultural form produced in dance, although transient, has structured content and is a visual manifestation of social relations that may be the subject of an elaborate aesthetic system. It seems likely that one of the many prerequisites for a successful dance career is the capacity for social communication through dance (“embodiment”).
Intriguingly, Darwin recorded in The Voyage of the Beagle (Chapter 19) his Australian encounter with the so-called White Cockatoo” aboriginal men and their performance in a corrobery, or “great dancing-party.” As Darwin notes, “Perhaps these dances originally represented actions, such as wars and victories; there was one called the Emu dance, in which each man extended his arm in a bent manner, like the neck of that bird. In another dance, one man imitated the movements of a kangaroo grazing in the woods, whilst a second crawled up, and pretended to spear him” [43]. It is worth noting in this context that the native Australians arrived on that continent approximately 50,000 y ago [44]. Similarly, Native American, also known for their complex dance culture [1], are estimated to have arrived in North America about 20,000 to 15,000 calendar years before the present [45]. We conjecture that both groups arrived on their respective continents already with a dance culture (alternatively there was parallel “cultural” evolution of this art form) that we speculate is likely to have originated before the African exodus. The earliest evidence for dance is derived from a cave painting in Creswell, England, that depicts dancing women and is dated approximately 13,000 y ago [46]. We hypothesize that this early evidence for dancing and its occurrence in groups geographically separated by thousands of years during our prehistory suggests a genetic basis for this behavior in H. sapiens.
Another aspect of dance is its spiritual side (e.g., sacred dancing across many diverse cultures) and the relationship of dance to altered states of consciousness; for example, in the Korean Salpuri dance, an ecstatic trance state is induced that results in changes in alpha wave activity [47]. Dances in which the participant (often a woman) enters into a trance and may lose consciousness have been used in a therapeutic setting (e.g., to dispossess “demons”) in diverse cultures, including a North African Jewish community [48]. We suggest the notion that the association we observe between SLC6A4 and dance is perhaps related to the need for altered consciousness states that subjects participating in and performing this art form sometimes have. Dancing may have its origins in shamanism, which sometimes used a potent and synergetic mix of music and dancing (and sometimes drugs) to alter consciousness [49]. Perhaps a prerequisite for some types of dancing, in both sacred and more modern “profane” versions as either an artistic performer or a participant, is the ability to enter into such a higher state of awareness.
The personality construct of absorption is generally measured using the TAS, a self-report personality instrument with good psychometric properties [50]. The TAS has been widely applied in research in an attempt to evaluate the significance of an individual's ability to attend intensely and imaginatively to stimuli. It has been found to correlate positively with spirituality and altered states of consciousness, such as an intrinsic religious orientation involving the internalization of religious tenets so they provide meaning and direction [51], and the frequency and type of reported spontaneous mystical, visionary, and paranormal experiences [18,50,52–55].
The short SLA6A4 promoter region polymorphism that we find associated with scores on the TAS characterizes a less efficient promoter, and in subjects carrying the short allele, less transporter protein is synthesized [19,56], which would lead to altered synaptic levels of serotonin. It is not surprising therefore that we observe an association between TAS scores (indicating in some individuals a propensity to have mystical, visionary experiences, and/or “artistic” sensitivity) and a common genetic polymorphism that modulates synaptic serotonin levels. A large body of evidence links hallucinogens, drugs that alter consciousness, to serotonin, and it is thought that hallucinogens stimulate 5-HT2A receptors, especially those expressed on neocortical pyramidal cells [57]. Altered serotonin levels in carriers of the SLC6A4 promoter region allele might predispose such individuals to a greater ability for imagery and attention to stimuli (especially to musical stimuli) that we hypothesize may provide part of the “hard wiring” that talented and devoted individuals need to perform in an art form that combines a unique combination of both musical and physical skills.
Individuals high in TPQ Reward Dependence tend to be tender-hearted, loving and warm, sensitive, dedicated, dependent, and sociable [58]. They seek social contact and are open to communication with other people. Typically, they find people they like everywhere they go and are sensitive to social cues, which facilitates warm social relations and understanding of others' feelings. The observed association between TPQ Reward Dependence scores and AVPR1a is consistent with the role of the arginine vasopressin receptor in social communication as demonstrated in extensive animal experiments [59]. The association between AVPR1a and Reward Dependence personality traits strengthens the notion that this gene contributes to dancing through its contribution to social communication.
Figure 3 summarizes our notions of how polymorphisms in the AVPR1a and SLC6A4 genes contribute to the dance phenotype.
Figure 3 Epistatic Interaction between AVPR1a and SLC6A4 Contributes to the Creative Dance Phenotype
Promoter region polymorphisms in the AVPR1a receptor region possibly contribute to regional differences in brain arginine receptor 1a expression patterns [42]. Vasopressin release, and subsequent AVPR1a receptor activation, is partially regulated by serotonin (5-HT) [28]. 5-HT is removed from the synapse by the serotonin transporter (SLC6A4), which plays a major role in regulation of synaptic levels of this neurotransmitter. In turn, synaptic SLC6A4 mRNA and protein levels are controlled in part by the presence or absence of a promoter region 44-bp insertion/deletion [19]. Subjects with polymorphic variants of these two genes are therefore predicted to show differences in serotonergic and vasopressin tone that contribute to differences in higher psychological constructs including TPQ Reward Dependence (associated with AVPR1a and AVPR1a × SLC6A4 gene × gene interaction) and TAS (associated with SLC6A4 and AVPR1a × SLC6A4 gene × gene interaction). Dancers score high on these two personality constructs, suggesting the hypothesis that the association between AVPR1a and SLC6A4 polymorphisms and dancing is likely mediated by the action of these two genes primarily on social communication (measured by TPQ Reward Dependence scores) and spirituality (measured by TAS scores). Similar to genes contributing to other complex traits, there are no “dancing” genes but rather common polymorphisms that contribute to simpler endophenotypes [77], such as TPQ Reward Dependence and TAS, that constitute some of the critical psychological underpinnings of the dance phenotype.
The proposed role of AVPR1a in contributing to the dance phenotype in humans as provisionally shown in the current report has a solid evolutionary basis. Across the vertebrates, vasopressin plays a key role in courtship behavior that frequently involves elements of song and dance. For example, male zebra finches (Taeniopygia guttata) sing directed song to females as an integral part of a courtship display that also includes elements of dance [60]. The choreography of the dance presumably conveys or enhances some part of the message that is carried by the individual's learned song, although the exact importance and function of the dance are not known. Furthermore, arginine vasopressin plays a key role courtship behavior in zebra finches [61] as well as in other bird species such as the territorial field sparrow (Spizella pusilla) [62], as it does in social behaviors in mammals and other vertebrates [22]. There is also evidence in humans that vasopressin is important in maternal and romantic love. Imaging studies have shown that brain areas rich in vasopressin receptors are activated when subjects are shown pictures designed to evoke feelings of attachment [63]. The observation discussed above that AVPR1a is associated with a temperament trait, Reward Dependence, also strengthens the conjectured role of this gene in human social communication. Thus, studies in humans as well as in many other animal species suggest to us the reasonable notion that variations in AVPR1a microsatellite structure might also predispose some individuals to excel in dancing. Darwin observed that vocalization is a form of emotional expression, but among neuroethologists, questions concerning the role of emotion in vocal (and dance) communication have been superseded by questions that concern sensorimotor integration [64]. We believe that the two genes we have identified with dancing in humans are likely involved in the emotional side of dance rather than in the sensorimotor mechanics of this complex phenotype. To quote West and King [65], “Animals do not perceive or communicate for the sake of perceiving or producing a display, but for the sake of managing a social environment.” In this context, it is easier to see how the AVPR1a receptor microsatellite polymorphisms contribute to human dance. Human dancing can be understood in part as a form of courtship and social communication that shares a surprisingly conserved evolutionary history, characterized by apparently common neurochemical and genetic mechanisms, with mating displays and affiliative behavior observed across the vertebrates.
Materials and Methods
Dancers.
One hundred seven dancers, who trained for a minimum of 10 h per week and perform regularly, were initially recruited for this study from professional dance classes and dance companies in Israel. From this group, DNA was obtained from 85 dancers and their parents and was successfully genotyped (82 females and three males). The main types of dance performed by the 85 dancers who were genotyped were classical ballet, modern or contemporary dance, and jazz ballet. The average age of the dancers was 19.30 ± 4.55 y (SD).
This project was approved by the Herzog Hospital Helsinki Committee and the Israeli Ministry of Health, Genetics Section, and all subjects gave informed consent.
Athletes.
Ninety-one (22 males and 69 females) competitive (nonperforming) athletes (and their parents) were recruited. The athletes competed regularly at the highest echelons of their respective sports in Israel and often abroad; 36 (32 females and four males) were endurance athletes, mostly runners and swimmers; 39 (24 females and 15 males) competed at a high level in ballgames such as basketball and volleyball; nine (eight females and one male) competed in technical sports such as sailing and fencing; and seven (five females and two males) competed in the martial arts. Sixty-seven female nonperforming athletes were recruited via sports unions and the Israeli National Sports Institute. Their average age was 21.47 ± 5.23 y.
Nonathletes/nondancers.
This group (n = 872 from 414 families) has been described in our previous studies [14,15]. They were primarily university students recruited from the Hebrew University campus, Mt. Scopus, Jerusalem. The average age of the subjects was 21.44 ± 4.37 y (range, 13−36 y).
DNA extraction and genotyping.
DNA was obtained from all family members and extracted with use of the MasterPure kit (Epicentre, Madison, Wisconsin, United States). Amplification of the RS1 and RS3 arginine vasopressin 1a microsatellites (AVPR1a) was achieved using the following pair of primers: RS1 [24,66] forward (fluorescent) 5′-AGG GAC TGG TTC TAC AAT CTG C-3′ and reverse 5′-ACC TCT CAA GTT ATG TTG GTG G-3′; RS3 [24,66] forward (fluorescent) 5′-CCT GTA GAG ATG TAA GTG CT-3′ and reverse 5′-TCT GGA AGA GAC TTA GAT GG-3′. Microsatellite haplotypes (RS1 and RS3) showed mild linkage disequilibrium (UNPHASED: global D′ = 0.339). The allele frequencies for RS1 and RS3 are shown in Table 1.
Each reaction mixture contained 0.5 μM primer and 20 ng of DNA. A ReddyMix master mix (Thermoprime plus DNA polymerase) was used (Abgene, Surrey, United Kingdom) at a magnesium concentration of 1.5−2.5 mM MgCl2. ReddyMix buffer consisted of 75 mM Tris-HCl (pH 8.8 at 25 °C), 20 mM (NH4)2SO4, and 0.01% (v/v) Tween 20. The sample was initially heated at 95 °C for 5 min followed by 30 cycles of 95 °C (30 s), 55 °C (30 s), and 72 °C (40 s), and a final extension step of 72 °C for 10 min. The PCR product was analyzed on an ABI 310 DNA analyzer (Applied Biosystems, Foster City, California, United States).
A subroutine of Merlin was used to test for conformity with Hardy–Weinberg equilibrium (HWE) and no departure from HWE was observed for either microsatellite.
5-HTTLPR.
PCR amplification was carried out using a ReddyMix kit (Abgene). The primers used were forward 5′-GGCGTTGCCGCTCTGAATGC-3′ and reverse 5′-GAGGGACTGAGCTGGACAACC-3′. The reaction mixture contained the following components: 0.5 μM primers, 20 ng of DNA, and 5% DMSO in a total volume of 10 μl. ReddyMix buffer consisted of 75 mM Tris-HCl (pH 8.8 at 25 °C), 20 mM (NH4)2SO4, and 0.01% (v/v) Tween 20. After an initial denaturation step of 94 °C for 5 min, amplification was carried out for 35 cycles (94 °C for 30 s, 55 °C for 30 s, and 72 °C for 90 s) in a PerkinElmer (Wellesley, California, United States) Cetus 9600 thermal cycler. A 5-min final extension at 72 °C was used. The reaction mixture was electrophoresed on a 3% agarose gel (Ameresco, Solon, Ohio, United States) with ethidium bromide to screen for genotypes.
VNTR.
PCR amplification was carried out for the intron 2 VNTR with the following primers: forward 5′- TCAGTATCACAGGCTGCGAG-3′ and reverse 5′-TGTTCCTAGTCTTACGCCAGTG-3′ [67]. The reaction mixture contained the following components: 0.5 μM primers, 20 ng of DNA, and 5% DMSO in a total volume of 10 μl. ReddyMix buffer consisted of 75 mM Tris-HCl (pH 8.8 at 25 °C), 20 mM (NH4)2SO4, and 0.01% (v/v) Tween 20. After an initial denaturation step of 94 °C for 5 min, amplification was carried out for 35 cycles (94 °C for 30 s, 61 °C for 30 s, and 72 °C for 30 s) in a PerkinElmer Cetus 9600 thermal cycler. A 10-min final extension at 72 °C was used.
The SLC6A4 repeats were in LD: global D′ = 0.71 (UNPHASED). There was no deviation from HWE. The allele frequency of the HTTLPR polymorphism is 50.4% long (plus 44-bp insertion) and 49.4% short (44-bp deletion); there were four individuals with a rare short allele [68] who were excluded from the analyses. The allele frequency of the VNTR polymorphism is 27.6% of the 10 repeat, 72.2% of the 12 repeat, and four individuals with a 9 repeat who were excluded from the genetic analyses.
The percentage of heterozygote parents is as follows: AVPR1a RS1 = 72%, RS3 = 82%; SLC6A4 HTTLPR = 58%, VNTR = 44%.
Quality of genotyping.
Quality of genotyping was determined as follows. (1) All families were initially screened for Mendelian consistency using seven highly polymorphic microsatellite markers. Problematic families who were not consistent with Mendelian inheritance (<1%) were excluded from the study. (2) Because all subjects in the current study were family members, genotype errors with SNPs appearing as Mendelian inconsistency, and automatically flagged by the statistical program, were reexamined either for data entry errors or by regenotyping such families for the aberrant SNPs. (3) In all cases of borderline classifications, when reading the ABI output, the PCR procedure was repeated. (4) Quality control and estimation of error rate (percent of miscalled genotypes) were evaluated by reanalysis of 5% of the families. The observed error rate is estimated to be less than 0.5%. (5) Because deviation from HWE in random samples may be indicative of problematic assays [69], the frequency of the alleles was examined for HWE. Additionally, the genotype and allele frequencies were compared with published results from other investigations, and were found to be similar.
Statistical methods.
We used the logistic-based variant of the transmission disequilibrium test (TDT), so-called ETDT [70], to assess association (and linkage) without confounding effect of population stratification. The TDT, in its simplest version, compares, for one allele, the number of times this allele is transmitted to the number of times where this allele is not transmitted to an affected offspring. Note that only heterozygous parents are informative. This approach can be extended to haplotypes.
The various tests we used are implemented in the program UNPHASED (http://www.rfcgr.mrc.ac.uk/~fdudbrid/software/unphased/). UNPHASED [71] is a suite of programs for association analysis of multilocus haplotypes from unphased genotype data. UNPHASED currently includes the following programs: UNPHASED, which is the graphical front end to the analysis programs; Tdtphase for TDT and HHRR analysis for nuclear families; Cocaphase, for case-control data; Qtphase, for quantitative traits in unrelateds; Pdtphase, for pedigree disequilibrium tests; and Qpdtphase, for quantitative trait pedigree disequilibrium tests.
Conditional analysis.
Here, the ETDT is adapted to test for an effect at a secondary locus or marker conditional on the association of a candidate disease locus in case–parent triads. Considering gametic haplotypes of a candidate (or established) disease locus and a neutral marker, it is expected that haplotypes with identical alleles at the candidate disease locus, but different alleles at the marker, have equal transmission probabilities. ETDT transmission probabilities can be estimated and, in this adaptation, can be tested for equality using a LRT called Conditional ETDT [72]. A significant difference in transmission of haplotypes identical at the candidate locus, but different at the secondary locus, provides evidence for the involvement of either the secondary locus or a locus in linkage disequilibrium with it.
Because some markers are microsatellites, with RS1 = 9 and RS3 = 17 alleles, including some rare alleles, we used a permutation procedure to confirm the asymptotic p-value. This procedure randomly permutes for parental alleles or haplotypes and the transmission and nontransmission status and computes the observed ETDT statistic. The p-value is the number of times a specific observed statistic is observed in n permutations divided by the number of permutations (n).
For single-locus and haplotype analysis, UNPHASED calculates overall global p-values that consider multiple testing of haplotypes. Those values are included in all of the tables (“Global p-Values”).
However, regarding the more complicated problem of how to correct for multiple testing in association studies when there are potentially approximately 30,000 genes in the human genome, the reader is referred to the insightful article by Neale and Sham that discusses this problem [73]. In the current study, the p-values are nominal and not corrected for multiple testing.
This research was partially supported by the Israel Science Foundation founded by the Israel Academy of Sciences and Humanities (RPE) and the Israel Association of University Women (RB-M). We would like to thank all of the subjects and their family members, who so willingly participated in this research.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. RB-M, AHZ, and RPE conceived and designed the experiments. RB-M, NC, EL, SH, SS, LN, and IG performed the experiments. RB-M, CD, EL, IG, PL, RG, and RPE analyzed the data. CD, AHZ, NC, SH, SS, LN, PL, and RG contributed reagents/materials/analysis tools. RB-M, CD, AHZ, and RPE wrote the paper.
Abbreviations
HWEHardy–Weinberg equilibrium
LRSlikelihood ratio test
SDstandard deviation
TASTellegen Absorption Scale
TDTtransmission disequilibrium test
TPQTridimensional Personality Questionnaire
==== Refs
References
Kaeppler AL 1978 Dance in anthropological perspective Annu Rev Anthropol 7 31 49
Kalbfleisch ML 2004 Functional neural anatomy of talent Anat Rec B New Anat 277 21 36 15052651
Zatorre R McGill J 2005 Music, the food of neuroscience? Nature 434 312 315 15772648
Macarthur DG North KN 2005 Genes and human elite athletic performance Hum Genet 116 331 339 15726413
Scott RA Wilson RH Goodwin WH Moran CN Georgiades E 2005 Mitochondrial DNA lineages of elite Ethiopian athletes Comp Biochem Physiol B Biochem Mol Biol 140 497 503 15694598
Wolfarth B Bray MS Hagberg JM Perusse L Rauramaa R 2005 The human gene map for performance and health-related fitness phenotypes: The 2004 update Med Sci Sports Exerc 37 881 903 15947712
Bentivoglio M 2003 Musical skills and neural functions. The legacy of the brains of musicians Ann N Y Acad Sci 999 234 243 14681147
Minciacchi D 2003 Part II: Brain sciences versus music. Introduction Ann N Y Acad Sci 999 215 217 14681145
Janata P Grafton ST 2003 Swinging in the brain: Shared neural substrates for behaviors related to sequencing and music Nat Neurosci 6 682 687 12830159
Hauser MD McDermott J 2003 The evolution of the music faculty: A comparative perspective Nat Neurosci 6 663 668 12830156
Gregersen PK Kowalsky E Kohn N Marvin EW 2001 Early childhood music education and predisposition to absolute pitch: Teasing apart genes and environment Am J Med Genet 98 280 282 11169569
American Psychiatric Association 1994 Diagnostic and statistical manual of mental disorders Washington (D. C.) American Psychiatric Association 943 p.
U.S. Department of Labor Bureau of Labor Statistics 2004 Dancers and choreographers. Washington (D. C.): U.S. Department of Labor Bureau of Labor Statistics Available: http://bls.gov/oco/pdf/ocos094.pdf . Accessed 7 September 2005.
Bachner-Melman R Zohar AH Bacon-Shnoor N Elizur Y Nemanov L 2005 Linkage between vasopressin receptor AVPR1A promoter region microsatellites and measures of social behavior in humans J Individ Differ 26 2 10
Bachner-Melman R Gritsenko I Nemanov L Zohar AH Dina C 2005 Dopaminergic polymorphisms associated with self-report measures of human altruism: A fresh phenotype for the dopamine D4 receptor Mol Psychiatry 10 333 335 15655563
Benjamin J Ebstein RP Belmaker RH 2002 Molecular genetics and the human personality Washington (D. C.) American Psychiatric Publishing 355 p.
Cloninger CR 1987 A systematic method for clinical description and classification of personality variants. A proposal Arch Gen Psychiatry 44 573 588 3579504
Tellegen A Atkinson G 1974 Openness to absorbing and self-altering experiences (“absorption”), a trait related to hypnotic susceptibility J Abnorm Psychol 83 268 277 4844914
Lesch KP Bengel D Heils A Sabol SZ Greenberg BD 1996 Association of anxiety-related traits with a polymorphism in the serotonin transporter gene regulatory region Science 274 1527 1531 8929413
Fiskerstrand CE Lovejoy EA Quinn JP 1999 An intronic polymorphic domain often associated with susceptibility to affective disorders has allele dependent differential enhancer activity in embryonic stem cells FEBS Lett 458 171 174 10481059
Lovejoy EA Scott AC Fiskerstrand CE Bubb VJ Quinn JP 2003 The serotonin transporter intronic VNTR enhancer correlated with a predisposition to affective disorders has distinct regulatory elements within the domain based on the primary DNA sequence of the repeat unit Eur J Neurosci 17 417 420 12542679
Hammock EA Young LJ 2004 Functional microsatellite polymorphism associated with divergent social structure in vole species Mol Biol Evol 21 1057 1063 15014156
Kim SJ Young LJ Gonen D Veenstra-VanderWeele J Courchesne R 2002 Transmission disequilibrium testing of arginine vasopressin receptor 1A (AVPR1A) polymorphisms in autism Mol Psychiatry 7 503 507 12082568
Wassink TH Piven J Vieland VJ Pietila J Goedken RJ 2004 Examination of AVPR1a as an autism susceptibility gene Mol Psychiatry 9 968 972 15098001
Insel TR Young LJ 2001 The neurobiology of attachment Nat Rev Neurosci 2 129 136 11252992
Borg J Andree B Soderstrom H Farde L 2003 The serotonin system and spiritual experiences Am J Psychiatry 160 1965 1969 14594742
Parrott AC 2004 MDMA (3,4-methylenedioxymethamphetamine) or ecstasy: The neuropsychobiological implications of taking it at dances and raves Neuropsychobiology 50 329 335 15539865
Galfi M Radacs M Juhasz A Laszlo F Molnar A 2005 Serotonin-induced enhancement of vasopressin and oxytocin secretion in rat neurohypophyseal tissue culture Regul Pept 127 225 231 15680491
Albers HE Karom M Smith D 2002 Serotonin and vasopressin interact in the hypothalamus to control communicative behavior Neuroreport 13 931 933 12004193
Koutedakis Y Jamurtas A 2004 The dancer as a performing athlete: Physiological considerations Sports Med 34 651 661 15335242
Rosenberg M 1965 Society and the adolescent self-image Princeton (New Jersey) Princeton University Press 326 p.
Nygard R Gjesme T 1973 Assessment of achievement motives: Comments and suggestions Scand J Educ Res 17 39 46
Manor I Corbex M Eisenberg J Gritsenkso I Bachner-Melman R 2004 Association of the dopamine D5 receptor with attention deficit hyperactivity disorder (ADHD) and scores on a continuous performance test (TOVA) Am J Med Genet B Neuropsychiatr Genet 127 73 77
Lowe N Kirley A Hawi Z Sham P Wickham H 2004 Joint analysis of the DRD5 marker concludes association with attention-deficit/hyperactivity disorder confined to the predominantly inattentive and combined subtypes Am J Hum Genet 74 348 356 14732906
Bachner-Melman R Zohar AH Nemanov L Heresco-Levy U Gritsenko I 2005 Association between the insulin-like growth factor 2 gene (IGF2) and scores on the Eating Attitudes Test in nonclinical subjects: A family-based study Am J Psychiatry In press.
Lachman HM Papolos DF Saito T Yu YM Szumlanski CL 1996 Human catechol-O-methyltransferase pharmacogenetics: Description of a functional polymorphism and its potential application to neuropsychiatric disorders Pharmacogenetics 6 243 250 8807664
Egan MF Goldberg TE Kolachana BS Callicott JH Mazzanti CM 2001 Effect of COMT Val108/158 Met genotype on frontal lobe function and risk for schizophrenia Proc Natl Acad Sci U S A 98 6917 6922 11381111
Deckert J Catalano M Syagailo YV Bosi M Okladnova O 1999 Excess of high activity monoamine oxidase A gene promoter alleles in female patients with panic disorder Hum Mol Genet 8 621 624 10072430
Caspi A McClay J Moffitt TE Mill J Martin J 2002 Role of genotype in the cycle of violence in maltreated children Science 297 851 854 12161658
Ebstein RP Novick O Umansky R Priel B Osher Y 1996 Dopamine D4 receptor (D4DR) exon III polymorphism associated with the human personality trait of Novelty Seeking Nat Genet 12 78 80 8528256
Risch NJ 2000 Searching for genetic determinants in the new millennium Nature 405 847 856 10866211
Hammock EA Young LJ 2002 Variation in the vasopressin V1a receptor promoter and expression: Implications for inter- and intraspecific variation in social behaviour Eur J Neurosci 16 399 402 12193181
Darwin C 2002 Voyage of the Beagle Mineola (New York) Dover Publications 524 p.
Bowler JM Johnston H Olley JM Prescott JR Roberts RG 2003 New ages for human occupation and climatic change at Lake Mungo, Australia Nature 421 837 840 12594511
Schurr TG Sherry ST 2004 Mitochondrial DNA and Y chromosome diversity and the peopling of the Americas: Evolutionary and demographic evidence Am J Hum Biol 16 420 439 15214060
Pickerell J 2004 July 19 Ice age Sistine Chapel found ScienceNOW 2004 3
Park JR Yagyu T Saito N Kinoshita T Hirai T 2002 Dynamics of brain electric field during recall of Salpuri dance performance Percept Mot Skills 95 955 962 12509202
Somer E 2000 Stambali: Dissociative possession and trance in a Tunisian healing dance Transcult Psychiatry 37 580 600
Nencini P 2002 The shaman and the rave party: Social pharmacology of ecstasy Subst Use Misuse 37 923 939 12180571
Tellegen A 2003 The MMPI-2 restructured clinical (RC) scales: Development, validation, and interpretation Minneapolis University of Minnesota Press 116 p.
Prohaska PR 2002 Religious beliefs and trait absorption scores: Implications for mental and physical health Dissert Abstr Int Sect B Sci Eng 62 4459
Dixon M Louise L Laurence JR 1996 A multivariate approach to the prediction of hypnotic susceptibility Int J Clin Exp Hypn 44 250 264 8690536
Irwin HJ 1985 Parapsychological phenomena and the absorption domain J Am Soc Psychical Res 1 11
Nelson PL 1989 Personality factors in the frequency of reported spontaneous praeternatural experiences J Transpersonal Psychol 21 193 209
Nelson PL 1992 Personality attributes as discriminating factors in distinguishing religio-mystical from paranormal experiences Imagination Cogn Pers 11 389 405
Heils A Teufel A Petri S Stober G Riederer P 1996 Allelic variation of human serotonin transporter gene expression J Neurochem 66 2621 2624 8632190
Nichols DE 2004 Hallucinogens Pharmacol Ther 101 131 181 14761703
Cloninger CR Svrakic DM Prybeck TR 1994 Cloninger CR, Svrakic DM, Prybeck TR (1994) The temperament and character inventory (TCI) A guide to its development and use. St. Louis: Center for Psychobiology of Personality 85 p.
Ferguson JN Young LJ Insel TR 2002 The neuroendocrine basis of social recognition Front Neuroendocrinol 23 200 224 11950245
Williams H 2001 Choreography of song, dance and beak movements in the zebra finch (Taeniopygia guttata)
J Exp Biol 204 3497 3506 11707499
Harding CF Rowe SA 2003 Vasotocin treatment inhibits courtship in male zebra finches; concomitant androgen treatment inhibits this effect Horm Behav 44 413 418 14644635
Goodson JL 1998 Territorial aggression and dawn song are modulated by septal vasotocin and vasoactive intestinal polypeptide in male field sparrows (Spizella pusilla)
Horm Behav 34 67 77 9735230
Bartels A Zeki S 2004 The neural correlates of maternal and romantic love Neuroimage 21 1155 1166 15006682
Cheng MF Durand SE 2004 Song and the limbic brain: A new function for the bird's own song Ann N Y Acad Sci 1016 611 627 15313797
West M King AP 1996 Eco-gen-actics: A systems approach to the ontogeny of avian communication Kroodsma DE Miller TE Ecology and evolution of acoustic communication in birds Ithaca (New York) Cornell University Press 20 38
Thibonnier M Graves MK Wagner MS Chatelain N Soubrier F 2000 Study of V(1)-vascular vasopressin receptor gene microsatellite polymorphisms in human essential hypertension J Mol Cell Cardiol 32 557 564 10756113
MacKenzie A Quinn J 1999 A serotonin transporter gene intron 2 polymorphic region, correlated with affective disorders, has allele-dependent differential enhancer- like properties in the mouse embryo Proc Natl Acad Sci U S A 96 15251 15255 10611371
Frisch A Finkel B Michaelovsky E Sigal M Laor N 2000 A rare short allele of the serotonin transporter promoter region (5-HTTLPR) found in an aggressive schizophrenic patient of Jewish Libyan origin Psychiatr Genet 10 179 183 11324943
Hosking L Lumsden S Lewis K Yeo A McCarthy L 2004 Detection of genotyping errors by Hardy-Weinberg equilibrium testing Eur J Hum Genet 12 395 399 14872201
Sham PC Curtis D 1995 An extended transmission/disequilibrium test (TDT) for multi-allele marker loci Ann Hum Genet 59 323 336 7486838
Dudbridge F 2003 Pedigree disequilibrium tests for multilocus haplotypes Genet Epidemiol 25 115 121 12916020
Koeleman BP Dudbridge F Cordell HJ Todd JA 2000 Adaptation of the extended transmission/disequilibrium test to distinguish disease associations of multiple loci: The Conditional Extended Transmission/Disequilibrium Test Ann Hum Genet 64 207 213 11246472
Neale BM Sham PC 2004 The future of association studies: Gene-based analysis and replication Am J Hum Genet 75 353 362 15272419
Hewitt PL Flett GL 1991 Perfectionism in the self and social contexts: Conceptualization, assessment, and association with psychopathology J Pers Soc Psychol 60 456 470 2027080
Hodgeson RJ Rachman S 1977 Obsessive-compulsive complaints Behav Res Ther 15 389 395 612339
Lennox RD Wolfe RN 1984 Revision of the Self-Monitoring Scale J Pers Soc Psychol 46 1349 1364 6737217
Gottesman II Gould TD 2003 The endophenotype concept in psychiatry: Etymology and strategic intentions Am J Psychiatry 160 636 645 12668349
|
16205790
|
PMC1239939
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Sep 30; 1(3):e42
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010042
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1620579110.1371/journal.pgen.001004505-PLGE-I-0255plge-01-03-12InterviewTurning the Tables—An Interview with Nicholas Wade InterviewGitschier Jane Jane Gitschier is in the Department of Medicine and Pediatrics and the Howard Hughes Medical Institute, University of California, San Francisco, California, United States of America. E-mail: [email protected]
9 2005 30 9 2005 1 3 e45Copyright: © 2005 Jane Gitschier.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Citation:Gitschier J (2005) Turning the tables—An interview with Nicholas Wade. PLoS Genet 1(3): e45.
==== Body
For many of us, keeping up with the literature poses a never-ending challenge. We rely on word of mouth, “news and views,” journal clubs, or E-mail alerts to stay abreast of our field. But the media are also potent arbiters of scientific advances. And in the realm of genetics, I can think of no better source than Nicholas Wade of the New York Times.
Nicholas Wade
What I like about reading Wade is that he gets right to the point—describing the discovery at hand in the first paragraph, yet roping me in for the full story. He's clear, crisp, and rarely misses the mark.
When visiting my father in rural Pennsylvania recently, I had the opportunity of interviewing Wade in situ at the venerable Times office building in Manhattan. The bus deposited me at the Port Authority Bus Terminal at the corner of 8th Avenue and 41st Street in blistering heat. I quickly walked two blocks north and turned the corner at 43rd Street to encounter the massive Times structure looming above, its signature globe lights forming a beacon to a set of revolving doors. I entered beneath the motto “All the news that's fit to print,” registered at the front desk, and took the elevator up to the fourth floor, where I was met by Wade.
Wade gave me a quick accounting of the enormous newsroom, which at that level, houses science and arts reporters and editors. On passing through a maze of cubicles, he commented that it looked just like any office building. “A lot messier,” I rejoined, as untidy stacks of papers were strewn everywhere. Inside a small conference room, I fiddled with my tape recorder, cognizant of a pro watching this neophyte. I looked across to a British man with a soft voice, lively blue eyes, and a puckish grin. He wasted no time turning the tables on me by asking, “Will the questions get too pointed?”
“No, they won't,” I replied. “I'm very sweet.”
“Your first mistake,” said he. And thus we began.
Jane Gitschier: How do you describe your beat?
Nicholas Wade: We're a small department, and the boundaries of the beats are flexible. By and large, I cover whatever I'm interested in. I manage to keep out of others' way because I keep close to the frontiers of biological research, particularly genetics and molecular biology, and no one else has quite the same interests. Many of my colleagues write about medicine, for example, but I write only about things that are perfectly useless, given that it takes some time to translate basic research into anything practical.
It's one of the challenges we have as a science section—to get people interested in things that are of purely intellectual consequence.
Gitschier: How do you envision your readership?
Wade: As a general policy, the newspaper is addressed to the intelligent and informed reader, but it's always with the idea of bringing news. We're not in the business of education. People can find general information from a dictionary or the Internet.
For the science section, although we should have the same readership as the main newspaper, I assume that readers have a certain amount of scientific knowledge or interest, and, of course, many readers of the section are scientists. So we can sometimes put in more technical detail than we would in the main newspaper.
Gitschier: The Times doesn't do a readership poll to ask how many people are actually reading Nicholas Wade's articles?
Wade: Our business office does do those polls, but they are kept secret from us. That's with the idea that the content of the newspaper should not be driven by polls or market results but rather by what the editors think is important.
Gitschier: How long have you worked for the New York Times?
Wade: Longer than I like to think! I came in 1981.
I was on the news section of Science, and before that with Nature. At both journals, I was mostly concerned with political stories that affected science. I found I had a great deal to learn when I came back to science writing.
Gitschier: And before that, were you a scientist yourself or a journalist in another area?
Wade: I guess I always wanted to be a writer, but I was interested in science, so I read science at university [Cambridge], though without any intention of becoming a scientist. I didn't plan to be a science writer either, but the two things came together. I got a job with Nature in London when I was quite young, and was sent off to the Washington office that Nature had just set up. After a few years, Science asked me to join them. I thought it would be fun to work with an American company. I worked for them for about ten years.
The Times asked me to join them as an editorial writer to cover science, technology, and medicine, and I did that for about ten years. Editorial writing is great fun, but it's rather a limited art form.
Gitschier: Can you describe editorial writing?
Wade: Editorials are unsigned pieces [on the editorial page] because they are intended to be the voice of the paper. That anonymity may give the writer's words extra authority, but the disadvantage is that you lose your byline, unfortunately. So as a writer, you essentially disappear from public view.
The only exception is when you advocate a position that the editors do not think should be the position of the paper. The piece is then called an editorial notebook and appears with your byline to make clear it's just your opinion, not theirs.
After writing editorials for ten years, I became science editor. That, too, is a job that deprives you of a byline because there is almost no time to write. The science editors handle both the science stories in the daily paper, and those in the weekly science section. As an editor, you get to see how the paper works, which is of great interest, but you cease reporting, and spend a lot of time improving other people's stories and attending meetings.
One of the great things about being a writer, particularly an editorial writer, is that you get to know a lot. The whole world comes through New York, and many people want to talk to the Times. Often the reporters and editors on the main paper are too busy to see them, so they end up talking to the editorial board, where the pressure of work is much less. So even if you don't cover foreign policy or defense, you can get to meet the leaders in these fields by sitting in on your colleagues' meetings.
But when I became an editor, I found I was one step back from the front line of the news. Being unable to research or write anything, my intellectual capital dwindled fast, until I began to feel I had gone from knowing almost everything about the world as an editorial writer to knowing almost nothing [as an editor]. On a newspaper, the most interesting job is reporting. I went back to writing as a reporter five or six years ago.
Gitschier: I'm interested in the process that you undergo in developing a story. First, how do you discover what's out there?
Wade: Mostly, we find out through the main journals that we watch. Most of them have now become sophisticated in preparing what they call “tip sheets,” or weekly lists of their most newsworthy articles, which are seen as a marketing tool for journals. They'll say to potential authors, “Send us your paper, and we'll get you mentioned in the press.” Tip sheets are useful but insidious because it's easy to rely on them too much and not read the stories in the rest of the journal. So it's an imperfect system.
Gitschier: What journals give you these tip sheets?
Wade: A lot of journals do it now. I look carefully at the Nature journals, Science, PNAS [Proceedings of the National Academy of Sciences], American Journal of Human Genetics, and the Cell journals.
Gitschier: And PLoS, of course!
Wade:
PLoS, of course!
Gitschier: Are you ever strong-armed by anyone to write about his or her work?
Wade: People sometimes do call up, but not as much as I'd like. Scientists are reticent about promoting their work to the press because they risk being criticized by their colleagues for doing so. But sometimes people call to say, “I've got something very exciting,” and send me the paper in advance. It's always very useful to hear from people when they are enthusiastic about a result.
Gitschier: Do you attend scientific meetings?
Wade: Yes, I do, but not as many as I'd like. When you go to a meeting, you're usually obligated to write a story about it, and many scientific meetings are very hard to write about for the general reader because the findings are often incremental advances and difficult to summarize. On the other hand, meetings are very useful for talking to people in person, so I try to go to as many as I can.
Gitschier: After you read a paper, what's your next step?
Wade: I usually start by talking to the authors, and then call others in the field to see if they share the author's interpretation of the finding. Much of this can also be done with E-mail. I try to keep talking with people until I feel I understand a paper and its strengths and weaknesses, and then I'm ready to write it up.
Gitschier: Do you do most of your writing at home?
Wade: It's more restful to write at home. But if I have a story that will appear in the paper the next day, it's usually easier to be in the office.
Gitschier: How many articles do you write per week?
Wade: Usually about one or two, or more if there's lots of news. I've been on book leave for much of this year.
Gitschier: What is your new book about?
Wade: It's on what genetics is telling us about human evolution, human nature, and prehistory. I'm trying to integrate information from the many different fields that bear on the human past—paleoanthropology, archeology, historical linguistics, and evolutionary psychology—all of which are now being informed and amplified by genetics.
Gitschier: Back to the process. What's next?
Wade: You have to sell a story to your editor. It's a quite small department, so if a reporter says, “This is an important subject,” it will probably go into the paper. But the question is at what length because space is at a premium. The editors who run the main paper, who tend to have a political/foreign affairs background, may not be as enthusiastic about science as we are. The science department's editors have to assess how much space they are likely to get for a story.
Gitschier: I want to learn more about which stories you choose to develop. When you look back on the stories that you've written about over the past five to six years, which ones leap out at you?
Wade: The first that comes to mind is the race to sequence the human genome. It was a good science story, but it was also of interest to general readers because of the rivalry between Celera and the university people.
Another story I found interesting was the genetics of human dispersal. We've had the picture of human origins as developed by the paleoanthropologists, and it's wonderful how well they've done with the material they've had available to them—just a handful of skulls. Then, the geneticists arrived on the scene and added a whole new dimension.
For example, there was a paper by Mark Stoneking about when man first started to wear clothes. He managed to figure out the date at which the human body louse, which lives in clothing, evolved from the human head louse, which lives just in hair. And the date of that divergence must give the time at which humans first started to wear reasonably close-fitting clothing. It's wonderful that genetics can provide that quite surprising insight.
Gitschier: My favorite recent story is Homo floresiensis.
Wade: Isn't that nice! The referees took a whole year, I think, to convince themselves that this was real. They started off by thinking this fossil must be a pathological Homo sapiens skull, but then realized it doesn't look like sapiens, so it must be erectus. But it was found with artifacts just like those made by modern humans. To assume the little Floresians made these artifacts contradicts almost everything that paleoanthropologists have been taught: that we didn't start to make tools until our brains were about twice the size of chimpanzees', which are approximately the size of Homo floresiensis. This is such a paradoxical finding!
I think the paleoanthropology community is going through the same learning process as the reviewers did. They started with the assumption that these were modern human artifacts and a pathological skull, but eventually came to accept that everything was the work of a downsized erectus.
Gitschier: I like it when people are forced to rethink their dogma! What about the other side of the coin—stories that you missed?
Wade: I think the main one in that category is RNAi [RNA interference], about which I've written only one story. I kept thinking, “This is fascinating, but the general reader won't be interested in the details of molecular biology, so let's wait till it advances more.” I think I was far too late.
Another thing that is very difficult for science reporters to tackle is the fact that most scientific research ends nowhere. People can be very enthusiastic about what they are doing, but just as most drugs fail in clinical trials, many advances that seem very promising don't lead anywhere. So after you have been mistaken a certain number of times, you tend to be a little cautious. Of course, it's then very easy to become far more skeptical than one should be.
Gitschier: Gene therapy, for example, is a field that many thought had promise. It had some successes and some spectacular failures.
Wade: That's a field that's been going on for about 15 years. And almost all the coverage throughout the first ten years kept saying gene therapy is great. But in retrospect, it was quite wrong—it wasn't great at all. There were technical obstacles that have still not been overcome. I think the lesson for reporters is that they should not get too caught up in scientists' enthusiasm. It's fine to report that scientists are enthused about some new finding or project, but reporters should remain detached about whether or not it will succeed.
Stem cells are a case in point. The hidden premise of proposals for stem cell therapy is that we needn't understand exactly what is going on because if you just put the cells in the right place they will know what to do. My fear is that we need to understand the total cell circuitry to get stem cells to do anything useful, and that won't happen for years.
Gitschier: By choosing to write up certain stories and ignoring others, you are making judgments. Are there things out there that you are not writing about because you simply don't agree with them?
Wade: The only criteria that reporters are trained to apply is “Is this news?” So it doesn't matter if you agree with it or not.
Gitschier: But is that the ethical thing to do?
Wade: If someone makes a newsworthy claim that I suspect is not true, I will try to see if there are skeptics and expose readers to both sides of the issue. A reporter's job is to give readers sufficient information to make up their own minds. In a news story, you should expose people to all the possibilities, but you don't have to decide which one is correct. The hard thing about writing editorials is that you have to decide.
Gitschier: It's such a responsibility, I would think.
Wade: If you try to figure out the consequences of every article, you'd never write anything.
Gitschier: Returning then to the question of editorial writing, what were some of the memorable topics that you had to write about?
Wade: There weren't that many scientific issues about which we could have an editorial opinion, since many issues in science are a matter of ascertainable fact, not opinion.
I was writing editorials during the Reagan administration, so there were many environmental issues to inveigh about, inspired by the likes of Ann Gorsuch and James Watt. During the Reagan military buildup, I also wrote many editorials about military hardware and procurement scandals. I remember having great difficulty making up my mind about a “big science” project dear to physicists, the superconducting supercollider. I wrote one editorial in favor of it, the next year one against it, and the third year one in the middle. Editorial writers have to do their learning in public, or at least I did. Life is much easier once you have developed your position on the issues.
Gitschier: Today, there would seem to be a lot of opportunities for editorial writing in genetics—embryonic stem cells, cloning, reproductive choices, and intelligent design, to name a few.
Wade: I think you're right. Of course, stem cells would be less of an issue if the government hadn't tried to restrict the research. Intelligent design is a good subject for editorials, though not, I think, for the science section because it has no scientific content. It's a debate that was settled in the 19th century. It's not our role to educate people, and I see no more reason to discuss whether intelligent design is an alternative to evolution than to discuss whether or not the earth is flat.
Gitschier: What about the urgency to write things?
Wade: There is nothing like a deadline for concentrating your mind. Some of the hardest stories are when you are asked to get a story at very short notice, such as late at night when the editors see the Washington Post has some story, and ask you to match it. If you don't have the home numbers of the people you need to talk to, you're out of luck. Most reporters know their beat well enough that they can match a story at short notice. Fortunately, it doesn't happen too often.
Gitschier: One of the things I like about my job as a geneticist is that there is always something new on the horizon. You must feel the same way.
Wade: Yes, and journalists have the luxury of being able to move from one field to another. If it's a slow week in genetics, I can write about cognitive science.
If you're not learning something new every day, you have no one but yourself to blame.
|
16205791
|
PMC1239940
|
CC BY
|
2021-01-05 07:59:37
|
no
|
PLoS Genet. 2005 Sep 30; 1(3):e45
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010045
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1619077910.1371/journal.pmed.0020264Research ArticleGenetics/Genomics/Gene TherapyMolecular Biology/Structural BiologyPharmacology/Drug DiscoveryCardiology/Cardiac SurgeryClinical PharmacologyPathologyGeneral MedicineCardiovascular MedicineGeneticsPathologyEffects of ADMA upon Gene Expression: An Insight into the Pathophysiological Significance of Raised Plasma ADMA Effects of ADMA upon Gene ExpressionSmith Caroline L
1
*Anthony Shelagh
1
Hubank Mike
1
Leiper James M
1
Vallance Patrick
1
1Centre for Clinical Pharmacology and Therapeutics, Division of Medicine, University College London, London, United KingdomBenjamin Ivor Academic EditorUniversity of UtahUnited States of America*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: University College London holds patents on DDAH as a drug target (not relevant to the present study).
Author Contributions: CLS, JML, and PV designed the study. CLS, SA, MH, JML, and PV analyzed the data. CLS, MH, JML, and PV contributed to writing the paper.
10 2005 4 10 2005 2 10 e2641 1 2005 30 6 2005 Copyright: © 2005 Smith et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
ADMA in Vascular Disease: More than a Marker?
Background
Asymmetric dimethylarginine (ADMA) is a naturally occurring inhibitor of nitric oxide synthesis that accumulates in a wide range of diseases associated with endothelial dysfunction and enhanced atherosclerosis. Clinical studies implicate plasma ADMA as a major novel cardiovascular risk factor, but the mechanisms by which low concentrations of ADMA produce adverse effects on the cardiovascular system are unclear.
Methods and Findings
We treated human coronary artery endothelial cells with pathophysiological concentrations of ADMA and assessed the effects on gene expression using U133A GeneChips (Affymetrix). Changes in several genes, including bone morphogenetic protein 2 inducible kinase (BMP2K), SMA-related protein 5 (Smad5), bone morphogenetic protein receptor 1A, and protein arginine methyltransferase 3 (PRMT3; also known as HRMT1L3), were confirmed by Northern blotting, quantitative PCR, and in some instances Western blotting analysis to detect changes in protein expression. To determine whether these changes also occurred in vivo, tissue from gene deletion mice with raised ADMA levels was examined. More than 50 genes were significantly altered in endothelial cells after treatment with pathophysiological concentrations of ADMA (2 μM). We detected specific patterns of changes that identify pathways involved in processes relevant to cardiovascular risk and pulmonary hypertension. Changes in BMP2K and PRMT3 were confirmed at mRNA and protein levels, in vitro and in vivo.
Conclusion
Pathophysiological concentrations of ADMA are sufficient to elicit significant changes in coronary artery endothelial cell gene expression. Changes in bone morphogenetic protein signalling, and in enzymes involved in arginine methylation, may be particularly relevant to understanding the pathophysiological significance of raised ADMA levels. This study identifies the mechanisms by which increased ADMA may contribute to common cardiovascular diseases and thereby indicates possible targets for therapies.
Pathophysiological concentrations of asymmetric dimethylarginine elicit significant changes in coronary artery endothelial cell gene expression and highlight specific molecular pathways for further investigation.
==== Body
Introduction
Asymmetric dimethylarginine (ADMA) is an endogenous inhibitor of all nitric oxide synthase (NOS) isoforms [1]. It is synthesised by the action of protein arginine methyltransferases (PRMTs), and following proteolysis, free ADMA is released into the cell cytosol and thence into plasma. Circulating concentrations of ADMA are increased in patients with renal failure [1], pulmonary hypertension, heart failure, hypercholesterolemia or a wide range of other cardiovascular risk factors [2–6]. In patients with end-stage renal failure, the plasma levels of ADMA predict mortality and cardiovascular outcome [7], and in a cohort of otherwise healthy Finnish men, those with the highest levels of ADMA had an increased risk of acute coronary events [8]. Increased circulating ADMA in pregnant women predicts an increased risk of pre-eclampsia and intrauterine growth retardation [9].
Despite these clinical observations and the increasing excitement surrounding the use of ADMA as a risk marker for vascular disease [3,7], it is still not clear whether ADMA has a causal role in pathophysiology. It has been argued that the concentration of ADMA in plasma is too low to be an effective inhibitor of NOS, and that the usual concentrations of arginine in cells should overcome any inhibitory effects of ADMA on NOS [10]. In order to determine how ADMA might exert effects on endothelial cells and produce pathology, we assessed the effects of ADMA on gene expression in human coronary endothelial cells.
Methods
Cell Culture
Human coronary artery endothelial cells (HCAEC) were purchased from Promocell and grown according to the manufacturer's instructions. HCAEC in 75-cm2 flasks at 70% confluency (passage 3 or 4) were treated for 24 h with complete media supplemented with asymmetric dimethylarginine (NGNG-dimethyl-L-arginine; ADMA; 0, 2, or 100 μM; Merck Biosciences, United Kingdom). This was repeated on three separate occasions with different batches of cells. RNA from each study was used as described below for GeneChip (Affymetrix, Santa Clara, California, United States) analysis. Our strategy for the GeneChip and subsequent analysis is outlined in Figure 1.
Figure 1 Flow Diagram Summarising the Methods Used in This Manuscript
GeneChip Experiments
ADMA-treated HCAEC from T75 flasks were harvested in 7.5 ml of TRIzol (Invitrogen, Carlsbad, California, United States), and total RNA was extracted; cDNA and subsequent cRNA synthesis were prepared as previously described [11]. The quality of the biotin-labelled cRNA transcripts was determined using a Bioanalyser 2100 (Agilent Technologies, Palo Alto, California, United States). Purified cRNA (15 μg) was fragmented and hybridised to human U133A GeneChips according to Affymetrix standard protocols (http://www.affymetrix.com). Labelled GeneChips were scanned, using a confocal argon ion laser (Agilent Technologies).
GeneChip Data Analysis
The U133A GeneChip contains oligonucleotides derived from approximately 22,000 human transcripts and includes control bacterial genes bioB, bioC, bioD, and cre. GeneChip data files scaled to 100 and normalised to the median prior to analysis with GeneSpring 7.2 software (Agilent; Figure 1). Genes were excluded if the signal strength did not significantly exceed background values and if expression did not reach a threshold value for reliable detection (based on the relaxed Affymetrix MAS 5.0 probability of detection (p ≤ 0.1; [12]) in each of the three separate studies. Finally, genes were excluded if the level of expression did not vary by more than 1.7-fold between ADMA-treated (2 or 100 μM) compared with untreated control HCAEC. The remaining genes were subjected to nonparametric Welch t-tests and are reported with their respective fold changes and p-values. The data have been submitted in a MIAME-compliant format to ArrayExpress at EBI (http://www.ebi.ac.uk/arrayexpress/).
Determination of ADMA Levels
HCAEC cells were grown in 75-cm2 flasks as described above for 24 h. Methylated arginines in the conditioned medium were quantitated by HPLC as previously described [1,13].
Cell Growth Assay
Cells were seeded into a 96-well microtitre plate at a density of 500 cells per well and grown in complete EC medium in the presence of 0, 2, or 100 μM ADMA. Cell growth was assayed (in triplicate) over a 4-d period using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, Wisconsin, United States).
Confirmation of Gene Changes
For genes of interest selected on the initial GeneChip analysis, further experiments were undertaken, with different batches of cells, to verify changes detected by global expression profiling and using different approaches to assess mRNA or protein.
RT-PCR
For RT-PCR cDNA was synthesised from the total RNA (approximately 1 μg) using Ready-To-Go You-Prime First-Strand Beads (Amersham Biosciences, Little Chalfont, United Kingdom) and supplemented with the 3′ gene-specific primers with β-actin as a control. RT-PCR used the primer sequences (Table 1) with β-actin as a control.
Table 1 Primer Sequences for Q-PCR and RT-PCR
The PCR products were separated in 1% agarose gel, and the intensity of the bands was measured; each sample was corrected for β-actin. PCR products used for Northern blotting were excised and purified from agarose gel using Qiaex II (Qiagen, Valencia, California, United States).
Quantitative-PCR
Quantitative PCR (Q-PCR) was carried out using a LightCycler (Roche, Alameda, California, United States) and LightCycler software version 3, LightCycler run 4.24. All reagents required for PCR (excluding cDNA and primers) were included in the LightCycler FastStart DNA Master SYBR Green 1 kit (Roche). Reverse transcription was performed as described above for RT-PCR. The PCR cycle settings were 95 °C for 5 min, followed by 45 cycles of 95 °C for 5 s, 58 °C for 10 s (60 °C for β-actin), 72 °C for 40 s, and 77 °C (82 °C for β-actin), where fluorescence was measured at the end of each cycle and is gene-specific. Standard curves were constructed for PRMT3 as described by the manufacturer's instructions and compared to the reference gene β-actin.
Northern Blotting
HCAEC were grown to 70% confluency in 6-well plates and treated with 0, 2, and 100 μM ADMA, 100 μM L-N5-(1-Iminoethyl)ornithine (L-NIO), or 100 μM symmetric dimethylarginine (SDMA) for 24 h; total RNA was extracted using TRIzol, and RNA was separated by gel electrophoresis and blotted onto Hybond N+ (Amersham Biosciences). The ribosomal protein S11 (RpS11), ribosomal protein L-27 (RpL-27), Calreticulin, and secretory carrier membrane protein 1 (SCAMP1) probes were generated by end-labelling 5′ oligonucleotides (Table 2) using T4 polynucleotide kinase.
Table 2 Oligonucleotide Sequences for 5′-End-Labelling Northern Blotting Experiments
The PCR products from RT-PCR reaction for β-actin, Smad5, bone morphogenetic protein receptor 1A (BMPR1A), and murine β-actin (mβ-actin); murine bone morphogenetic protein 2 inducible kinase (mBMP2K), and murine protein arginine methyltransferase 3 (mPRMT3) were purified and labelled with Redivue deoxycytidine 5′-[α-32P]-triphosphate (Amersham Biosciences), using a random primed DNA labelling kit (Roche).
Western Blotting
HCAEC were treated with either 0, 2, or 100 μM ADMA for 24 h in 6-well plates and harvested in lysis buffer as described previously [11], the protein concentrations of the lysates were determined by protein assay (Bio-Rad, Hercules, California, United States), and cell lysates were resolved by 12% SDS polyacryamide gel electrophoresis with equal amounts of protein loaded into each lane. Anti-PRMT3 (Upstate Biotechnology, Charlottesville, Virginia, United States) and Anti-BMP2K (Orbigen, San Diego, California, United States) were used with anti-rabbit secondary antibody coupled to horseradish peroxidase and detected with the ECL+ detection system (Amersham Pharmacia, Piscataway, New Jersey, United States). Densitometry of the bands was determined, and results are shown as the mean densitometry, where n = 4 with inset of a typical Western blot.
Pathway Mapping and Gene Ontology Analysis
In order to determine whether ADMA had affected expression of genes in pathways related to the genes identified on the initial analysis, lists of genes that were changed more than 1.7-fold compared to control (irrespective of p-value) were examined using Gene Ontology (Affymetrix) data mining for biological process (at level 3), and Expression Analysis Systematic Explorer (EASE) biological theme analysis were conducted online at http://david.niaid.nih.gov using DAVID [14]. DAVID-EASE [15] generates an EASE score predicting the likelihood of genes mapping to specific biological processes (determined by Gene Ontology consortium) from a given list of changed genes, therefore enabling global themes in gene expression following ADMA treatment to be identified [15].
Dimethylarginine Dimethylaminohydrolase 1 Gene Deletion Mice
ADMA is metabolised to citrulline and dimethylamine by the action of dimethylarginine dimethylaminohydrolase (DDAH). We have created knockout mice that lack DDAH1. DDAH1 heterozygous knockout mice (details to be published elsewhere) have approximately 2-fold higher plasma ADMA levels compared to wild-type litter-mates and thus provide an excellent model to test the effects of moderately raised ADMA levels in vivo. Northern blotting was carried out using RNA extracted from the brain, heart, and kidney of 12–14-wk-old DDAH1 heterozygous knockout and wild-type litter-mates with probes for mBMP2K and mPRMT3, and results are expressed relative to mβ-actin.
Statistical Analyses
Q-PCR, Northern blot, and Western blot densitometry data for the treated HCAEC was analysed by one-way analysis of variance (ANOVA) coupled to Bonferroni posttest, and the Bonferroni posttest p-values are reported. The Northern blots from the DDAH1 gene deletion mice were compared with unpaired t-test and the p-values are reported.
Results
Changes in HCAEC Gene Expression in Response to ADMA on U133A GeneChips
A total of 979 genes changed in expression between ADMA-treated and control cells. Following the Welch t-test with a cutoff of p < 0.05, 56 genes were identified as having shown a statistically significant change between the untreated and 2-μM ADMA-treated cells, and 86 genes changed between the untreated and 100-μM ADMA greater than 1.7-fold; 11 genes showed statistically significant changes at both concentrations of ADMA compared to untreated cells (Figure 2A; Tables 3 and 4).
Figure 2 Endothelial Gene Expression Changes in Response to ADMA
(A) Changes in HCAEC gene expression in response to ADMA. Hierarchical clustering of 131 genes significantly up- or downregulated (p < 0.05, Welch t-test) by greater than 1.7-fold on the U133A GeneChips in response to 2 μM ADMA (56 genes) or 100 μM ADMA (86 genes) compared with untreated cells with 11 genes changing with both concentrations of ADMA. Individual arrays are shown for each treatment group with blue representing low expression, and red high expression on the scale bar.
(B) Treatment of HCAEC with either 2 μM or 100 μM ADMA had no significant effect on cell viability; n = 6.
Table 3 Genes That Changed by More Than 1.7-Fold in 2 μM ADMA Treated HCAEC Compared to Untreated
Table 4 Genes that Changed by More than 1.7-Fold in 100 μM ADMA-Treated HCAEC Compared to Untreated
Table 4 Continued
Effects of ADMA upon HCAEC Viability
Basal levels of ADMA and SDMA in HCAEC media were 0.17 ± 0.01 μM and 0.22 ± 0.02 μM, respectively, and arginine levels exceeded 300 μM. No changes were observed in HCAEC viability over 72 h in the presence of either 2 or 100 μM ADMA (Figure 2B).
Confirming Transcriptional Changes
To determine the reliability of changes identified by GeneChip analysis, four genes were selected from those that showed a statistically significant change in either the 2 or 100 μM sample compared to control (Figure 3A). These were SCAMP1, Calreticulin, (RpL27) and RpS11. In studies on a different batch of HCAEC, Northern blotting confirmed that expression of these genes changed in response to ADMA (Figure 3B).
Figure 3 Confirmation of Gene Expression Changed by GeneChip Analysis
(A) SCAMP1; Calreticulin, ribosomal protein L-27 (RpL27), and RpS11 signal-to-noise ratios derived from U133A GeneChip analysis where n = 3 (untreated versus 100 μM: SCAMP1 changed 3.16-fold, p = 0.031; and untreated versus 2 μM: Calreticulin changed 2.96-fold, p = 0.047; RpS11, 8.065-fold, p = 0.033; RpL27, 4.39-fold, p = 0.048).
(B) SCAMP1; Calreticulin, RpL27, and RpS11 mRNA levels are elevated by ADMA (2 and 100 μM; *p < 0.05 and **p < 0.01). SDMA (100 μM) and L-NIO (100 μM) did not elicit changes in gene expression as shown by Northern blotting, where mRNA was corrected for differences in β-actin mRNA expression.
In order to elucidate the mechanism of ADMA action, HCAEC were also treated with the potent NOS inhibitor L-NIO and SDMA, which is not a NOS inhibitor or DDAH substrate, but is a naturally occurring methylarginine that competes with arginine for the cationic amino acid transporter [10,16]. Interestingly neither SDMA nor L-NIO elicited significant changes in the expression of RpS11, RpL27, SCAMP1, or Calreticulin (Figure 3B).
Genes of Specific Interest Identified by GeneChip
Protein arginine methyltransferase
PRMT3 was selected as a gene of interest, because of its involvement in ADMA synthesis [17]. It changed 1.8-fold on the GeneChip untreated versus 100 μM groups (p = 0.0339; Figure 4A). In a separate series of follow-up studies, HCAEC were treated with ADMA, and PRMT3 gene expression was determined by Q-PCR. PRMT3 mRNA levels increased following either low- or high-dose ADMA treatment (Figure 4B untreated versus 2 μM, p < 0.05; and untreated versus 100 μM, p < 0.01, n = 6). PRMT3 protein expression also increased in response to ADMA treatment (Figure 4C; untreated versus 2 μM; and untreated versus 100 μM, p < 0.05, n = 4). When HCAEC were treated for 24 h with 2 μM SDMA or L-NIO, neither treatment significantly affected PRMT3 expression, whereas 100 μM SDMA or L-NIO caused a small increase in PRMT3 expression (Figure 4B; n = 6).
Figure 4 ADMA Alters PRMT3 Gene Expression and Protein Levels
(A) PRMT3 levels are changed by ADMA on U133A GeneChips where n = 3 (untreated versus 100 μM: 1.88-fold increase, p = 0.034).
(B) PRMT3 mRNA is changed in HCAEC following 24-h exposure to ADMA (2 and 100 μM) but not SDMA (100 μM) or L-NIO (100 μM), measured by Q-PCR (*p < 0.05 and **p < 0.01, where n = 8).
(C) PRMT3 protein levels are increased in HCAEC following 24-h treatment with ADMA (2 and 100 μM) as determined by Western blotting, using a commercially available PRMT3 antibody (Upstate Biotechnology), where equal amounts of protein were loaded in each lane. Densitometry of the PRMT3 was carried for each of the blots from four separate experiments. ADMA (2 and 100 μM) treatment for 24 h significantly increased the levels of PRMT3 (*p < 0.05). The inset blot is a representative from four separate experiments.
Bone morphogenetic protein 2 inducible kinase
We identified that bone morphogenetic protein 2 inducible kinase (BMP2K) changed on the GeneChip in response to ADMA (2 μM and 100 μM ADMA increased expression by 2.304-fold [p = 0.00128] and 2.695-fold [p < 0.001], respectively; Figure 5A). In a separate series of experiments on a different batch of HCAEC this increase in BMP2K expression was confirmed by RT-PCR (data not shown). Western blotting also revealed that BMP2K protein levels were increased in response to ADMA (Figure 5B); densitometry of these blots indicated that there was a significant increase (untreated versus 2 μM, p < 0.05; and untreated versus 100 μM, p < 0.01, n = 4).
Figure 5 ADMA Alters BMP2K Gene Expression and Protein Levels
(A) BMP2K is increased more than 2-fold in HCAEC following 24-h ADMA treatment (2 μM and 100 μM ADMA increased expression by 2.304-fold [p = 0.00128] and 2.695-fold [p < 0.001] respectively).
(B) BMP2K protein levels are increased in HCAEC following 24-h treatment with ADMA (2 and 100 μM) as determined by Western blotting, using a commercially available BMP2K antibody (Orbigen). Densitometry of the BMP2K band was carried for each of the blots from four separate experiments (untreated versus 2 μM: p < 0.05; and untreated versus 100 μM: p < 0.01, n = 4). The inset blot is representative of four separate experiments, where equal amounts of protein were loaded in each lane.
Identification of Genes Involved in Bone Morphogenetic Protein Signalling
Having confirmed changes in BMP2K expression, the total list of 765 genes that changed greater than 1.7-fold in response to 100 μM ADMA, was reexamined to identify additional genes in the bone morphogenetic protein (BMP) signalling pathway affected by ADMA. This analysis identified Smad5 and BMPR1A (Figures 6 and 7). In a separate set of studies Northern blotting confirmed that mRNA was increased in HCAEC after 24 h by either 2 or 100 μM ADMA for Smad5 and BMPR1A (Smad5 untreated versus 2 μM and untreated versus 100 μM; p < 0.05, n = 4; Figure 6A; and BMPR1A untreated versus 2 μM, p < 0.05 and untreated versus 100 μM; p < 0.01, n = 4; Figure 6B).
Figure 7 Involvement of ADMA-Affected Genes in the BMP Signalling Pathway
Squares in yellow show genes that increased significantly on the GeneChip, whilst squares in blue show genes decreased by ADMA compared to untreated HCAEC, and “?” represents an unknown mechanism of action.
Figure 6 Genes Involved in BMP Signalling are Altered by ADMA
Smad5 (A) and BMPR1A (B) are elevated in HCAEC following 24-h treatment with ADMA (2 and 100 μM), as shown by Northern blotting where mRNA was corrected for differences in β-actin mRNA expression (*p < 0.05; ** p < 0.01, where n = 4).
Identification of Global Changes in Gene Expression
All genes that changed by more than 1.7-fold (irrespective of p-value) were entered into DAVID-EASE. EASE probability scores were generated based upon the number of genes for each biological process altered in response to ADMA (Tables 5 and 6), where these biological processes were defined by enriched Gene Ontology categories. These gene lists indicated that ADMA affects genes involved in metabolism, RNA splicing, transcription, and cell cycle regulation.
Table 5 Identification of Global Changes in Gene Expression in HCAEC Following Treatment with 2 μM ADMA
Table 6 Identification of Global Changes in Gene Expression in HCAEC Following Treatment with 100 μM ADMA
Gene Deletion Mice
To examine whether the effects observed in the cell culture model were relevant to the in vivo situation, we determined the expression level of certain genes in DDAH heterozygous knockout mice that have 2-fold elevation in plasma ADMA levels. Total RNA from DDAH1 heterozygous knockout was probed for mBMP2K and mPRMT3 by Northern blotting and corrected for mβ-actin expression (Figure 8). Levels of mBMP2K for brain, heart, and kidney, respectively, were 42 ± 13.8% (p = 0.047), 33.3 ± 18.6% (p = 0.038), 74.0 ± 21.4% (p = 0.007), higher in DDAH1 heterozygous mice (n = 9) compared with wild-type litter-mates (n = 5). A similar trend was seen for the expression of mPRMT3 (data not shown).
Figure 8 BMP2K Is Increased in DDAH1 Gene Deletion Mice
Expression of BMP2K mRNA from brain, heart, and kidney is increased in 12-wk DDAH1 heterozygous knockout mice compared to wild-type litter-mates (p = 0.0473, p = 0.0379, and p = 0.0070, respectively); mRNA was corrected for differences in β-actin mRNA expression, where n = 5 wild type, and n = 9 DDAH heterozygous.
Discussion
ADMA is an endogenous inhibitor of NOSs [1] and there is an association between increased plasma levels of ADMA and renal disease [1], pulmonary hypertension [5], preeclampsia [9], and the progression of atherosclerosis [18,19]. Whilst the concentration of ADMA in plasma of healthy adults varies between 0.4 and 1 μM, it may increase to 1.45–4.0 μM with certain diseases, and this increase is thought to be causally involved in pathophysiology [1,6,7,9,20]. In the present study we detected substantial changes in gene expression in HCAEC after 24 h of exposure to concentrations of ADMA similar to those reported in pathophysiological states. Furthermore, we identified specific pathways of gene activation that give insight into the mechanisms by which ADMA may contribute to disease. Surprisingly, some of these changes appear to be independent of blockade of the L-arginine:nitric oxide (NO) pathway.
Low Concentrations of ADMA Alter Gene Expression
Acute administration of ADMA to healthy individuals elicits a transient fall in heart rate and cardiac output and increases blood pressure [2], but little is known of the potential longer-term effects of raised ADMA. Zoccali et al. reported that a 1-μM increment in ADMA above the upper limits for healthy individuals was associated with increased risk of cardiovascular mortality [7], and levels around 2 μM seem to be associated with a number of diverse cardiovascular pathologies [3,8]. In the current study, HCAEC were treated with 2 or 100 μM ADMA. We repeated GeneChip analysis in three separate studies and observed reproducible changes in the expression of a subset of genes in response to low- and high -dose ADMA. Because there is always a possibility of false positives being identified on arrays, we tested the reproducibility of the GeneChip approach by selecting four genes significantly upregulated by ADMA treatment. In a separate series of studies we confirmed an increase in mRNA levels by Northern blotting at both concentrations of ADMA. Thus it is clear that pathophysiological concentrations of ADMA (2 μM) affect endothelial cell gene expression even in the presence of very high arginine concentrations (>300 μM). Endogenous ADMA is metabolised by DDAH, and DDAH activity is the major determinant of plasma ADMA concentrations [2]. To determine whether the effects we saw in vitro would be reproduced in vivo we examined DDAH1 knockout mice. At least one of the gene changes we saw in vitro also occurs in vivo since DDAH1 heterozygous knockout mice have increased concentrations of ADMA in plasma and showed upregulation of BMP2K in several tissues.
ADMA inhibits NOSs with an IC50 of about 5 μM, the precise potency depending on the prevailing concentration of arginine [21]. It is known that adding endogenous NO to endothelial cells alters gene expression [22,23] and that inhibitors of endogenous NO generation can alter expression of specific genes, at least under conditions of endothelial cell activation with cytokines. To determine whether the effects we observed could be accounted for by inhibition of NOS, we treated cells with a highly potent NOS inhibitor (L-NIO) and with SDMA, an endogenous dimethylarginine that has no effect on NOS but which can block arginine transport [10,16]. Neither SDMA nor L-NIO replicated the effects of ADMA on gene expression. We have not undertaken a full GeneChip analysis of responses to L-NIO, so we do not know how much overlap there would be between the effects of L-NIO and ADMA, but of those genes examined we saw a discordance between responses to the two inhibitors. This raises the intriguing possibility that some of the actions of ADMA may be independent of effects of NO, possibly due to other actions such as the ability of ADMA to increase superoxide generation [24,25]. Other authors have reported differences in efficacy of ADMA in cell culture systems compared to other NOS inhibitors that have more potent effects on isolated NOS [26], and further studies will be required to identify to what extent NO-independent effects contribute to the overall action of ADMA. Interestingly, a NOS-independent effect of ADMA on angiotensin-converting enzyme has recently been suggested [25]. NOS-independent effects of ADMA may explain why, in some situations where ADMA concentrations are raised, L-arginine does not restore normal endothelial function [27] and why ADMA can exert effects, even in the presence of high endogenous arginine concentrations.
Patterns of Gene Change
Mapping genes changed by ADMA to identify global changes in biological processes [15], indicated that ADMA treatment may have significant effects on genes involved in cell cycle regulation, cell proliferation, DNA repair, transcriptional regulation, and metabolism. The full biological significance of the range of genes affected is not yet known, but our data demonstrate the potential for elevated ADMA to affect endothelial (and likely other cellular) function in disease. In the present study, we focussed on two pathways of potential importance—BMP pathways and PRMTs.
BMP Pathways and ADMA
Analysis of the U133A GeneChips revealed that BMP2K was induced more than 2-fold in response to either 2 or 100 μM ADMA. The increase in gene expression was mirrored by an increase in BMP2K protein, and the effect was also seen in our high-ADMA mouse model. By relaxing parameters (to exclude false negatives) a search for other genes involved in BMP signalling revealed that Smad5 and BMPR1A were also amongst the transcripts increased by GeneChip analysis, and these changes were confirmed by Northern blotting. The finding of changes in the BMP signalling pathway is important since the ADMA/DDAH pathway seems to be involved in animal models of pulmonary hypertension [28,29], and mutations in the BMP receptor 2 (BMPR2) are associated with familial pulmonary hypertension in humans [30].
The changes in BMP pathways may also be important in understanding some of the effects of renal failure. ADMA accumulates in renal failure and fulfils many of the criteria of a uraemic toxin [1,18]. In addition to effects on cardiovascular risk, ADMA may contribute to renal osteodystrophy, a process in which BMPs have been implicated [31]. Indeed an earlier study that showed that ADMA reduces osteoblast differentiation and decreases osteocalcin expression [32]. The present study confirms osteocalcin as a gene downregulated by ADMA, and since activation of BMP2K attenuates osteocalcin expression and reduces osteoblast differentiation [33], it is possible that the effects of ADMA on osteocalcin may be secondary to induction of BMP2K (Figure 7). Whatever the mechanisms, identification of a link between ADMA and BMP pathways may be relevant to the increased vascular calcification seen in renal disease [31].
ADMA and Arginine Methylation
We observed that PRMT3 gene expression was elevated following exposure to ADMA; this was confirmed by Q-PCR, and PRMT3 protein expression also increased. This is the first report that ADMA may alter the expression of enzymes involved in its own synthesis. There are presently five known PRMTs that asymmetrically methylate arginine residues and two (PRMT5 and PRMT7) that symmetrically methylate arginine residues. PRMT3 has a wide tissue distribution, is expressed in highly vascular tissues, including heart and lung [17], and expression may be increased by oxidised-LDL [34]. Our observations indicate that ADMA can induce a similar increase in PRMT3 expression.
The roles of methylation of arginine residues in proteins are not yet well defined, but studies of PRMT3 in fission yeast have shown that it associates with proteins involved in the translational machinery and that the S2 ribosomal protein, a component of the yeast 40S ribosome, is a specific substrate for PRMT3 [35]. PRMT3 is the only PRMT known to interact with the translational machinery, and it is interesting that we have found several genes, including ribosomal proteins RpS11 and RpL27, involved in translational control, that were also altered in response to ADMA treatment. Amongst the genes changed greater than 1.7-fold in response to ADMA was methionine adenosyltransferase II α, which catalyses the production of S-adenosylmethionine, the methyl donor for the PRMT reaction [36]. The role of ADMA in regulating arginine methylation in protein deserves further study.
Summary
Increased circulating concentrations of ADMA have been reported in cardiovascular and other disorders, and intracellular concentrations may vary independently of circulating levels. In the present study we have demonstrated that relatively small changes in the concentration of ADMA affect gene expression in endothelial cells. Identification of pathways regulated by ADMA may aid our understanding of how ADMA contributes to a wide range of pathologies. Two pathways of specific interest have been identified—BMP signalling and enzymes involved in arginine methylation. The effects on BMP signalling may be particularly important in renal disease and in the link between raised ADMA and pulmonary hypertension.
Supporting Information
Accession Numbers
The microarray data have been loaded into the EBI MIAMExpress database (http://www.ebi.ac.uk/miamexpress/) and have been assigned the accession number E-MEXP-377.
Patient Summary
Background
Diseases of the circulation system are common and cause many deaths. Medical conditions associated with damage to the blood vessels include heart failure, high blood pressure, stroke, and kidney failure. The lining of the blood vessels plays an active role in maintaining their health. A substance called asymmetric dimethylarginine (ADMA) is found naturally in the vessel lining, both in healthy people and in people with vascular disease, but in the latter it is present at higher levels. Thus raised ADMA may be a marker of vascular disease. This means it could be used to help identify people with a circulation problem. However, it has not been clear whether ADMA actually causes any damage, i.e., whether it is more than just a marker.
What Did the Researchers Do and Find?
The researchers are trying to find out whether elevated ADMA levels can cause vascular disease. In this study, they treated cells from blood vessel linings with levels of ADMA equal to those found in people with vascular disease and measured how gene activity changed in response. They found that a number of genes were more active when the cells were exposed to the elevated ADMA levels. Some of these were interesting because other studies suggest that they might be involved in lung, heart, and kidney disease.
What Do the Results Mean for Patients?
This area of research is still at an exploratory stage. Additional studies need to examine which function (if any) the genes that respond to elevated ADMA levels play in vascular disease. If they do play active roles, drugs that inhibit them might help to prevent or treat vascular disease.
Where Can I Get More Information?
For general information on cardiovascular disease see information provided by the following organisations.
The British Heart Foundation:
http://www.bhf.org.uk/hearthealth/index_home.asp?SecID=1
The American Academy of Family Physicians:
http://familydoctor.org/292.xml
The National Heart, Lung, and Blood Institute:
http://www.nhlbi.nih.gov/health/public/heart/index.htm
New York Online Access to Health:
http://www.noah-health.org/en/blood/vascular/index.html
University College London:
http://www.ucl.ac.uk/medicine/clinical-pharmaco/research
The authors would like to thank Dr. R. C. Chambers (University College London) and Ms. D. Fletcher (Institute of Child Health) for their help with the GeneChip analysis and Dr. M. Malaki and Dr. M. Nandi for their assistance with the transgenic mice. This work was funded by British Heart Foundation Grant RG 2000/07. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Citation: Smith CL, Anthony S, Hubank M, Leiper JM, Vallance P (2005) Effects of ADMA upon gene expression: An insight into the pathophysiological significance of raised plasma ADMA. PLoS Med 2(10): e264.
Abbreviations
ADMAasymmetric dimethylarginine
BMPbone morphogenetic protein
BMPR1Abone morphogenetic protein receptor 1A
BMP2Kbone morphogenetic protein 2 inducible kinase
DDAHdimethylarginine dimethylaminohydrolase
HCAEChuman coronary artery endothelial cells
L-NIOL-N5-(1-Iminoethyl)ornithine
mβ-actinmurine β-actin
mBMP2Kmurine bone morphogenetic protein 2 inducible kinase
mPRMT3murine protein arginine methyltransferase 3
NOnitric oxide
NOSnitric oxide synthase
PRMTprotein arginine methyltransferase
Q-PCRquantitative PCR
RpL27ribosomal protein L27
RpS11ribosomal protein S11
SCAMP1secretory carrier membrane protein 1
SDMAsymmetric dimethylarginine
Smad5SMA-related protein 5
==== Refs
References
Vallance P Leone A Calver A Collier J Moncada S Accumulation of an endogenous inhibitor of nitric oxide synthesis in chronic renal failure Lancet 1992 339 572 575 1347093
Achan V Broadhead M Malaki M Whitley G Leiper J Asymmetric dimethylarginine causes hypertension and cardiac dysfunction in humans and is actively metabolized by dimethylarginine dimethylaminohydrolase Arterioscler Thromb Vasc Biol 2003 23 1455 1459 12805079
Cooke JP Asymmetrical dimethylarginine: The Uber marker? Circulation 2004 109 1813 1818 15096461
Usui M Matsuoka H Miyazaki H Ueda S Okuda S Increased endogenous nitric oxide synthase inhibitor in patients with congestive heart failure Life Sci 1998 62 2425 2430 9651109
Gorenflo M Zheng C Werle E Fiehn W Ulmer HE Plasma levels of asymmetrical dimethyl-L-arginine in patients with congenital heart disease and pulmonary hypertension J Cardiovasc Pharmacol 2001 37 489 492 11300662
Boger RH Bode-Boger SM Szuba A Tsao PS Chan JR Asymmetric dimethylarginine (ADMA): A novel risk factor for endothelial dysfunction: Its role in hypercholesterolemia Circulation 1998 98 1842 1847 9799202
Zoccali C Bode-Boger S Mallamaci F Benedetto F Tripepi G Plasma concentration of asymmetrical dimethylarginine and mortality in patients with end-stage renal disease: A prospective study Lancet 2001 358 2113 2117 11784625
Valkonen VP Paiva H Salonen JT Lakka TA Lehtimaki T Risk of acute coronary events and serum concentration of asymmetrical dimethylarginine Lancet 2001 358 2127 2128 11784629
Savvidou MD Hingorani AD Tsikas D Frolich JC Vallance P Endothelial dysfunction and raised plasma concentrations of asymmetric dimethylarginine in pregnant women who subsequently develop pre-eclampsia Lancet 2003 361 1511 1517 12737861
Tsikas D Boger RH Sandmann J Bode-Boger SM Frolich JC Endogenous nitric oxide synthase inhibitors are responsible for the L-arginine paradox FEBS Lett 2000 478 1 3 10922458
Chambers RC Leoni P Kaminski N Laurent GJ Heller RA Global expression profiling of fibroblast responses to transforming growth factor-beta1 reveals the induction of inhibitor of differentiation-1 and provides evidence of smooth muscle cell phenotypic switching Am J Pathol 2003 162 533 546 12547711
Seo J Bakay M Chen YW Hilmer S Shneiderman B Interactively optimizing signal-to-noise ratios in expression profiling: Project-specific algorithm selection and detection p -value weighting in Affymetrix microarrays Bioinformatics 2004 20 2534 2544 15117752
Teerlink T Nijveldt RJ de Jong S van Leeuwen PA Determination of arginine, asymmetric dimethylarginine, and symmetric dimethylarginine in human plasma and other biological samples by high-performance liquid chromatography Anal Biochem 2002 303 131 137 11950212
Dennis G Sherman BT Hosack DA Yang J Gao W DAVID: Database for Annotation, Visualization, and Integrated Discovery Genome Biol 2003 4 P3 12734009
Hosack DA Glynn Dennis G Sherman BT Lane BC Lempicki RA Identifying biological themes within lists of genes with EASE Genome Biol 2003 4 R70 14519205
Closs EI Basha FZ Habermeier A Forstermann U Interference of L-arginine analogues with L-arginine transport mediated by the y+ carrier hCAT-2B Nitric Oxide 1997 1 65 73 9701046
Tang J Gary JD Clarke S Herschman HR PRMT 3, a type I protein arginine N-methyltransferase that differs from PRMT1 in its oligomerization, subcellular localization, substrate specificity, and regulation J Biol Chem 1998 273 16935 16945 9642256
Kielstein JT Bode-Boger SM Frolich JC Haller H Boger RH Relationship of asymmetric dimethylarginine to dialysis treatment and atherosclerotic disease Kidney Int Suppl 2001 78 S9 S13 11168975
Zoccali C Endothelial damage, asymmetric dimethylarginine and cardiovascular risk in end-stage renal disease Blood Purif 2002 20 469 472 12207095
Stuhlinger MC Abbasi F Chu JW Lamendola C McLaughlin TL Relationship between insulin resistance and an endogenous nitric oxide synthase inhibitor JAMA 2002 287 1420 1426 11903029
MacAllister RJ Whitley GS Vallance P Effects of guanidino and uremic compounds on nitric oxide pathways Kidney Int 1994 45 737 742 7515129
Faruqi TR Erzurum SC Kaneko FT DiCorleto PE Role of nitric oxide in poly(I-C)-induced endothelial cell expression of leukocyte adhesion molecules Am J Physiol 1997 273 H2490 H2497 (5 Pt 2) 9374789
Hemish J Nakaya N Mittal V Enikolopov G Nitric oxide activates diverse signaling pathways to regulate gene expression J Biol Chem 2003 278 42321 42329 12907672
Cardounel AJ Xia Y Zweier JL Endogenous methylarginines modulate superoxide as well as nitric oxide generation from neuronal nitric oxide synthase: Differences in the effects of monomethyl and dimethyl arginines in the presence and absence of tetrahydrobiopterin J Biol Chem 2005 280 7540 7549 15574418
Suda O Tsutsui M Morishita T Tasaki H Ueno S Asymmetric dimethylarginine causes arteriosclerotic lesions in endothelial nitric oxide synthase-deficient mice: Involvement of renin-angiotensin system and oxidative stress Arterioscler Thromb Vasc Biol 2004 24 1682 1688 15217805
Smirnova IV Kajstura M Sawamura T Goligorsky MS Asymmetric dimethylarginine upregulates LOX-1 in activated macrophages: Role in foam cell formation Am J Physiol Heart Circ Physiol 2004 287 H782 H790 15016631
Cross JM Donald AE Kharbanda R Deanfield JE Woolfson RG Acute administration of L-arginine does not improve arterial endothelial function in chronic renal failure Kidney Int 2001 60 2318 2323 11737605
Arrigoni FI Vallance P Haworth SG Leiper JM Metabolism of asymmetric dimethylarginines is regulated in the lung developmentally and with pulmonary hypertension induced by hypobaric hypoxia Circulation 2003 107 1195 1201 12615801
Millatt LJ Whitley GS Li D Leiper JM Siragy HM Evidence for dysregulation of dimethylarginine dimethylaminohydrolase I in chronic hypoxia-induced pulmonary hypertension Circulation 2003 108 1493 1498 12952847
Deng Z Morse JH Slager SL Cuervo N Moore KJ Familial primary pulmonary hypertension (gene PPH1 ) is caused by mutations in the bone morphogenetic protein receptor-II gene Am J Hum Genet 2000 67 737 744 10903931
Hruska KA Saab G Chaudhary LR Quinn CO Lund RJ Kidney-bone, bone-kidney, and cell–cell communications in renal osteodystrophy Semin Nephrol 2004 24 25 38 14730507
Xiao ZS Quarles LD Chen QQ Yu YH Qu XP Effect of asymmetric dimethylarginine on osteoblastic differentiation Kidney Int 2001 60 1699 1704 11703587
Kearns AE Donohue MM Sanyal B Demay MB Cloning and characterization of a novel protein kinase that impairs osteoblast differentiation in vitro J Biol Chem 2001 276 42213 42218 11500515
Boger RH Sydow K Borlak J Thum T Lenzen H LDL cholesterol upregulates synthesis of asymmetrical dimethylarginine in human endothelial cells: Involvement of S-adenosylmethionine-dependent methyltransferases Circ Res 2000 87 99 105 10903992
Bachand F Silver PA PRMT3 is a ribosomal protein methyltransferase that affects the cellular levels of ribosomal subunits EMBO J 2004 23 2641 2650 15175657
Casellas P Jeanteur P Protein methylation in animal cells. II. Inhibition of S-adenosyl-L-methionine:protein(arginine) N-methyltransferase by analogs of S-adenosyl-L-homocysteine Biochim Biophys Acta 1978 519 255 268 667065
|
16190779
|
PMC1240048
|
CC BY
|
2021-01-05 10:38:14
|
no
|
PLoS Med. 2005 Oct 4; 2(10):e264
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020264
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1619077810.1371/journal.pmed.0020294Research ArticleAllergy/ImmunologyEpidemiology/Public HealthImmunology and allergyRespiratory MedicineAsthmaLatitude, Birth Date, and Allergy Latitude, Birth Date, and AllergyWjst Matthias
1
*Dharmage Shyamali
2
André Elisabeth
1
Norback Dan
3
Raherison Chantal
4
Villani Simona
5
Manfreda Jure
6
Sunyer Jordi
7
Jarvis Deborah
8
Burney Peter
8
Svanes Cecilie
9
1Gruppe Molekulare Epidemiologie, Institut für Epidemiologie, Forschungszentrum für Umwelt und Gesundheit, Munich, Germany,2Department of Public Health, School of Population Health, The University of Melbourne, Carlton, Victoria, Australia,3Department of Medical Science, Occupational and Environmental Medicine, Uppsala University, Uppsala, Sweden,4Service des Maladies Respiratoires, Centre François Magendie, Hôpital du Haut-Lévêque, Pessac, France,5Università egli Studi di Pavia, Dip. di Scienze Sanitarie Applicate e Psicocomportamentali, Pavia, Italy,6University of Manitoba, Department of Medicine, Winnipeg, Manitoba, Canada,7Respiratory and Environmental Health Research Unit, Municipal Institute of Medical Research, Barcelona, Spain,8Department of Public Health Sciences, Capital House, London, Great Britain,9Det medisinske fakultet, Lungemedisin, Universitet I Bergen, Bergen, Norway,Barnes Peter Academic EditorNational Heart and Lung InstituteUnited Kingdom*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: All authors participated in the local study management, DJ and PB planned and organised the study centre. MW developed the hypothesis for this paper, conducted the analysis, and wrote the first draft of the paper following two working group meetings in Malaga 2003 and in Bergen 2004, chaired by CS. All authors participated in the local study management, commented on the analysis, and revised the paper.
10 2005 4 10 2005 2 10 e29414 3 2005 25 7 2005 Copyright: © 2005 Wjst et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Teasing Out the Effects of Latitude and Birth Date on Allergy
Background
The space and time distribution of risk factors for allergic diseases may provide insights into disease mechanisms. Allergy is believed to vary by month of birth, but multinational studies taking into account latitude have not been conducted.
Methods and Findings
A questionnaire was distributed in 54 centres to a representative sample of 20- to 44-y-old men and women mainly in Europe but also including regions in North Africa, India, North America, Australia, and New Zealand. Data from 200,682 participants were analyzed. The median prevalence of allergic rhinitis was 22%, with a substantial variation across centres. Overall, allergic rhinitis decreased with geographical latitude, but there were many exceptions. No increase in prevalence during certain winters could be observed. Also, no altered risk by birth month was found, except borderline reduced risks in September and October. Effect estimates obtained by a multivariate analysis of total and specific IgE values in 18,085 individuals also excluded major birth month effects and confirmed the independent effect of language grouping.
Conclusion
Neither time point of first exposure to certain allergens nor early infections during winter months seems to be a major factor for adult allergy. Although there might be effects of climate or environmental UV exposure by latitude, influences within language groups seem to be more important, reflecting so far unknown genetic or cultural risk factors.
A large international survey of risk factors associated with allergy refutes previous work suggesting that allergy varies by month of birth
==== Body
Introduction
Allergy prevalence has been on the rise in many countries, while causal risk factors are still unknown [1]. The spatial and temporal distribution of risk factors may offer an insight into the mechanism of disease.
Birth month has been claimed to be associated with allergy. More than half of the studies summarized in our first analysis of birth month and allergy in 1992 [2] showed a positive association of month of birth with various allergy outcomes [3–19]. A few studies missed at that time, as well as most consecutive studies [20–36] do not show a consistent relationship.
Birth month has been used as a proxy for early allergen exposure but may also be associated with upper respiratory infections during certain winter months. At least in Europe, exposure to outdoor allergens is expected to occur in annually fixed flowering intervals, while episodes of respiratory infections are encountered in autumn and winter months with variation between years. Any autumn or winter season of birth effect could give further support the hygiene hypothesis [37] that postulates a reduction of natural infection, which is responsible for an over-reactive immune system, finally leading to allergy.
Geographical latitude so far has been associated with different diseases such as Crohn disease [38] or type I diabetes [39] but only sporadically with allergy [40]. Latitude is usually described as a proxy for UV solar exposure, as radiation reaching the earth's surface varies inversely with latitude. It may also reflect climatic differences responsible for different pollen seasons, as well as different building construction. In addition, many other factors are associated with geographical latitude in Europe, such as genetic influences or cultural differences in raising children.
The aim of this analysis was, therefore, to further delineate latitude and birth date effects on the prevalence of allergy defined by markers such as allergic rhinitis (AR), sensitization to grass or dust, and total IgE levels.
Methods
Sample
The methods for the European Community Respiratory Health Study (ECRHS) I were published earlier [41], with protocols and questionnaires available from the study Web site (http://www.ecrhs.org). Briefly, ECRHS I participating centres were each selected from an area defined by pre-existing administrative boundaries, with a population of at least 150,000 people. An up-to-date sampling frame was used to randomly select at least 1,500 men and 1,500 women aged 20 to 44 y. All individuals were sent a questionnaire enquiring about respiratory symptoms and attacks of asthma in the last 12 mo, current use of asthma medication, and nasal allergies including hayfever (ECRHS I screening). This sample consists of 54 centres with 200,682 participants. For this analysis, the study centre Aarhus and part of Erfurt probands were excluded because of unreliable birth dates, in addition to all individuals with wrong or missing birth dates and all born on the 29th of February. Also, only birth years from 1945 until 1973 were included, as all other birth years did not have enough observations to be reliable. The final dataset included 186,723 individuals (Table 1). The main outcome variable in this dataset was the response to the question “Do you have any nasal allergies including ‘hayfever'?” Given 16,000 exposed persons in a single month compared to 16,000 born in a reference month with an assumed disease prevalence of 22% and a given α of 0.05, an increase of 1% in the exposed group would have been found with a power of 57% in a two-tailed test, while an increase of 2% would have been found with a power of 99%.
Table 1 Lifetime Prevalence of AR and Prevalence of Positive RAST Values by Centre in ECRHS I Screening
A random sample of these individuals was selected to take part in the full study (ECRHS I main, in joint papers denominated stage II), during which they were invited to visit a local testing centre, answer a more detailed questionnaire, provide a blood sample for measurement of specific IgE and total IgE, perform baseline spirometry, and undergo bronchial challenge with methacholine. Informed consent was obtained from 60% of invited participants [42]. All study protocols were approved by the local ethics committees. Blood samples were all handled in a similar manner and analysed in a central laboratory (Pharmacia, Uppsala, Sweden). Briefly, blood samples were centrifuged and serum stored at −20 °C until total IgE and specific IgE analysis (Radio AllergoSorbent Test [RAST]), of which only mixed grass and house dust mite were used in this study. RAST class 1 or greater was considered as a positive antibody test. All total IgE values were log-transformed and tested on a continuous scale. Numbers under analysis are given in Table 1.
Exposure
The following birth-related variables were defined from the questionnaire: sinus(day of year [1,...,365]); day of month (1,…,31); day of week (1 = Sunday,.., 7 = Saturday); month of year (1 = Jan,…,12 = Dec); season (1 = Dec/Jan/Feb, …, 4 = Sep/Oct/Nov); annual season (1/1945, …, 4/1973) and year of birth (1945,…,1973). Reported in this study are only month, year, and quarter of birth year, as none of the other variables showed any additional information.
Geographical latitude was obtained from the route planning software Mapsonic (http://www.viamichelin.com, where GPS degrees were obtained for a random inner city point and were included as a continuous variable as well as dichotomized into quartile groups. Non-European locations were taken from The World Gazetteer (http://www.world-gazetteer.com) or Encarta (http://encarta.msn.com). For Figures 1 and 2, the original latitude values are used, while for all regression models only the absolute latitude values are taken. If primary language was not self evident, the CIA's World Factbook (http://www.cia.gov/cia/publications/factbook/fields/2098.html) was used as a reference. The city of Montreal was classified as English, although English/French bilingual questionnaires were used. Additional data included in this study are obtained from http://rimmer.ngdc.noaa.gov/mgg/coast/getcoast.html, http://www.polleninfo.org, and http://www.dssresearch.com/toolkit/spcalc/power_p2.asp.
Figure 1 World Map and European Map of Study Centres
Figure 2 Prevalence of AR in ECRHS I Screening by Centre
Centres are sorted by increasing geographical latitudes from left to right.
Analysis
In an initial step, Trellis barcharts and boxplots [43] were created for all date and latitude variables in the ECRHS I screening sample and contingency tables analysed using global χ2 tests. In a second step this analysis was repeated in ECRHS I main study for AR, total IgE, RAST grass, and RAST house dust. Heterogeneity across centres was assessed using meta-analysis with latitude included as a random effect [44]. Next, generalised linear equations were fitted with a binomial outcome for categorical and Gaussian outcome for IgE values. Analyses were conducted for single risk factors alone (Table 2), followed by the joint inclusion of all factors (Table 3). Open-source R software 2.0.1 was used for all analyses (http://www.r-project.org).
Table 2 Crude Risk for AR, Total IgE, and Sensitization against Grass or House Dust Mite in ECRHS I Main Study
Table 3 Adjusted Risk for AR, Total IgE, and Sensitization against Grass or House Dust Mite in ECRHS I Main Study
Results
A history of AR was reported in the screening questionnaire with a range from 2.4% in Nancy to 41.1% in Melbourne (Table 1). Participants reported on median a prevalence of 22.2%, while the prevalence in ECRHS I main was slightly higher, reaching 27.4% (Table 1).
There was a substantial variation between centres in the frequency of sensitization against mixed grass allergen, ranging from 10.3% in Tartu up to 37.1% in Basel. Dust mite sensitization was high in Melbourne, Norwich, and Pessac, and low in Albacete, Reykjavik, and Uppsala.
All centres (Figure 1) were then ordered by geographical latitude (Figure 2). Latitudes covered in Europe include +37 ° (Huelva in Spain) up to +64 ° (Reykjavik in Iceland). Overall, AR decreased with geographical latitude, but there were many exceptions (Figures S1–S12).
The analysis of birth dates in ECRHS I screening did not show any significant effect, neither for day of year, day of month, nor day of week. Crude risks by birth month in ECRHS I main were also not increased, while in the adjusted analysis only September (OR 0.75; 0.56–1.01) and October (OR 0.75; 0.56–1.00) reached borderline significant reduced risks compared to birth month January.
When tested for heterogeneity according to centre, only birth month May showed significant centre differences (p = 0.021). There may be even more altered risks at the single centre level (like Huelva in January, Montpellier in February, etc.; see Figures S9 for ECRHS I screening and Figure S20 for ECRHS I main), but the overall number of associations does not seem to exceed the number expected by chance.
AR prevalence was high even in the earliest birth cohorts (Figures S1–S11 for ECRHS I screening and Figures S12–S22 for ECRHS I main). Analysing birth date in increasing 3-mo intervals did not reveal any major spike that could be attributed to a seasonal influence of winter months or the known influenza A pandemic years in 1957 and 1968 [45].
Finally, effects were analysed by generalized linear regression models in ECRHS I main (Table 2) that also controlled for confounding effects (Table 3). Factors with significant influence on AR in the adjusted model are a history of AR in the parents, current smoking, latitude, and language. There was also a significant effect by year of birth, but none by month of birth except for borderline reduced risks by being born in September or October. Latitude was negatively associated with AR (overall OR 0.85 per ten-degree increase, p < 0.0001 if included as a continuous variable).
Language appeared to be an independent factor from latitude and was in some instances an even stronger predictor for AR (Table 3). Language group affects not only AR and specific sensitization to all tested allergens, but also total IgE values. The most prominent effect compared to English language was seen for the Spanish language.
While risks for total and specific IgE were usually concordant, risks dissociated in the Italian group (OR for IgE grass 0.55, but 1.69 for total IgE), raising the question of further modifying factors of total IgE in Italy. AR and grass specific IgE were also discordant in the French centres, raising questions about labelling of AR in France (or the specificity of the tested allergens for AR there).
Discussion
We describe a high prevalence of AR with a substantial variation across centres but do not find any major risk by being born in a particular month or during a particular season.
The main advantage of this study is the use of standardized interviews and identical laboratory methods, which leads to the conclusion that the geographical differences are real and not an artefact of non-comparable methods [46]. There is also less concern about non-response, since non-responder differences in the main outcome variables were relatively small [46]. Another benefit is the power of the study due to the high number of participants.
Previous studies of birth months showed mixed results. Due to the different definitions and outcomes used, studies can roughly be grouped into those with a positive [5–7,9,10,12–19,21,24,26,29,30,32–36], negative [3,4,8,27,47], or even unclear outcome [20,22,23,28,31]. Without applying formal criteria of a meta-analysis, these studies are difficult to sort. Some are case-only studies, others depend on cross-sectional data, and only a few are cohort studies. A detailed knowledge of local circumstances would be necessary to integrate all results into a larger framework. Also, this study agrees that there might be relevant birth month effects in single centres, but it questions any global effect.
Taking into account publication bias, it may be understandable that in published studies the positive results outnumber negative ones. Most of the previous studies showed an association with allergic sensitization (and not so often with AR), which may indicate subclinical effects that may gain importance only when occurring in combination with additional risk factors. One of the main advantages of our study—the standardized allergen test protocol—might be a disadvantage where the effects of local allergens might have been missed. An exposure matrix constructed by flowering season in all participating European centres did not result in a different risk estimate (unpublished data; see Materials and Methods).
A further difference in comparison with many previous studies is the higher age of our study participants. It may be possible that more marked symptoms exist in children, that are being lost in adulthood. We are sharing, however, a methodological problem with most previous studies, as we are assuming that current residence is identical with residence of birth, which might not always be true. For those participating in our main study, migration was limited [48] and is probably not leading to a distortion of study results.
Another methodological restriction may be the use of self-reported “hayfever.” This term might be used in a different way across Europe, and there might be secular changes in the labelling. The more or less negative finding for any particular birth month (together with the results for specific IgE against grass) in our study, however, is rather consistent across centres. It is possible that reporting of AR during “symptom” months might be increased, but there is no indication that any differential reporting is also associated with birth date. It is therefore unlikely that specific allergen exposure outdoors directly after birth has a major impact on the development of AR.
So far, no data confirm that allergen exposure has been increased in parallel with AR during the recent decades. This would be a rather likely explanation, as global climatic changes with warming may result in higher grass pollen exposure [49], and construction of better isolated houses may support dust mite growth. A study on peak allergen exposure, however, showed only an insignificant increased sensitization [50], and even allergen avoidance trials are far from being conclusive [51].
Any effect of winter month epidemics, in particular of the influenza A episodes in 1957 and 1968 [45], could not be shown. An increasing trend of asthma incidence by birth year has already been described in this study [52] and is now also found for AR. It may be emphasized that the prevalence in the cohorts born directly after World War II is already high and stable over all birth years. Unfortunately, time and cohort effects cannot be discriminated, but an effect by infection epidemics directly before or after birth is unlikely.
Data pertaining to the geographical distribution of allergic diseases are rare. The International Study of Asthma and Allergies in Childhood (ISAAC) compared the worldwide distribution of AR in children, which varied across centres from 0.8% to 14.9% in the 6- to 7-y-olds and from 1.4% to 39.7% in the 13- to 14-y-olds [53]. Although the prevalence in 13- to 14-y-old children was slightly lower, a direct comparison of 13 centres included in both ECRHS and ISAAC showed good agreement. A worldwide meta-analysis of ISAAC centres showed a negative association of latitude and symptoms of AR with a −0.05% decrease per degree (−0.11; 0.00) in 6- to 7-y-olds and with −0.09% (−0.18; −0.01) in 13- to 14-y-olds; effects were attributed to climatic differences such as indoor humidity or altitude [40]. A study in Australia also reported a negative association of latitude and asthma [39], while effects in this study are interpreted by UV solar radiation. This view may be also supported by a recent study of vitamin D supplementation and allergy [54].
AR decreased with geographical latitude, but it is unclear why the Spanish, Portuguese, and Greek centres make such an exception from this rule. Surprisingly, in another study the lowest serum vitamin D metabolite concentrations were seen in southern European countries, which could be explained by attitudes toward sunlight exposure [55].
The lowest prevalence of AR in the ISAAC was found in Eastern Europe and South and Central Asia, and—as in the ECRHS—a high prevalence was reported for English-speaking centres. It is intriguing that inclusion of preferential language into the multivariate model did not resolve the latitude effect, and even increased it.
Language may be a marker for genetic traits [56], and there is indeed a genetic heterogeneity in this study by language where a history of AR in the family does not countervail for the latitude effect. A risk factor operating within language borders therefore seems to be even more relevant than geographical latitude alone.
Supporting Information
Figure S1 ECRHS I Screening AR Prevalence by Birth Quarter and Centre
(298 KB PDF)
Click here for additional data file.
Figure S2 ECRHS I Screening AR Prevalence by Birth Quarter and Latitude/Longitude
(105 KB PDF)
Click here for additional data file.
Figure S3 ECRHS I Screening AR Prevalence by Birth Quarter and Language
(105 KB PDF)
Click here for additional data file.
Figure S4 ECRHS I Screening AR Prevalence by Birth Quarter and Sex
(32 KB PDF)
Click here for additional data file.
Figure S5 ECRHS I Screening AR Prevalence by Month of Birth and Centre
(87 KB PDF)
Click here for additional data file.
Figure S6 ECRHS I Screening AR Prevalence by Month of Birth and Latitude/Longitude
(59 KB PDF)
Click here for additional data file.
Figure S7 ECRHS I Screening AR Prevalence by Month of Birth and Language
(54 KB PDF)
Click here for additional data file.
Figure S8 ECRHS I Screening AR Prevalence by Month of Birth and Sex
(48 KB PDF)
Click here for additional data file.
Figure S9 ECRHS I Screening AR Prevalence by Month and Year of Birth
(66 KB PDF)
Click here for additional data file.
Figure S10 ECRHS I Screening AR Prevalence by Quarter and Month of Birth
(66 KB PDF)
Click here for additional data file.
Figure S11 ECRHS I Screening AR Prevalence by Quarter and Month of Birth
(51 KB PDF)
Click here for additional data file.
Figure S12 ECRHS I Main Median Log Total IgE by Birth Quarter and Centre
(180 KB PDF)
Click here for additional data file.
Figure S13 ECRHS I Main Median Log Total IgE by Birth Quarter and Latitude/Longitude
(120 KB PDF)
Click here for additional data file.
Figure S14 ECRHS I Main Median Log Total IgE by Birth Quarter and Language
(85 KB PDF)
Click here for additional data file.
Figure S15 ECRHS I Main Median Log Total IgE by Birth Quarter and Sex
(58 KB PDF)
Click here for additional data file.
Figure S16 ECRHS I Main Median Log Total IgE by Month of Birth and Centre
(70 KB PDF)
Click here for additional data file.
Figure S17 ECRHS I Main Median Log Total IgE by Month of Birth and Latitude/Longitude
(61 KB PDF)
Click here for additional data file.
Figure S18 ECRHS I Main Median Log Total IgE by Month of Birth and Language
(53 KB PDF)
Click here for additional data file.
Figure S19 ECRHS I Main Median Log Total IgE by Month of Birth and Sex
(47 KB PDF)
Click here for additional data file.
Figure S20 ECRHS I Main Median Log Total IgE by Month and Year of Birth
(67 KB PDF)
Click here for additional data file.
Figure S21 ECRHS I Main Median Log Total IgE by Month and Quarter of Birth
(66 KB PDF)
Click here for additional data file.
Figure S22 ECRHS I Main Median Log Total IgE by Absolute Latitude and Language
(58 KB PDF)
Click here for additional data file.
Patient Summary
Background
Allergy is becoming more common in many parts of the world, but there is no satisfactory explanation for this, or for why the number of people with allergies varies so much between countries. If scientists knew more about where and when the risks of allergy are highest, it might help them understand more about the condition and its causes. Some research has suggested that the month in which one is born can affect one's risk of getting an allergy, although other studies have not found this. The month of birth would determine how old a baby was when it first encountered an allergy-causing substance (like pollen) or an infection (like a cold), and this might turn out to be important.
What Did The Researchers Do?
They looked at one type of allergy: allergic rhinitis, the inflammation of the membranes inside the nose that is usually called hayfever. They gave a questionnaire to patients in 54 regions, mainly in Europe. More than 200,000 people responded, and nearly a quarter of them had allergic rhinitis. Overall, the condition was found to be more common further away from the Equator, although there were some exceptions. However, the month of birth did not seem to make a difference in the likelihood of getting allergy. The researchers did find a variation in allergy rates according to which languages people spoke. They suggest that this means there are genetic and cultural factors involved in allergy risk.
What Does This Mean?
This study was larger and used more reliable methods than some earlier research, so we can be more confident of the conclusions. The mystery of why allergy is becoming more common has not been solved, but the researchers' conclusion that genetic and cultural factors are more important than geographical factors is an advance in our knowledge.
More Information Online
General information for suffers of allergy, including hayfever, may be found on these Web sites.
The American Academy of Family Physicians:
http://familydoctor.org/083.xml
The American Academy for Asthma, Allergy and Immunology:
http://www.aaaai.org
Allergy UK, formerly the British Allergy Foundation:
http://www.allergyuk.org/allergy_whatis.html
For the big picture on asthma worldwide, try the World Allergy Organization:
http://www.worldallergy.org
We wish to thank Michelle Emfinger for proof-reading of the manuscript and Deepayan Sarkar for help with the lattice R package. The ECRHS study is a joint project by many participants and funded by many sources.
Project Leader: Peter Burney; Statistician: Sue Chinn; Principal Investigator: Deborah Jarvis; Project Coordinator: Jill Knox; Principal Investigator: Christina Luczynska; Assistant Statistician: J Potts; Data Manager: S Arinze.
Steering Committee: Josep M Antó, Institut Municipal d'Investigació Mèdica (IMIM-IMAS), Universitat Pompeu Fabra (UPF); Peter Burney, King's College London; Isa Cerveri, University of Pavia; Susan Chinn, King's College London; Roberto de Marco, University of Verona; Thorarinn Gislason, Iceland University Hospital; Joachim Heinrich, GSF—Institute of Epidemiology; Christer Janson, Uppsala University; Deborah Jarvis, King's College London; Jill Knox, King's College London; Nino Künzli, formerly University of Basel, now University of Southern California Los Angeles; Bénédicte Leynaert, Institut National de la Santé et de la Recherche Médicale (INSERM); Christina Luczynska, King's College London; Françoise Neukirch, Institut National de la Santé et de la Recherche Médicale (INSERM); J Schouten, University of Groningen; Jordi Sunyer, Institut Municipal d'Investigació Mèdica (IMIM-IMAS), Universitat Pompeu Fabra (UPF); Cecilie Svanes, University of Bergen; Vermeire, University of Antwerp; Matthias Wjst, GSF Institute of Epidemiology.
Principal Investigators and Senior Scientific Team
Australia: Melbourne (M Abramson, EH Walters, J Raven, S Dharmage). Belgium: South Antwerp and Antwerp City (P Vermeire, J Weyler, M Van Sprundel, V Nelen). Canada: Halifax (D Bowie), Hamilton (MR Sears, HC Siersted); Montreal (MR Becklake, P Ernst); Prince Edward Island (L Sweet, L Van Til); Vancouver (M Chan-Yeung, H Dimich-Ward); Winnipeg (J Manfreda, NR Anthonisen). Estonia: Tartu (R Jogi, A Soon). France: Paris (F Neukirch, B Leynaert, R Liard, M Zureik); Grenoble (I Pin, J Ferran-Quentin). Germany: Erfurt (J Heinrich, M Wjst, C Frye, I Meyer). Iceland: Reykjavik (T Gislason, E Bjornsson, D Gislason, T Blondal, KB Jorundsdottir). Italy: Turin (M Bugiani, P Piccioni, E Caria, A Carosso, E Migliore, G Castiglioni); Verona (R de Marco, G Verlato, E Zanolin, S Accordini, A Poli, V Lo Cascio, M Ferrari); Pavia (A Marinoni, S Villani, M Ponzio, F Frigerio, M Comelli, M Grassi, I Cerveri, A Corsico). Netherlands: Groningen and Geleen (J Schouten, M Kerkhof). Norway: Bergen (A Gulsvik, E Omenaas, C Svanes, B Laerum). Spain: Barcelona (JM Antó, J Sunyer, M Kogevinas, JP Zock, X Basagana, A Jaen, F Burgos); Huelva (J Maldonado, A Pereira, JL Sanchez); Albacete (J Martinez-Moratalla Rovira, E Almar); Galdakao (N Muniozguren, I Urritia), Oviedo (F Payo). Sweden: Uppsala (C Janson, G Boman, D Norback, M Gunnbjornsdottir); Goteborg (K Toren, L Lillienberg, AC Olin, B Balder, A Pfeifer-Nilsson, R Sundberg); Umea (E Norrman, M Soderberg, K Franklin, B Lundback, B Forsberg, L Nystrom). Switzerland: Basel (N Künzli, B Dibbert, M Hazenkamp, M Brutsche, U Ackermann-Liebrich). United Kingdom: Norwich (D Jarvis, B Harrison); Ipswich (D Jarvis, R Hall, D Seaton).
Funders
Financial support for ECRHS I centres: Allen and Hanbury, Belgian Science Policy Office, National Fund for Scientific Research; Ministère de la Santé, Glaxo France, Institut Pneumologique d'Aquitaine, Contrat de Plan Etat-Région Languedoc-Rousillon, CNMATS, CNMRT (90MR/10, 91AF/6), Ministre delegué de la santé, RNSP, France; Health Canada, Province of Prince Edward Island, Glaxo Canada; GSF, and the Bundesministerium für Forschung und Technologie, Bonn, Germany; Ministero dell'Università e della Ricerca Scientifica e Tecnologica, CNR, Regione Veneto (RSF n. 381/05.93), Italy; Norwegian Research Council (101422/310); Dutch Ministry of Wellbeing, Public Health and Culture, Netherlands; Ministero Sanidad y Consumo FIS (91/0016060/00E-05E and 93/0393), and grants from Hospital General de Albacete, Hospital General Juan Ramón Jiménenz, Consejeria de Sanidad Principado de Asturias, Spain; The Swedish Medical Research Council, the Swedish Heart Lung Foundation, and the Swedish Association against Asthma and Allergy; Swiss National Science Foundation (4026–28099); National Asthma Campaign, British Lung Foundation, Department of Health, South Thames Regional Health Authority, UK; US Department of Health, Education and Welfare Public Health Service (2 S07 RR05521–28) and Victorian Asthma Foundation.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Citation: Wjst M, Dharmage S, André E, Norback D, Raherison C, et al. (2005) Latitude, birth date, and allergy. PLoS Med 2(10): e294
Abbreviations
ARallergic rhinitis
ECRHSEuropean Community Respiratory Health Study
ISAACInternational Study of Asthma and Allergies in Childhood
RASTRadio AllergoSorbent Test
==== Refs
References
Meltzer EO Hamilos DL Hadley JA Lanza DC Marple BF Rhinosinusitis: Establishing definitions for clinical research and patient care J Allergy Clin Immunol 2004 114 155 212 15577865
Wjst M Dold S Reitmeir P Stiepel E von Mutius E Month of birth and allergic disease at the age of 10 Clin Exp Allergy 1992 22 1026 1031 1468028
Reed CE The failure of antepartum or neonatal exposure to grass pollen to influence the later development of grass sensitivity J Allergy 1958 29 300 301 13563063
Kleiner H Arkins JA Lauwasser M Correlation between date of birth and pollen sensitivity Ann Allergy 1975 34 310 314 1124867
Soothill JF Stokes CR Turner MW Norman AP Taylor B Predisposing factors and the development of reaginic allergy in infancy Clin Allergy 1976 6 305 319 963861
Bjoerksten F Suoniemi I Dependence of immediate hypersensitivity on the month of birth Clin Allergy 1976 6 165 171 1277440
Pearson DJ Freed DL Taylor G Respiratory allergy and month of birth Clin Allergy 1977 7 29 33 872354
Smith JM Springett VH Atopic disease and month of birth Clin Allergy 1979 9 153 157 445784
Kemp AS Relationship between the time of birth and the development of immediate hypersensitivity to grass-pollen antigens Med J Aust 1979 1 263 264 449781
Bjorksten F Suoniemi I Koski V Neonatal birch-pollen contact and subsequent allergy to birch pollen Clin Allergy 1980 10 585 591 7438417
Anderson HR Bailey PA Bland JM The effect of birth month on asthma, eczema, hayfever, respiratory symptoms, lung function, and hospital admissions for asthma Int J Epidemiol 1981 10 45 51 7239762
Pedersen PA Weeke ER Month of birth in asthma and allergic rhinitis Scand J Prim Health Care 1983 1 97 101 6545046
Korsgaard J Dahl R Sensitivity to house dust mite and grass pollen in adults. Influence of the month of birth Clin Allergy 1983 13 529 535 6640888
David TJ Beards SC Asthma and the month of birth Clin Allergy 1985 15 391 395 4028390
Carosso A Ruffino C Bugiani M The effect of birth season on pollenosis Ann Allergy 1986 56 300 303 3963521
Beck HI Hagdrup HK Atopic dermatitis, house dust mite allergy and month of birth Acta Derm Venereol 1987 67 448 451 2449000
Businco L Cantani A Farinella F Businco E Month of birth and grass pollen or mite sensitization in children with respiratory allergy: A significant relationship Clin Allergy 1988 18 269 274 3396196
Quoix E Bessot JC Kopferschmitt-Kubler MC Fraisse P Pauli G Positive skin tests to aero-allergens and month of birth Allergy 1988 43 127 131 3364625
Troise C Voltolini S Delbono G Ebbli A Negrini AC Allergy to Parietaria pollen and month of birth Allergol Immunopathol (Madr) 1989 17 201 204 2816662
Tan KY Voorhorst R Hay fever and month of birth Allergol Immunopathol (Madr) 1978 6 9 18 685791
Peinado Perez J Fernandez Hidalgo D Pantoja Bajo A Perez Alvarez T Olmos Berrocoso A [Birth month as a predisposing factor in allergic disease] An Esp Pediatr 1983 19 29 32 6638709
Tseng RY Lam CW Davies DP Cord blood IgE and month of birth Arch Dis Child 1988 63 573 574
Strachan DP Seagroatt V Cook DG Chest illness in infancy and chronic respiratory disease in later life: An analysis by month of birth Int J Epidemiol 1994 23 1060 1068 7860157
Karachaliou FH Panagiotopoulou K Manousakis M Sinaniotis K Papageorgiou F Month of birth, atopic disease, and sensitization to common aeroallergens in Greece Pediatr Allergy Immunol 1995 6 216 219 8822396
Schaefer F Wingen AM Hennicke M Rigden S Mehls O Growth charts for prepubertal children with chronic renal failure due to congenital renal disorders. European Study Group for Nutritional Treatment of Chronic Renal Failure in Childhood Pediatr Nephrol 1996 10 288 293 8792392
Eriksson NE Holmen A Skin prick tests with standardized extracts of inhalant allergens in 7099 adult patients with asthma or rhinitis: Cross-sensitizations and relationships to age, sex, month of birth and year of testing J Investig Allergol Clin Immunol 1996 6 36 46
Saryan JA Bouras A Month of birth and sensitization to dust mites in New England Allergy Asthma Proc 1996 17 65 69 8934795
Singh BP Malhotra M Sridhara S Gaur SN Relevance of birth month in respiratory allergy Allergy 1997 52 232 233 9105534
Erel F Karaayvaz M Caliskaner Z Ozanguc N The allergen spectrum in Turkey and the relationships between allergens and age, sex, birth month, birthplace, blood groups and family history of atopy J Investig Allergol Clin Immunol 1998 8 226 233
Chew FT Goh DY Teo J Quak SH Lee BW Month of birth and childhood atopic diseases in the tropics Allergy 1998 53 962 968 9821476
de Montis G [Reagin sensitizations and date of birth. Differences in function according to family history] Arch Pediatr 1999 6 259 262 10191890
Kusunoki T Asai K Harazaki M Korematsu S Hosoi S Month of birth and prevalence of atopic dermatitis in schoolchildren: Dry skin in early infancy as a possible etiologic factor J Allergy Clin Immunol 1999 103 1148 1152 10359898
Vovolis V Grigoreas C Galatas I Vourdas D Is month of birth a risk factor for subsequent development of pollen allergy in adults? Allergy Asthma Proc 1999 20 15 22 10076705
Saitoh Y Dake Y Shimazu S Sakoda T Sogo H Month of birth, atopic disease, and atopic sensitization J Investig Allergol Clin Immunol 2001 11 183 187
Guerra S Sherrill DL Cottini M Michetti G Allegra L On the association between date of birth and pollen sensitization: Is age an effect modifier? Allergy Asthma Proc 2002 23 303 310 12476539
Dik N Tate RB Manfreda J Anthonisen NR Risk of physician-diagnosed asthma in the first 6 years of life Chest 2004 126 1147 1153 15486376
Strachan DP Allergy and family size: A riddle worth solving Clin Exp Allergy 1997 27 235 236 9088647
Armitage EL Aldhous MC Anderson N Drummond HE Riemersma RA Incidence of juvenile-onset Crohn's disease in Scotland: Association with northern latitude and affluence Gastroenterology 2004 127 1051 1057 15480983
Staples JA Ponsonby AL Lim LL McMichael AJ Ecologic analysis of some immune-related disorders, including type 1 diabetes, in Australia: Latitude, regional ultraviolet radiation, and disease prevalence Environ Health Perspect 2003 111 518 523 12676609
Weiland SK Husing A Strachan DP Rzehak P Pearce N Climate and the prevalence of symptoms of asthma, allergic rhinitis, and atopic eczema in children Occup Environ Med 2004 61 609 615 15208377
Burney PG Luczynska C Chinn S Jarvis D The European Community Respiratory Health Survey Eur Respir J 1994 7 954 960 8050554
Sunyer J Basagana X Burney P Anto JM International assessment of the internal consistency of respiratory symptoms. European Community Respiratory Health Study (ECRHS) Am J Respir Crit Care Med 2000 162 930 935 10988108
Cleveland WS Visualizing data 1993 Summit (New Jersey) Hobart Press 360
DerSimonian R Laird N Meta-analysis in clinical trials Control Clin Trials 1986 7 177 188 3802833
Reid AH Taubenberger JK The origin of the 1918 pandemic influenza virus: A continuing enigma J Gen Virol 2003 84 2285 2292 12917448
European Community Respiratory Health Survey Variations in the prevalence of respiratory symptoms, self-reported asthma attacks, and use of asthma medication in the European Community Respiratory Health Survey (ECRHS) Eur Respir J 1996 9 687 695 8726932
Schafer T Przybilla B Ring J Kunz B Greif A Manifestation of atopy is not related to patient's month of birth Allergy 1993 48 291 294 8328665
Tobias A Soriano JB Chinn S Anto JM Sunyer J Symptoms of asthma, bronchial responsiveness and atopy in immigrants and emigrants in Europe. European Community Respiratory Health Survey Eur Respir J 2001 18 459 465 11589342
Williams R Climate change blamed for rise in hay fever Nature 2005 434 1059
Kihlstrom A Lilja G Pershagen G Hedlin G Maternal pollen allergy may be more important than birch pollen exposure during pregnancy for atopic airway disease in the child Pediatr Allergy Immunol 2004 15 497 505 15610362
Simpson A Simpson B Custovic A Craven M Woodcock A Stringent environmental control in pregnancy and early life: The long-term effects on mite, cat and dog allergen Clin Exp Allergy 2003 33 1183 1189 12956752
Sunyer J Anto JM Tobias A Burney P Generational increase of self-reported first attack of asthma in fifteen industrialized countries. European Community Respiratory Health Study (ECRHS) Eur Respir J 1999 14 885 891 10573237
Strachan D Sibbald B Weiland S Ait-Khaled N Anabwani G Worldwide variations in prevalence of symptoms of allergic rhinoconjunctivitis in children: The International Study of Asthma and Allergies in Childhood (ISAAC) Pediatr Allergy Immunol 1997 8 161 176 9553981
Hyppoenen E Sovio U Wjst M Patel S Pekkanen J Infant vitamin d supplementation and allergic conditions in adulthood: Northern Finland birth cohort 1966 Ann N Y Acad Sci 2004 1037 84 95 15699498
van der Wielen RP Lowik MR van den Berg H de Groot LC Haller J Serum vitamin D concentrations among elderly people in Europe Lancet 1995 346 207 210 7616799
Cavalli-Sforza LL Genes, peoples and languages Sci Am 1991 265 104 110 1785035
|
16190778
|
PMC1240049
|
CC BY
|
2021-01-05 11:13:37
|
no
|
PLoS Med. 2005 Oct 4; 2(10):e294
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020294
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1623197510.1371/journal.pmed.0020295Research ArticleInfectious DiseasesMicrobiologyAllergy/ImmunologyVaccinesParasitologyInfectious DiseasesImmunology and allergyVaccination with Recombinant Aspartic Hemoglobinase Reduces Parasite Load and Blood Loss after Hookworm Infection in Dogs Vaccination with Hookworm HemoglobinaseLoukas Alex
1
*Bethony Jeffrey M
2
Mendez Susana
2
Fujiwara Ricardo T
2
Goud Gaddam Narsa
2
Ranjit Najju
1
Zhan Bin
2
Jones Karen
2
Bottazzi Maria Elena
2
Hotez Peter J
2
*1Division of Infectious Diseases and Immunology, Queensland Institute of Medical Research, Brisbane, Queensland, Australia,2Department of Microbiology and Tropical Medicine, The George Washington University Medical Center, Washington, District of Columbia, United States of AmericaYazdanbakhsh Maria Academic EditorLeiden University Medical Centerthe Netherlands*To whom correspondence should be addressed. E-mail: [email protected] (AL); E-mail: [email protected] (PJH)
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: AL, JMB, MEB, and PJH designed the study. AL, RTF, GNG, NR, BZ, performed experiments. AL, PJH, JMB, SM, and KJ analyzed the data. AL, PJH, JMB, and SM contributed to writing the paper.
10 2005 4 10 2005 2 10 e2958 4 2005 13 7 2005 Copyright: © 2005 Loukas et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
The End of the Line for Hookworm? An Update on Vaccine Development
Getting Closer to a Vaccine for Hookworm
Background
Hookworms infect 730 million people in developing countries where they are a leading cause of intestinal blood loss and iron-deficiency anemia. At the site of attachment to the host, adult hookworms ingest blood and lyse the erythrocytes to release hemoglobin. The parasites subsequently digest hemoglobin in their intestines using a cascade of proteolysis that begins with the Ancylostoma caninum aspartic protease 1, APR-1.
Methods and Findings
We show that vaccination of dogs with recombinant Ac-APR-1 induced antibody and cellular responses and resulted in significantly reduced hookworm burdens (p = 0.056) and fecal egg counts (p = 0.018) in vaccinated dogs compared to control dogs after challenge with infective larvae of A. caninum. Most importantly, vaccinated dogs were protected against blood loss (p = 0.049) and most did not develop anemia, the major pathologic sequela of hookworm disease. IgG from vaccinated animals decreased the catalytic activity of the recombinant enzyme in vitro and the antibody bound in situ to the intestines of worms recovered from vaccinated dogs, implying that the vaccine interferes with the parasite's ability to digest blood.
Conclusion
To the best of our knowledge, this is the first report of a recombinant vaccine from a hematophagous parasite that significantly reduces both parasite load and blood loss, and it supports the development of APR-1 as a human hookworm vaccine.
Vaccination of dogs with a recombinant protease produced by hookworms can reduce blood loss when these dogs are infected with the hookworm Ancylostoma caninum.
==== Body
Introduction
Hookworms infect more than 700 million people in tropical and subtropical regions of the world. The major species infecting humans are Necator americanus and Ancylostoma duodenale. The parasites feed on blood, causing iron-deficiency anemia, and as such, are a major cause of disease burden in developing countries [1]. Unlike other human helminthiases, worm burdens do not generally decrease with age; in fact, recent findings revealed that the heaviest worm burdens are found among the elderly [2,3]. Whereas anthelminthic chemotherapy with benzimidazole drugs is effective in eliminating existing adult parasites, re-infection occurs rapidly after treatment [4], making a vaccine against hookworm disease a desirable goal.
Canines can be successfully vaccinated against infection with the dog hookworm, Ancylostoma caninum, by immunization with third-stage infective larvae (L3) that have been attenuated with ionizing radiation [5–7]. Subsequently, varying levels of vaccine efficacy have been reported for the major antigens secreted by hookworm L3 using hamsters [8,9] and dogs [10]. Despite obtaining encouraging levels of protection with larval antigens, only partial reductions in parasite load (fecal egg counts and adult worm burdens) were reported. Moreover, protective antigens from the larval stage are only expressed by L3, and not adult worms, rendering antibodies against these L3 secretions useless against parasites that have successfully reached adulthood in the gut and begun to feed on blood. We therefore suggest that an ideal hookworm vaccine would require a cocktail of two recombinant proteins, one targeting the infective larva and the second targeting the blood-feeding adult stage of the parasite [11].
Of the different families of proteins expressed by blood-feeding parasitic helminths, proteolytic enzymes have shown promise as intervention targets for vaccine development [12,13]. Proteases are pivotal for a parasitic existence, mediating fundamental physiologic processes such as molting, tissue invasion, feeding, embryogenesis, and evasion of host immune responses [12,14]. Parasite extracts enriched for proteases protect sheep against the blood-feeding nematodes Haemonchus contortus [15–18] and Ostertagia ostertagi [19]; however, significant protective efficacy has not been shown with a purified recombinant protease from nematodes of livestock.
Hookworms feed by burying their anterior ends in the intestinal mucosa of the host, rupturing capillaries and ingesting the liberated blood. Erythrocytes are lysed by pore formation [20], releasing hemoglobin (Hb) into the lumen of the parasite's intestine, where it is degraded by a semi-ordered pathway of catalysis that involves aspartic, cysteine, and metalloproteases [21]. Vaccination of dogs with a catalytically active recombinant cysteine hemoglobinase, Ac-CP-2, induced antibodies that neutralized proteolytic activity and provided partial protection to vaccinees by reducing egg output (a measure of intestinal worm burden) and worm size, but significant reductions of adult worm burdens and/or blood loss were not observed [22]. Anemia is the primary pathology associated with hookworm infection, and an ultimate human hookworm vaccine would limit the amount of blood loss caused by feeding worms and maintain normal levels of Hb. This is particularly important in young children as well as women of child-bearing age, in whom menstrual, and particularly fetal, Hb demands are considerable, rendering these populations most vulnerable to the parasite [1].
Here we describe vaccination of dogs with the aspartic hemoglobinase of A. caninum, Ac-APR-1 [21,23] and show that vaccination resulted in the production of neutralizing antibodies, significantly reduced egg counts, and significantly reduced adult worm burdens. Most importantly, Hb levels of vaccinated dogs were significantly higher than those of dogs that were vaccinated with adjuvant alone after parasite challenge. These data show that aspartic hemoglobinases, particularly APR-1, are efficacious vaccines against canine hookworm disease, providing strong support for further investigation and development of APR-1 as a recombinant vaccine against human hookworm disease.
Methods
Expression of Recombinant Ac-APR-1 in Pichia pastoris
The entire open reading frame of Ac-APR-1 encoding the zymogen (spanning Ser-17 to the C-terminal Phe-446) but excluding the predicted signal peptide was cloned into the expression vector pPIC-Zα (Invitrogen, Carlsbad, California, United States) using the XbaI and EcoRI sites. Yeast, P. pastoris X 33, was transformed with the vector encoding the Ac-APR-1 zymogen as recommended by the manufacturer (Invitrogen) with modifications. Protein disulfide isomerase (PDI) gene in the vector pPIC3.5 (a gift from Mehmet Inan, University of Nebraska, Lincoln, Nebraska, United States) was cut with SacI and transformed into P. pastoris X 33 cells which were already transformed with Ac-apr-1 following the manufacturer's instructions. Eight transformed colonies were picked from YPD plates containing Geneticin (0.5–1.0 mg·ml−1) and Zeocin (1.0 mg·ml−1) and tested for Ac-APR-1 expression following the manufacturer's instructions. The highest expressing colony was selected and transferred to suspension culture in flasks containing BMG medium (buffered minimal glycerol: 1.34% yeast nitrogen base, 0.00004% d-biotin, 1% w/v glycerol, and 100 mM potassium phosphate, [pH 6.0]). Suspension cultures were then transferred to a Bioflo 3000 fermentor (New Brunswick Scientific, Edison, New Jersey, United States) utilizing a 5-l vessel as described [8]. The recombinant protein was secreted into culture medium and affinity purified on nickel-agarose as described elsewhere [8]. Progress of purification was monitored using SDS-PAGE gels stained with Coomassie Brilliant Blue and immunoblots using monoclonal antibodies to the vector-derived myc epitope. Recombinant Ac-APR-1 was treated with PNGase F and O-glycosidase, according to the manufacturer's instructions (Enzymatic CarboRelease kit; QA-Bio, San Mateo, California, United States), under denaturing conditions to remove any N-linked and O-linked oligosaccharides. Deglycosylation was performed only to confirm the presence of N-linked sugars on the recombinant molecule. All remaining studies were conducted with the glycoprotein.
Activation and Hemoglobinolytic Activity of Recombinant APR-1
The unactivated zymogen was used for vaccination. A small amount of the purified protein, however, was buffer exchanged into 100 mM sodium formate (pH 3.6)/0.15 M NaCl using a PD10 desalting column (Amersham Biosciences, Little Chalfont, United Kingdom) to facilitate proteolytic activation and removal of the pro-region. One microgram of purified, activated protease was then added to 10 μg of dog Hb in the same buffer and incubated at 37 °C for 2 h. Cleavage of Hb was assessed visually by staining SDS-PAGE gels with Coomassie Brilliant Blue.
Animal Husbandry
Purpose-bred, parasite naive, male beagles aged 8 ± 1 wk were purchased from Marshall Farms (North Rose, New York, United States), identified by ear tattoo, and maintained in the George Washington University Animal Research Facility as previously described [24]. The experiments were conducted according to a protocol approved by the University Animal Care and Use Committee (IACUC 48–12,0 [12,1]E). Before the first vaccination and after each subsequent one, a blood sample was obtained from each dog.
Vaccine Study Design and Antigen-Adjuvant Formulation
The vaccine trial was designed to test Ac-APR-1 zymogen formulated with the adjuvant AS03 [25], obtained from GlaxoSmithKline (a kind gift from Drs. Joe Cohen and Sylvie Cayphas; GSK Biologicals, Rixensart, Belgium). To make six doses of Ac-APR-1 formulated with AS03, 600 μg of recombinant protein (1.5 ml of Ac-APR-1 at a concentration of 0.4 mg·ml−1) was mixed with 1.2 ml of 20 mM Tris-HCl, 0.5 M NaCl (pH 7.9), and 1.5 ml of AS03; the contents of the tube were vortex mixed for 30 sec then shaken at low speed for 10 min. Dogs were immunized with 100 μg of formulated antigen in a final volume of 0.5 ml. AS03-only control was prepared as described above, with PBS included instead of Ac-APR-1.
Canine Immunizations and Antibody Measurements
Five beagles were immunized three times with AS03-formulated Ac-APR-1 by intramuscular injection. The vaccine was administered on days 0, 21, and 42, beginning when the dogs were 62 ± 4 d of age. As negative controls, five beagles were also injected intramuscularly with an equivalent amount of AS03 using the identical schedule. Blood was drawn at least once every 21 d and serum was separated from cells by centrifugation. Enzyme-linked immunosorbent assays (ELISA) were performed as previously described [24]. Recombinant Ac-APR-1 was coated onto microtiter plates at a concentration of 5.0 μg·ml−1. Dog sera were titrated between 1:100 and 1:2 × 106 to determine endpoint titers (the highest dilution of test group [APR-1] sera that gave a mean O.D. of ≥3× the mean optical density (OD) of sera from the control group). Anti-canine IgG1, IgG2, and IgE antibodies conjugated to horseradish peroxidase (Bethyl Laboratories, Montgomery, Texas, United States) were used at a dilution of 1:1,000. Blood was collected from dogs before immunizations and 7 d after the third vaccination but before L3 challenge.
Stimulation of and Cytokine Measurements from Cultured Whole Blood
Lymphoproliferation assays were performed using a whole blood microassay as previously described [26]. Briefly, 25 μl of heparinized blood was diluted in 200 μl of RPMI 1640 medium (Gibco, Invitrogen) supplemented with 3% antibiotic/antimycotic solution (Gibco). All tests were performed in triplicate in 96-well flat-bottomed culture plates using recombinant APR-1 at a concentration of 25 μg·ml−1 and concanavalin A (ConA; Sigma-Aldrich, St. Louis, Missouri, United States) at 80 μg·ml−1. Incubation was carried out in a humidified 5% CO2 atmosphere at 37 °C for 2 d (ConA-stimulated cultures) and 5 d (APR-1). Cells were pulsed for 6 h with 1.0 μCi of [3H] thymidine (PerkinElmer Life And Analytical Sciences, Boston, Massachusetts, United States) and harvested onto glass fiber filters. Radioactive incorporation was determined by liquid scintillation spectrometry. Proliferation responses were expressed as stimulation indices, SI (where SI = mean proliferation of stimulated cultures/mean proliferation of unstimulated cultures). For cytokine analyses, whole blood (collected as described above) was diluted 1:8 in RPMI supplemented with 3% antibiotic/antimycotic solution in a 48-well flat-bottomed culture plate with a final volume of 1.0 ml per well. Cells were stimulated by the addition of 25 μg·ml−1 of recombinant APR-1. After 48 h of incubation at 37 °C, 700 μl of supernatant was removed from each well and stored at −20 °C until required for the cytokine assay. IL-4, IL-10, and IFN-γ were measured using a capture ELISA assay for dogs (R & D Systems, Minneapolis, Minnesota, United States) following the manufacturer's instructions. Biotin-labeled detection antibodies were used (100 ng·ml−1), revealed with streptavidin-HRP (Amersham Biosciences), and plates were developed with OPD (O-Phenylenediamine) substrate system (Sigma-Aldrich).
Hb Measurements
To determine Hb concentrations of experimental dogs, 1–2 ml of blood were collected in EDTA and analyzed using a QBC VetAutoread Hematology System and VETTEST Software (IDEXX Laboratories, Westbrook, Maine, United States).
Hookworm Infections and Parasite Recovery
Two weeks after the final immunization, dogs were anaesthetized using a combination of ketamine and xylazine (20 mg·kg−1 and 10 mg·kg−1 respectively) and infected via the footpad with 500 A. caninum L3 as described elsewhere [22]. Quantitative hookworm egg counts (McMaster technique) were obtained for each dog 3 d per wk from days 12–26 postinfection. Four weeks postinfection, the dogs were killed by intravenous injection of barbiturate, and adult hookworms were recovered and counted from the small and large intestines at necropsy [24]. The sex of each adult worm was determined as described elsewhere [8]. Approximately 1–2 cm lengths of small intestine were removed and stored in formalin for future histopathologic analysis.
Statistical Methods
In most cases, the small size of the samples did not enable us to determine if values were normally distributed, so the following non-parametric tests were used: Mann-Whitney U was used to test whether two independent samples (groups) came from the same population, and the Kruskal Wallis H test was used to determine if several independent samples came from the same population. Normally distributed variables were tested in the following manner: The independent-samples t-test procedure was used to compare the means for two groups, and an analysis of variance was used to test the hypothesis that several means are equal, followed by a Dunnet post hoc multiple comparison t-test to compare the vaccine treatment groups against the control group. Differences were considered statistically significant if the calculated p-value was equal to or less than 0.1 (two-sided). The percentage reduction or increase in adult hookworm burden in the vaccinated group was expressed relative to the control group as described elsewhere [24].
Immunohistochemistry
Adult hookworms were recovered at necropsy from vaccinated dogs and control dogs, washed briefly, then fixed and sectioned as previously described [22]. To observe whether IgG from vaccinated but not control dogs, bound to APR-1 lining the intestinal microvillar surface of worms in situ, sections were probed with Cy3-conjugated rabbit anti-dog IgG (Jackson Immunoresearch, West Grove, Pennsylvania, United States) at a dilution of 1:500 as described elsewhere [27]. Sections were visualized using a Leica IM 100 inverted fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
Effect of Anti–Ac-APR-1 IgG on Proteolytic Activity
Canine IgG was purified from sera of vaccinated dogs using protein A-agarose (Amersham Biosciences) as previously described [23]. Purified IgG (0.2 μg) was incubated with 1.0 μg of recombinant Ac-APR-1 for 45 mins prior to assessing catalytic activity of APR-1 against the fluorogenic substrate o-aminobenzoyl-IEF-nFRL-NH2 as described previously [23]. The aspartic protease inhibitor, pepstatin A, was included at a final concentration of 1.0 μM as a positive control for enzymatic inhibition. Data was recorded from triplicate experiments and presented as relative fluorescence units using a TD700 fluorometer (Turner Designs, Sunnyvale, California, United States).
Results
Secretion of Catalytically Active Ac-APR-1 by P. pastoris
Yeast secreted the APR-1 zymogen into culture medium at an approximate concentration of 1.0 mg·l−1 (Figure 1A). In the absence of co-expression with the PDI chaperone, the amount of APR-1 secreted by P. pastoris was approximately half that obtained here (not shown). Ac-APR-1 has one potential glycosylation site at Asn-29 of the zymogen (after removal of the signal peptide), and treatment with PNGase F decreased the size of the recombinant protein by the expected size (2–3 kDa; not shown). The activated recombinant protease readily digested canine Hb at acidic pH (Figure 1B), confirming that Ac-APR-1 expressed in yeast is catalytically active and digested Hb with similar efficiency to recombinant Ac-APR-1 produced in baculovirus (data not shown).
Figure 1 P. pastoris Secrete Ac-APR-1 Zymogen that Autoactivates at Low pH and Degrades Canine Hb
SDS-PAGE gel stained with Coomassie Brilliant Blue showing purification of recombinant APR-1 zymogen from P. pastoris culture supernatant.
(A) Lane 1, molecular weight markers; lane 2, concentrated culture supernatant; lane 3, flow-through from a nickel-IDA column; lane 4, 5 mM imidazole wash; lane 5, 20 mM imidazole column eluate; lane 6, 60 mM imidazole eluate; and lane 7, 1 M imidazole eluate. Purified recombinant APR-1 zymogen was activated by buffer exchange into 0.1 M sodium formate/0.1 M NaCl (pH 3.6).
(B) Lane 1, molecular weight markers; lane 2, 5.0 μg of canine Hb (pH 3.6); and lane 3, 5.0 μg of canine Hb (pH 3.6) incubated with 0.2 μg of recombinant APR-1.
Recombinant Ac-APR-1 Is Immunogenic in Dogs
AS03 was used as an adjuvant based on its ability to induce a higher IgG1 response and greater reduction in hookworm egg counts when used to vaccinate dogs in a head-to-head comparison of a cysteine hemoglobinase formulated with four different adjuvants [22]. Dogs immunized with recombinant Ac-APR-1 formulated with AS03 produced IgG1 and IgG2 antibody responses as measured by ELISA using the recombinant protein (Figure 2). IgE titers were low (<1:1,500) and were not sustained past challenge. We did not adsorb IgG from serum before measuring IgE in this study; however, in previous trials IgG was removed and we did not see a difference in antigen-specific IgE titers. For vaccinated dogs, maximum IgG2 titers of 1:121,500 were attained by all five dogs after the second vaccination. High titers persisted through challenge and decreased to 1:26,098 by necropsy. IgG1 titers peaked at 1:13,500 after the third vaccination in all four dogs and dropped to 1:3,600 by necropsy. Dogs immunized with adjuvant alone did not generate detectable immune responses greater than 1:500, even after larval challenge.
Figure 2 The Geometric Mean Titers of the IgG1 and IgG2 Antibody Responses of Dogs Vaccinated with Recombinant Ac-APR-1 Formulated with AS03 or AS03 Alone
LC, day on which dogs were challenged with hookworm L3; N, day of necropsy; V1, V2, and V3, days on which animals were vaccinated.
Dogs rapidly acquire resistance to hookworm with maturity. A single dog was therefore removed from the control group (for all analyses) because its weight was greater than the acceptable range at all time points after the first vaccination (mean plus or minus three standard errors).
Vaccination Induces Antigen-Specific Cell Proliferation and Cytokine Production
Vaccination with APR-1 induced a high level of lymphocyte/leukocyte proliferation compared with control dogs when cells were stimulated with APR-1 (p < 0.01, t-test). Cells from both vaccinated and control dogs proliferated equally when stimulated with mitogen (Figure 3A and 3B). No significant proliferation to APR-1 was observed before the immunization process. Immunization with APR-1 elicited antigen-specific production of IFN-γ (p = 0.03, t-test) (Figure 3C). In contrast, we did not detect significant production of IL-4 or IL-10 after stimulation with APR-1 in either vaccinated or control groups (not shown).
Figure 3 Canine Cellular Immune Response to Vaccination with Recombinant Ac-APR-1
Cell proliferation of whole blood cells from vaccinated (APR-1) and control dogs (AS03) when stimulated with concanavalin A (A) or recombinant Ac-APR-1 (B) before (day 0) and after the final immunization (day 51). The p-value comparing the mean differences between the vaccinated group and controls is denoted. Detection of secreted IFN-γ in whole blood cultures taken from vaccinated and control dogs before and after immunization (C). Mean cytokine concentrations are indicated in pg·ml−1 with standard error bars. Statistically significant differences are indicated above the bars by p-values. APR, stimulated with recombinant APR-1; NS, non-stimulated cultures.
Vaccination with Ac-APR-1 Decreases Fecundity of Female Hookworms
Dogs develop age- and exposure-related immunity to A. caninum [5], so we therefore observed egg counts from vaccinated animals up to 26 d postchallenge, after which we often observe a significant decrease in egg counts in some dogs. Because of daily variation in egg counts from infected dogs (A. Loukas. S. Mendez, and P. Hotez, unpublished data), we analyzed the data in two ways. Firstly, the median egg counts for days 21, 23, and 26 postinfection were used to compare worm fecundity between vaccinated and control groups. A 70% decrease in median egg counts was observed in dogs vaccinated with Ac-APR-1 (2,650 eggs per gram of feces g]) compared with dogs that were vaccinated with adjuvant alone (8,725 epg) when median egg counts were calculated for the three time points measured after larval challenge (Figure 4A). We then compared geometric mean values of egg counts between the two groups (Figure 4B), and showed that mean egg counts of the vaccinated animals remained lower than the control animals as worms became fecund by day 21, implying that fecundity of female worms diminished significantly as they began to feed on blood containing anti–APR-1 antibodies. By day 26 postchallenge, there was an 85% reduction in mean egg counts between the two groups. For statistical analyses, we transformed egg counts into log values and ran the test in two ways: (1) comparing the log transformed epgs in the last three egg counts by analysis of variance (Kruskall-Wallis) revealed no significant differences among the groups for the last three egg counts when each time point was considered individually; and (2) comparing pooled data from the last three egg counts using a Mann-Whitney test (APR-1 versus control), revealed a statistically significant difference (p = 0.018).
Figure 4 Vaccination with APR-1 Reduces Fecal Egg Counts of Dogs after Challenge Infection with Hookworms
Statistically significant reduction (p = 0.018) in median fecal egg counts sampled on days 21, 23, and 26 of dogs vaccinated with APR-1 compared to dogs that received adjuvant alone.
(A). Geometric mean values of fecal egg counts from vaccinated and control dogs between challenge infection and necropsy.
(B). Error bars refer to the standard error of the mean.
Vaccination with Ac-APR-1 Significantly Reduces Adult Hookworm Burdens
A statistically significant difference at the p ≤ 0.1 level (p = 0.095; Mann-Whitney U test) was detected for a one-sided test between median adult worm burdens recovered from vaccinated dogs (182) compared with control dogs (270) but not for a two-sided test (p = 0.190) (Figure 5). Percentage reduction of the median worm counts was 33% when data from both sexes of worms were combined, 30% for male worms (p = 0.111 [2-sided] or p = 0.056 [1-sided]) and 40% for female worms (p = 0.1905 [2-sided] or p = 0.0952 [1-sided]), again supporting the enhanced effect of the vaccine on female worms given their increased nutritional requirements for egg production.
Figure 5 Vaccination with APR-1 Reduces Adult Worm Burdens of Dogs after Challenge Infection with Hookworms
Statistically significant reduction at the p < 0.1 level (p = 0.065) in median adult worm (both sexes) burdens of dogs vaccinated with APR-1 compared to dogs that received adjuvant alone (A). Reductions are also shown when only male (B) (p = 0.111) and only female (C) (p = 0.1905) worms were considered; however, statistically significant reductions were not achieved for single sex analyses. Bars represent the median value for each group.
Vaccination with APR-1 Protects against Anemia
Hb levels in four of the five dogs that were vaccinated with APR-1 were significantly elevated when compared with control dogs (adjuvant alone) after challenge infection (Figure 6). The median Hb concentration of vaccinated dogs for the last two time points (0 and 7 d prior to necropsy) was 12.45 g·dl−1 compared with 9.5 g·dl−1 for the control dogs that were immunized with adjuvant alone (p = 0.049; Mann-Whitney U test). A decline in Hb levels was seen in all of the control dogs after challenge infection; the decline was marked in three of the four dogs. Four of the five dogs that were vaccinated with APR-1 did not show a similar decline, and had Hb levels within (or very close to) the normal clinical range of 12–14 g·dl−1. One dog (C5) from the vaccinated group did become anemic (Hb concentration was 9.6 g·dl−1), and this animal had more female worms (120 compared with a mean of 88 female worms for the group) and more male worms (87 compared with a mean of 80 male worms for the group). However, using both Spearman and Pearson tests, we did not detect a significant correlation between worm burdens (for either or both sexes) and Hb status of the vaccinated dogs.
Figure 6 Vaccination of Dogs with APR-1 Reduces Blood Loss and Protects against Anemia
Hb concentrations of vaccinated dogs were significantly (p = 0.049) greater than those of control dogs when blood was drawn after larval challenge (0 and 7 d before necropsy [post]) but not when blood was drawn 5 d before larval challenge (pre).
Anti–APR-1 Antibodies Are Ingested by and Bind to the Intestine of Feeding Hookworms
The site of anatomical expression of Ac-APR-1 within adult hookworms has been previously reported by us to be the microvillar surface of the gut [21,23]. To determine whether vaccination of dogs induced circulating antibodies that bound to the intestinal lumen during infection, parasites were removed from vaccinated dogs, fixed, sectioned, and probed with anti-dog IgG conjugated to Cy3. Worms recovered from dogs immunized with Ac-APR-1 but not from dogs immunized with adjuvant alone reacted with Cy3-conjugated anti-dog IgG (Figure 7), indicating that anti–APR-1 antibodies were ingested with the blood-meal of the worm and subsequently bound specifically to the intestine of the parasite in situ.
Figure 7 Antibodies Bind In Situ to the Intestines of Hookworms that Feed on Vaccinated Dogs
Detection of antibodies that bound to the gut of worms recovered from vaccinated dogs (A and B) but not control dogs (C and D) by immunofluorescence. Binding was detected using Cy3-conjugated rabbit anti-dog IgG, allowing only detection of antibodies that had bound in situ while parasites were feeding on blood from vaccinated or control dogs. ic. intestinal contents; in, hookworm intestine; mv, intestinal microvillar surface; ro, reproductive organs.
IgG from Dogs Vaccinated with Ac-APR-1 Neutralizes Proteolytic Activity In Vitro
Purified IgG from dogs that were immunized with Ac-APR-1 reduced the catalytic activity of the enzyme by 71%, compared with just 6% reduction when an equivalent amount of IgG from dogs immunized with adjuvant alone was assessed (Table 1). The aspartic protease inhibitor, pepstatin A, inhibits catalytic activity of APR-1 [23] and was therefore used as a positive control to obtain 100% inhibition for comparative purposes.
Table 1 Reduction in Cleavage of the Fluorogenic Substrate o-Aminobenzoyl-IEF-nFRL-NH2 When 1.0 μg of Recombinant Ac-APR-1 Was Pre-Incubated with 0.2 μg of IgG Purified from Sera of Dogs Vaccinated with APR-1/AS03 or AS03 Alone (Control)
Discussion
Here we describe protective vaccination of dogs with a recombinant aspartic hemoglobinase, a pivotal enzyme in the initiation of Hb digestion in the gut of canine hookworms [12,21]. We show that APR-1 provides the best efficacy thus far reported for a recombinant vaccine aimed at reducing hookworm egg counts, intestinal worm burdens, and hookworm-induced blood loss.
The vaccine efficacy of recombinant Ac-APR-1 expressed in baculovirus-infected insect cells was described earlier by us [24]; however, this initial vaccine trial was hampered by limited availability of the recombinant protein: Suboptimal doses were used and antibody responses (titers <10,000) were first observed just 1 wk following the third (and final) immunization, and only in some dogs. Despite the weak antibody responses, a statistically significant reduction in mean (18%, p < 0.05) and median (23%) hookworm burdens were observed. In addition there was a shift of adult hookworms from the small intestine to the colon [24]. However, no reduction in the mean fecal egg counts were observed, and hematologic parameters were not assessed. The improved immunogenicity of APR-1 observed in this study might also be attributed to use of the adjuvant AS03 compared with alhydrogel in the previous study. We have shown in a head-to-head comparison of a hookworm cysteine hemoglobinase formulated with different adjuvants (including alhydrogel) that AS03-formulated protein generated higher antibody titers and afforded greater protection to vaccinated dogs [22]. In this study, we show that yeast-derived APR-1 provides the best efficacy thus far reported for a recombinant vaccine aimed at reducing hookworm load and potential transmission. Moreover, we show that vaccination protects against the pathology associated with worm-induced blood loss, or hookworm disease.
Hookworms bury their anterior ends into the intestinal mucosa to feed, secreting anticoagulants to promote blood flow and stop clot formation at the site of attachment (reviewed in [28]). Numerous anticoagulant peptides have been reported from hookworms [29–31], and their combined activities result in “leakage” of blood around the attachment site and into the host intestine [32]. It is not known whether the majority of blood loss during a hookworm infection is due to leakage around the feeding site or from ingested blood that enters the parasite's alimentary canal for nutritional purposes. To address this, attempts have been made to measure blood lost from the anus of A. caninum (i.e., blood that has passed through the parasite's alimentary canal); varying calculations have been proposed ranging from 0.14–0.8 ml blood expelled over 24 h per adult worm (reviewed in [32]). Whatever the true figure is, significant blood loss occurs via this route, supporting the hypothesis that vaccination with APR-1 damages that parasite's intestine and results in decreased blood intake (and blood loss) by feeding worms.
The immunological parameters required for vaccine-induced protection against hookworm infection were, until recently, poorly defined. Protection against A. caninum by vaccination of dogs with radiation-attenuated L3 was reported many years ago [5]; however, it was not until recently that murine [8,33] and canine [34] studies revealed the protective mechanisms of the irradiated larval vaccine at a cellular level. These studies suggested that a T-helper type-2 response is induced by vaccination with irradiated L3; however the authors did not prove that a T-helper type-1 response abrogates protection. In our study reported here, dogs vaccinated with APR-1 generated strong memory responses to the recombinant antigen and did not secrete Th-2 cytokines but instead secreted IFN-γ in response to stimulation with recombinant APR-1. Moreover, the dominant antibody isotype induced by vaccination was IgG2, suggesting that a Th-1-like response was generated. Unlike the clear association between IgG2 and type I cytokines such as IFN-γ in mice and humans, little is known about this association in dogs. Experimental evidence using the canine model suggests that immune responses (Th1 versus Th2) are, however, linked to isotype production. For example, animals infected with and protected against visceral leishmaniasis (Th1 response) or Salmonella (also a Th1 response) mount a higher IgG2 than IgG1 response [35,36]. Our data [34] show that dogs immunized with irradiated hookworm larvae demonstrated a stronger production of IgG1 (also supported by [37]) which accompanied IL-4 production, implying a Th2 cytokine response in dogs is accompanied by the same immunoglobulin isotypes seen in humans and mice. Based on the current data, we cannot conclude that a Th-1 response to APR-1 is required to obtain protection; however, it does not inhibit the development of a protective memory response. It should also be considered that successful immunity to the different developmental stages of hookworms might require very different immune response phenotypes, not unlike those seen in schistosomiasis [38]. Further studies will explore the effects of vaccination with APR-1 formulated with different adjuvants and co-factors (e.g., cytokines) that will promote a Th2 response.
Hematophagous helminths require blood as a source of nutrients to mature and reproduce. Female schistosomes ingest 13 times as many erythrocytes and ingest them about nine times faster than male worms [39]. Moreover, mRNAs encoding Hb-degrading proteases of schistosomes are overexpressed in female worms [40]. Although similar studies have yet to be performed for hookworms, female hookworms are bigger than males and lay up to 10,000 eggs per day, implying that they have a greater metabolism and therefore greater demand for erythrocytes.
Ac-APR-1 degrades Hb in the gut lumen of the worm, and it is therefore not surprising that interruption of the function of APR-1 via the action of neutralizing antibodies has a deleterious effect on the establishment of worms, particularly females and their subsequent egg production. We observed a similar (although not as pronounced) phenomenon when dogs were vaccinated with the cysteine hemoglobinase, Ac-CP-2, followed by challenge infection with A. caninum L3 [22]. Vaccination with CP-2, however, did not result in reduced adult worm burdens or reduced blood loss, essential attributes of an efficacious hookworm vaccine.
Vaccination of livestock and laboratory animals with aspartic proteases of other nematodes, as well as trematode helminths, has resulted in antifecundity/antiembryonation effects. Immunization of sheep with the intestinal brush border complex, H-gal-GP, confers high levels of protection (both antiparasite and antifecundity) against H. contortus and at least three different protease activities, including aspartic proteases, have been detected in this extract [16,41]. Immunization of sheep with aspartic protease-enriched fractions of H. contortus membranes resulted in 36% reduction in adult worms and 48% reduction in fecal egg output [17]. Vaccination of sheep with denatured H. contortus proteases or recombinant proteases expressed in bacteria, however, did not confer protection, suggesting that conformational epitopes are important in protection [17]. Vaccination of mice with recombinant aspartic protease of the human blood fluke, Schistosoma mansoni, resulted in 21%–38% reduction in adult parasites after challenge with infective cercariae; however a reduction in eggs deposited in the liver (the cause of most pathology in schistosomiasis) was not detected [42]. Protective efficacy of aspartic proteases has been observed against fungal pathogens as well. Vaccination of mice with secreted aspartic proteases of Candida albicans, known virulence factors in candidiasis, protected animals against a lethal challenge infection and inhibited colonization of fungi in the kidneys [43]. Moreover, passive transfer of serum from vaccinated animals conferred protection, pointing towards an antibody-mediated protective mechanism.
Almost all of the pathology and morbidity of human hookworm infection results from intestinal blood loss caused by large numbers of adult hookworms. Depending on host iron and protein stores, a range of hookworm intensities, equivalent to burdens of 40 to 160 worms, is associated with Hb levels below 11 g·dl−1, the World Health Organization threshold for anemia. In Tanzania, Nepal, and Vietnam where host iron stores are generally depleted, there is a direct correlation between the number of adult hookworms in the intestine and host blood loss [1,44]. Therefore the optimal hookworm vaccine will be one that either prevents L3 from developing into adult blood-feeding hookworms, or one that blocks the establishment, survival, and fecundity of the adult parasites in the intestine [3,45]. Achieving both goals will likely require a vaccine cocktail comprised of an L3 antigen, such as ASP-2 now under clinical development [46,47], and an adult gut protease, such as APR-1.
An effective hookworm vaccine need not attain 100% efficacy. Unlike many unicellular organisms that reproduce asexually within the host, nematodes need to sexually reproduce. Therefore, small numbers of adult worms will generate fewer eggs to contaminate the environment, and subsequently reduce transmission. More importantly, because hookworms are blood feeders, a partial reduction in adult worm burden equates to a decrease in pathology, notably iron-deficiency anemia [44]. Mathematical modeling of schistosomiasis in China showed that elimination of the parasite could be attained using an antifecundity vaccine that targets egg output with 75% efficacy [48], and it is likely that a similar scenario applies to long-term elimination of soil-transmitted helminths such as hookworms. An orthologue of Ac-APR-1 has been reported from the major human hookworm, N. americanus [23]. Na-APR-1 is structurally and antigenically very similar to Ac-APR-1 and also functions as a hemoglobinase [23]. For this reason, we believe that APR-1 is now the major vaccine antigen from the adult stage of the parasite, and as such, Na-APR-1 should undergo process development and enter into Phase I clinical trials as a vaccine for human hookworm infection. This vaccine strategy is now being implemented for a larval hookworm antigen, with Phase 1 human trials using ASP-2 formulated with Alhydrogel already underway [49]. Based on the data reported here, APR-1 may also be selected for downstream process development, manufactured under good clinical manufacturing processes, and tested in the clinic.
Supporting Information
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession numbers for the gene products mentioned in this paper are Ac-APR-1 (U34888) and Na-APR-1 (AJ245459).
Patient Summary
Background
Hookworms are parasites of the intestines. They can infect many animals, including dogs, cats, and people. Worldwide, about one person in five has a hookworm infection. Most of these one billion people live in tropical countries. Hookworm is not spread from person to person, because at one stage of its lifecycle, the parasite needs to be in the soil. In areas where hookworm is common, people who have contact with soil that contains human feces are at high risk of infection; because children play on soil and often go barefoot, they have the greatest risk. Infection leads to blood loss and a decrease in the amount of iron, and this causes anemia (i.e., because of a lack of iron, the blood cannot carry oxygen efficiently). There are effective drugs to treat the infection, but they do not prevent the patient from becoming re-infected. Making a vaccine against hookworm is therefore a priority. Some vaccines for use in animals have already been developed, but their effectiveness is limited to one stage of the hookworm's lifecycle. The aim is to find a vaccine that works against more than one of the stages that the parasite passes through in its lifecycle.
What Did the Researchers Do and Find?
The researchers focused on two enzymes the parasite needs in order to live. Building on earlier research and using a species of hookworm that affects dogs, the researchers aimed to make these enzymes the “target” of a vaccine. They first vaccinated dogs, then infected them with hookworm. These dogs had fewer parasites than dogs that had not been vaccinated. Most importantly, vaccinated dogs were protected against blood loss, and most did not develop anemia. Laboratory tests confirmed that the target enzymes had been damaged.
What Do These Findings Mean?
This is the best result so far for a hookworm vaccine used in dogs. The authors believe that, as well as reducing parasite numbers, the vaccine reduces the ability of the parasite to take in blood, which would explain the reduction in anemia. The researchers have called for trials to begin with a vaccine targeted against similar enzymes in the species of hookworm that most commonly affects humans.
Where Can I Get More Information Online?
The US Centers for Disease Control have a fact sheet on hookworm:
http://www.cdc.gov/ncidod/dpd/parasites/hookworm/factsht_hookworm.htm.
The Sabin Vaccine Institute has an overview of the Human Hookworm Vaccine Initiative:
http://www.sabin.org/hookworm.htm.
This work was supported by a grant from the Bill and Melinda Gates Foundation awarded to the Sabin Vaccine Institute. AL is supported by a Career Development Award from the National Health and Medical Research Council of Australia. JMB is supported by an International Research Scientist Development Award (1K01 TW00009) from the Fogarty Center. For technical assistance and/or helpful advice, we thank Yan Wang, Lilian Bueno, Azra Dobardzic, Reshad Dobardzic, Andre Samuel, Sonia Ahn, Aaron Witherspoon, Clay Winters, Estelle Schoch, John Hawdon, and Philip Russell. We would like to acknowledge Joe Cohen and Sylvie Cayphas of GlaxoSmithKline Biologicals (Rixensart, Belgium) for providing AS03 and technical assistance with formulation.
Citation: Loukas A, Bethony JM, Mendez S, Fujiwara RT, Goud GN, et al. (2005) Vaccination with recombinant aspartic hemoglobinase reduces parasite load and blood loss after hookworm infection. PLoS Med 2(10): e295.
Abbreviations
APR-1
Ancylostoma caninum aspartic protease 1
AS03GlaxoSmithKline Adjuvant System 01
ELISAenzyme-linked immunosorbent assay
epgeggs per gram of feces
Hbhemoglobin
L3third stage larvae
==== Refs
References
Hotez PJ Brooker S Bethony JM Bottazzi ME Loukas A Hookworm infection N Engl J Med 2004 351 799 807 15317893
Bethony J Chen J Lin S Xiao S Zhan B Emerging patterns of hookworm infection: influence of aging on the intensity of Necator infection in Hainan Province, People's Republic of China Clin Infect Dis 2002 35 1336 1344 12439796
Loukas A Constant SL Bethony JM Immunobiology of hookworm infection FEMS Immunol Med Microbiol 2005 43 115 124 15681140
Albonico M Smith PG Ercole E Hall A Chwaya HM Rate of reinfection with intestinal nematodes after treatment of children with mebendazole or albendazole in a highly endemic area Trans R Soc Trop Med Hyg 1995 89 538 541 8560535
Miller TA Effect of age of the dog on immunogenic efficiency of double vaccination with x-irradiated Ancylostoma caninum larvae Am J Vet Res 1965 26 1383 1390 5882555
Miller TA Influence of age and sex on susceptibility of dogs to primary infection with Ancylostoma caninum
J Parasitol 1965 51 701 704 5857264
Miller TA Persistence of immunity following double vaccination of pups with x-irradiated Ancylostoma caninum larvae J Parasitol 1965 51 705 711 5857265
Goud GN Zhan B Ghosh K Loukas A Hawdon JM Cloning, yeast expression, isolation and vaccine testing of recombinant Ancylostoma -secreted protein 1 (ASP–1) and ASP-2 from Ancylostoma ceylanicum
J Infect Dis 2004 189 919 929 14976610
Mendez S Zhan B Goud GN Ghosh K Dobardzic A Effect of combining the larval antigens Ancylostoma secreted protein 2 (ASP-2) and metalloprotease 1 (MTP-1) in protecting hamsters against hookworm infection and disease caused by Ancylostoma ceylanicum
Vaccine 2005 23 3123 3130 15837211
Hotez PJ Ashcom J Zhan B Bethony J Loukas A Effect of vaccination with a recombinant fusion protein encoding an astacin-like metalloprotease (MTP-1) secreted by host-stimulated Ancylostoma caninum third-stage infective larvae J Parasitol 2003 89 853 855 14533704
Brooker S Bethony JM Rodrigues L Alexander N Geiger S Epidemiological, immunological and practical considerations in developing and evaluating a human hookworm vaccine Expert Rev Vacc 2005 4 35 50
Williamson AL Brindley PJ Knox DP Hotez PJ Loukas A Digestive proteases of blood-feeding nematodes Trends Parasitol 2003 19 417 423 12957519
Dalton JP Brindley PJ Knox DP Brady CP Hotez PJ Helminth vaccines: From mining genomic information for vaccine targets to systems used for protein expression Int J Parasitol 2003 33 621 640 12782060
Tort J Brindley PJ Knox D Wolfe KH Dalton JP Proteinases and associated genes of parasitic helminths Adv Parasitol 1999 43 161 266 10214692
Knox DP Redmond DL Newlands GF Skuce PJ Pettit D The nature and prospects for gut membrane proteins as vaccine candidates for Haemonchus contortus and other ruminant trichostrongyloids Int J Parasitol 2003 33 1129 1137 13678629
Knox DP Skuce PJ Newlands GF Redmond DL Kennedy MW Harnett W Nematode gut peptidases, proteins and vaccination Parasitic nematodes: Molecular biology, biochemistry and immunology 2001 New York CAB International 247 268
Smith WD Skuce PJ Newlands GF Smith SK Pettit D Aspartyl proteases from the intestinal brush border of Haemonchus contortus as protective antigens for sheep Parasite Immunol 2003 25 521 530 15053773
Smith WD Newlands GF Smith SK Pettit D Skuce PJ Metalloendopeptidases from the intestinal brush border of Haemonchus contortus as protective antigens for sheep Parasite Immunol 2003 25 313 323 14507329
Geldhof P Claerebout E Knox D Vercauteren I Looszova A Vaccination of calves against Ostertagia ostertagi with cysteine proteinase enriched protein fractions Parasite Immunol 2002 24 263 270 12060320
Don TA Jones MK Smyth D O'Donoghue P Hotez P A pore-forming haemolysin from the hookworm, Ancylostoma caninum
Int J Parasitol 2004 34 1029 1035 15313129
Williamson AL Lecchi P Turk BE Choe Y Hotez PJ A multi-enzyme cascade of hemoglobin proteolysis in the intestine of blood-feeding hookworms J Biol Chem 2004 279 35950 35957 15199048
Loukas A Bethony JM Williamson AL Goud GN Mendez S Vaccination of dogs with a recombinant cysteine protease from the intestine of canine hookworms diminishes the fecundity and growth of worms J Infect Dis 2004 189 1952 1961 15122534
Williamson AL Brindley PJ Abbenante G Prociv P Berry C Cleavage of hemoglobin by hookworm cathepsin D aspartic proteases and its potential contribution to host specificity FASEB J 2002 16 1458 1460 12205047
Hotez PJ Ashcom J Bin Z Bethony J Williamson A Effect of vaccinations with recombinant fusion proteins on Ancylostoma caninum habitat selection in the canine intestine J Parasitol 2002 88 684 690 12197114
Stoute JA Slaoui M Heppner DG Momin P Kester KE A preliminary evaluation of a recombinant circumsporozoite protein vaccine against Plasmodium falciparum malaria. RTS,S Malaria Vaccine Evaluation Group N Engl J Med 1997 336 86 91 8988885
Shifrine M Taylor NJ Rosenblatt LS Wilson FD Comparison of whole blood and purified canine lymphocytes in a lymphocyte-stimulation microassay Am J Vet Res 1978 39 687 690 646204
Williamson AL Brindley PJ Abbenante G Datu BJ Prociv P Hookworm aspartic protease, Na-APR-2, cleaves human hemoglobin and serum proteins in a host-specific fashion J Infect Dis 2003 187 484 494 12552433
Bungiro R Cappello M Hookworm infection: New developments and prospects for control Curr Opin Infect Dis 2004 17 421 426 15353961
Stassens P Bergum PW Gansemans Y Jespers L Laroche Y Anticoagulant repertoire of the hookworm Ancylostoma caninum
Proc Natl Acad Sci U S A 1996 93 2149 2154 8700900
Cappello M Vlasuk GP Bergum PW Huang S Hotez PJ
Ancylostoma caninum anticoagulant peptide: A hookworm-derived inhibitor of human coagulation factor Xa Proc Natl Acad Sci U S A 1995 92 6152 6156 7597095
Chadderdon RC Cappello M The hookworm platelet inhibitor: Functional blockade of integrins GPIIb/IIIa (alphaIIbbeta3) and GPIa/IIa (alpha2beta1) inhibits platelet aggregation and adhesion in vitro J Infect Dis 1999 179 1235 1241 10191228
Roche M Layrisse M The nature and causes of “hookworm anemia” Am J Trop Med Hyg 1966 15 1029 1102 5334793
Girod N Brown A Pritchard DI Billett EE Successful vaccination of BALB/c mice against human hookworm (Necator americanus) The immunological phenotype of the protective response Int J Parasitol 2003 33 71 80 12547348
Fujiwara RT Loukas A Mendez S Williamson AL Bueno LL Vaccination with irradiated Ancylostoma caninum third stage larvae induces a Th2-like response in dogs Vaccine 2005 In press
de Oliveira Mendes C Paraguai de Souza E Borja-Cabrera GP Maria Melo Batista L Aparecida dos Santos M IgG1/IgG2 antibody dichotomy in sera of vaccinated or naturally infected dogs with visceral leishmaniosis Vaccine 2003 21 2589 2597 12744895
Chabalgoity JA Moreno M Carol H Dougan G Hormaeche CE
Salmonella typhimurium as a basis for a live oral Echinococcus granulosus vaccine Vaccine 2000 19 460 469 11027809
Boag PR Parsons JC Presidente PJ Spithill TW Sexton JL Characterisation of humoral immune responses in dogs vaccinated with irradiated Ancylostoma caninum
Vet Immunol Immunopathol 2003 92 87 94 12628766
Pearce EJ MacDonald AS The immunobiology of schistosomiasis Nat Rev Immunol 2002 2 499 511 12094224
Lawrence JD The ingestion of red blood cells by Schistosoma mansoni
J Parasitol 1973 59 60 63 4687511
Hu W Yan Q Shen DK Liu F Zhu ZD Evolutionary and biomedical implications of a Schistosoma japonicum complementary DNA resource Nature Genet 2003 35 139 147 12973349
Knox DP Smith WD Vaccination against gastrointestinal nematode parasites of ruminants using gut-expressed antigens Vet Parasitol 2001 100 21 32 11522403
Verity CK McManus DP Brindley PJ Vaccine efficacy of recombinant cathepsin D aspartic protease from Schistosoma japonicum
Parasite Immunol 2001 23 153 162 11240906
Vilanova M Teixeira L Caramalho I Torrado E Marques A Protection against systemic candidiasis in mice immunized with secreted aspartic proteinase 2 Immunology 2004 111 334 342 15009435
Stoltzfus RJ Dreyfuss ML Chwaya HM Albonico M Hookworm control as a strategy to prevent iron deficiency Nutr Rev 1997 55 223 232 9279058
Hotez PJ Zhan B Bethony JM Loukas A Williamson A Progress in the development of a recombinant vaccine for human hookworm disease: The Human Hookworm Vaccine Initiative Int J Parasitol 2003 33 1245 1258 13678639
Bethony J Loukas A Smout M Brooker S Mendez S Antibodies against a secreted protein from hookworm larvae reduce the intensity of infection in humans and vaccinated laboratory animals FASEB J 2005 E-pub ahead of print
Goud GN Bottazzi ME Zhan B Mendez S Deumic V Expression of the Necator americanus hookworm larval antigen Na-ASP-2 in Pichia pastoris and purification of the recombinant protein for use in human clinical trials Vaccine 2005 23 4754 4764 16054275
Williams GM Sleigh AC Li Y Feng Z Davis GM Mathematical modelling of schistosomiasis japonica: Comparison of control strategies in the People's Republic of China Acta Trop 2002 82 253 262 12020899
Hotez P Bethony J Bottazzi ME Brooker S Buss P Hookworm: “The great infection of mankind.” PLoS Med 2005 2 177 181
|
16231975
|
PMC1240050
|
CC BY
|
2021-01-05 10:38:15
|
no
|
PLoS Med. 2005 Oct 4; 2(10):e295
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020295
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1618779610.1371/journal.pmed.0020296Research ArticleCancer BiologyGenetics/Genomics/Gene TherapyOncologyOncologyA Critical Reassessment of the Role of Mitochondria in Tumorigenesis Role of Mitochondria in TumorigenesisSalas Antonio
1
2
*Yao Yong-Gang
3
Macaulay Vincent
4
Vega Ana
2
5
Carracedo Ángel
1
2
5
Bandelt Hans-Jürgen
6
1Unidade de Xenética, Instituto de Medicina Legal, Facultade de Medicina, Universidad de Santiago de Compostela, Galicia, Spain,2Centro Nacional de Genotipado (CeGen), Hospital Clínico Universitario, Santiago de Compostela, Galicia, Spain,3Key Laboratory of Cellular and Molecular Evolution, Kunming Institute of Zoology, Chinese Academy of Sciences, Kunming, Yunnan, China,4Department of Statistics, University of Glasgow, Glasgow, Scotland, United Kingdom,5Fundación Pública Galega de Medicina Xenómica (FPGMX), Hospital Clínico Universitario, Universidad de Santiago de Compostela, Galicia, Spain,6Department of Mathematics, University of Hamburg, Hamburg, GermanyTurnbull Doug Academic EditorUniversity of Newcastle upon TyneUnited Kingdom*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: AS, YGY, VM, AV, AC, and HJB designed the study, analyzed the data, and contributed to writing the paper.
11 2005 4 10 2005 2 11 e2962 5 2005 25 7 2005 Copyright: © 2005 Salas et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Casting Doubt on the Role of Mitochondria in Tumorigenesis
Background
Mitochondrial DNA (mtDNA) is being analyzed by an increasing number of laboratories in order to investigate its potential role as an active marker of tumorigenesis in various types of cancer. Here we question the conclusions drawn in most of these investigations, especially those published in high-rank cancer research journals, under the evidence that a significant number of these medical mtDNA studies are based on obviously flawed sequencing results.
Methods and Findings
In our analyses, we take a phylogenetic approach and employ thorough database searches, which together have proven successful for detecting erroneous sequences in the fields of human population genetics and forensics. Apart from conceptual problems concerning the interpretation of mtDNA variation in tumorigenesis, in most cases, blocks of seemingly somatic mutations clearly point to contamination or sample mix-up and, therefore, have nothing to do with tumorigenesis.
Conclusion
The role of mitochondria in tumorigenesis remains unclarified. Our findings of laboratory errors in many contributions would represent only the tip of the iceberg since most published studies do not provide the raw sequence data for inspection, thus hindering a posteriori evaluation of the results. There is no precedent for such a concatenation of errors and misconceptions affecting a whole subfield of medical research.
The role of mitochondria in tumorigenesis remains unclear; in this paper Salas and colleagues raise concerns over many published studies
==== Body
Introduction
For more than two decades human mitochondrial DNA (mtDNA) has widely been used as a versatile tool to investigate different genetic aspects such as the origin and migration patterns of human populations or criminal casework in the forensic field. Specific mutations in the mtDNA genome are also suggested to be responsible for human diseases such as Leber hereditary optic neuropathy, myoclonic epilepsy associated with ragged-red fiber disease, MELAS syndrome, deafness, and inherited adult-onset diabetes (see [1] for a recent review). In the past few years, the putative role of mtDNA in cancer has received special attention. While many studies seem to support an active role of mtDNA in tumorigenesis [2,3], there are many caveats, and the issue has been highly debated [4–6].
Owing to the multiple steps involved in mtDNA analysis, systematic errors in mtDNA sequences are often found in the anthropological and forensic literature [7–11]. Thus one should also expect to detect similar problems in clinical investigations. More than half of published mtDNA sequencing studies contain obvious errors, no matter in which journal the investigation is published [12]. The consequences of such errors can be more or less dramatic depending on the subject or particular case under study. In the forensic context, a single mistake can lead to the false exclusion of an individual as the source of the biological material left at a crime scene or to a mismatch in comparisons of an mtDNA profile with forensic databases. Systematic errors can also lead to biological dogmas such as the maternal inheritance of mtDNA being brought into question [13]. In an oncogenetic context, flawed sequence data can lead to a conclusion of false association between seemingly causal variants and tumor instability.
A phylogenetic approach to the analysis of mtDNA profiles (in which the sequences under consideration are compared with the current database of complete sequences that make up the global mtDNA phylogeny) has been shown to be useful for assessing the accuracy of mtDNA data [14]. In the clinical context, such an approach allows mtDNA sequences to be assigned to haplogroups alias monophyletic clades (that is, groups of all mtDNA sequences derived from a common ancestor), according to the haplogroup-specific mutations they harbor, and offers clues for pinpointing flaws. The mutational processes that lead to a cancer (or the mutations accumulating during cancer proliferation) could hardly reproduce by chance (mutation by mutation) the long evolutionary routes between distant mtDNA haplogroups. Therefore, when a tumor sample is apparently distinguished from the corresponding normal tissue sample by (nearly) all the mutations distinguishing two very different haplogroups, then the only conclusion is that one of the two samples was contaminated or exchanged by mistake.
Methods
MtDNA Databases for the Interpretation of Human Population Variation
The use of large worldwide databases is of great help for identifying matching sequences and confirming membership to specific haplogroups, as well as for obtaining information about their geographical distribution (phylogeography). In order to study mtDNA variation properly it is necessary to take the full body of published mtDNA studies into consideration. While MITOMAP (http://www.mitomap.org/) provides a useful (but incomplete) listing of single mutations that have appeared in the older medical literature, direct reference to population databases of complete or nearly complete mtDNA sequences allows the inference of mutations that have occurred on evolutionary pathways between reconstructed ancestral sequences. Thus, a snapshot of the global mtDNA phylogeny and some of its representatives in all continents is given by the complete sequences of Ingman et al. [15], although the accompanying diagrams are devoid of the information that a medical geneticist would need, namely, a reconstruction of the coding-region mutations along the estimated phylogeny. The data of Herrnstadt et al. [16,17], which comprise only the coding region, give additional information, with the emphasis on European mtDNA, while Kong et al. [18] data contain East Asian complete genomes. Coble et al. [19] provided a considerable number of new complete mtDNA genomes, which were preselected according to frequent control-region haplotypes found in Europe. Most recently, Palanichamy et al. [20] obtained 75 complete sequences from haplogroup N sampled in India, and Achilli et al. [21] published 62 complete mtDNAs, covering most of the basal variation of haplogroup H.
The deeper parts of the global mtDNA phylogeny are expressed through a system of nested haplogroups, which are encoded by strings of letters and numbers in alternation, following specific rules [22]. In what follows we make use of this haplogroup nomenclature [18,20] in order to reference the pertinent sections of the phylogeny.
Results
The Unsuitability of an “Allelic” Approach
Uniparental markers, such as mtDNA, were newcomers to the field of human genetics, where classical nuclear markers had been predominant. Consequently, the analysis of the new markers proceeded in the traditional way by treating the segregating nucleotides at each polymorphic position in the sequence as alleles and treating each position independently so that haplotypes were disrupted, and the strong association of certain mutations along the phylogeny was disregarded completely. The cumulative listing of mtDNA mutations observed in patients or controls [23,24] is therefore not only rather uninformative but often misleading.
The strategy followed by Nishikawa et al. [25] will serve as a paradigmatic example of what is conceptually inappropriate. These authors sequenced the entire mtDNA genome of two individuals with hepatocellular carcinoma (HCC) and a liver specimen from one control, as well as the D-loop region (only nucleotides 100–600) of another six controls and 69 HCC specimens of Japanese subjects. They used an arbitrary mtDNA complete genome deposited in GenBank (accession number J01415) as a reference sequence. This sequence differs from the revised Cambridge reference sequence (rCRS [26]) by four mutations: A4985G, C11335T, C14766T (constituting three of the 11 errors of the original Cambridge reference sequence [27]), and A750G; it is actually an artificial sequence phylogenetically related to haplogroup H. When comparing the mtDNA of the control liver specimen with the J01415 sequence, they found only three differences. Judging from the context in their paper, the three differences, however, must be A263G, 315+C, and T489C. Thus, all sequences were erroneously scored at the four positions 750, 4985, 11335, and 14766. The presence of the substitution T489C indicates that the control lineage belongs to one of the haplogroups M and J, very likely to the former because haplogroup J is virtually absent in Japanese, but this is at odds with the meager number of differences to the rCRS. Even more alarmingly, since all samples screened for region 100–600 were claimed to harbor those three mutations in the D-loop, we would then expect no single sequence from haplogroup N (which embraces the East Eurasian haplogroups A, B, F, etc.) in 78 individuals. This is very unlikely when we consider the haplogroup distribution pattern in large Japanese mtDNA datasets [28], which testify to more than 30% of haplogroup N sequences. It therefore seems that contamination with some haplogroup M sequence had affected the samples. Finally, the authors compared the mtDNAs of two cancerous tissue specimens with sequence J01415 and found as many as 67 and 77 mutations, respectively (half of them also present in the paired noncancerous tissue specimens of the two patients). The two cancerous tissues seem to share several mutations, especially in the region 11000–16000 of the mtDNA genome rather than with their matched noncancerous tissues (see their Figure 1), which again would make contamination or sample mix-up plausible as a cause for the incidence of seemingly somatic mtDNA mutations in HCCs. There is no consistent way to allocate the mutations that separate the rCRS from the root of haplogroup H (or R) or the mutations distinguishing M and R in their Figure 1 (see [25]). It seems that massive oversight of mutations on the one hand, and contamination on the other, have shaped the picture presented. Since the total data obtained by the authors are not reported mutation by mutation, the likely causes of the sequencing dilemma cannot be reconstructed more precisely.
Figure 1 Portion of the Worldwide mtDNA Phylogeny That Explains the Most Relevant Contamination/Sample Mix-Up Episodes Erroneously Interpreted as mtDNA Instabilities in Several Kinds of Tumors or Detected as Germline Mutations, as Commented in the Text
Capital letter-number codes designate haplogroups; bold bars indicate intermediate branching points as inferred from the total mtDNA phylogeny. Variation at position 16519, length polymorphisms of long C-stretches in HVS-I and II, and dinucleotide repeats at 522–523 are disregarded. All mutations are transitions except for those suffixed by A, G, C, or T (transversion) or del (deletion) or + (insertion of the specified nucleotide). Parallel mutations are underlined. Somatic (red squares) and germline (pink circles) mutations are indicated on the left side of each position in the tree as they appear in their original studies.
Yeh et al. [29] studied mtDNA in papillary thyroid carcinomas (PTCs). A whole paragraph of the discussion is dedicated to two (C7521A, A10398G) of the three missense mutations found in PTC cases. The authors failed to recognize that both C7521A and A10398G are familiar mutations in the mtDNA phylogeny. The latter is shared by nearly all haplotypes outside haplogroup N, whereas the former is common to virtually all members of the major African haplogroups L0, L1, and L2. It then seems that the authors' statement that “the 10398A>G and 7521G>A variants might not be totally innocuous…” (p. 2064) and that “one might speculate that these somatic mtDNA mutations are low penetrance modifiers of tumour risk…”(p. 2064) is most implausible, given the ubiquity of this mutational pair on the African continent. It is also remarkable that a meager number of 30 controls and nine fetal tissues without heteroplasmy at positions 7521 and 10398 led the authors to speculate that “this suggests that when these mutations are somatic, they are specific to PTCs (p. 2064).” As we shall see below, seemingly heteroplasmic or somatic mutations may be the result of contamination or sample mix-up and therefore would necessitate more sequencing and cloning efforts.
Yeh et al. [29] also followed a paradigm that is nearly ubiquitous in all those studies about the role of mtDNA and tumorigenesis, namely, the straightforward comparison of a patient group with an arbitrary control group, by counting mutations relative to the rCRS. A seemingly larger number of mutations in the patient group would normally reflect sampling effects in that different parts of the phylogeny are covered by the mtDNAs of patients compared to controls. In other words, controls and cases do not necessarily represent the same population and ethnic matching; this provokes a well-known effect in popular association studies that leads to spurious association between probands and the polymorphism/mutation under study. For example, only one (G15179A) out of 16 mutations that were found in PTC but not in controls and fetal tissues [29] is apparently a mutation not yet reported in normal mtDNA genomes from worldwide studies of the past 5 y. Worse, at least three PTC mtDNAs contribute more than one mutation to the list, which are inherited from the particular basal branch of the worldwide phylogeny the mtDNA belongs to. One haplogroup L1b lineage is responsible for mutations T710C and T3308C that are characteristic of this haplogroup as well as mutation T7389C specific to the superhaplogroup L1. Similarly, potential Native American mtDNAs from haplogroups D1 and B2 could have contributed two mutations each. The phylogenetic linkage of mutations therefore violates the tacit assumption of independence behind any claims of “significance.”
The most recent claim that one “can now add cancer to the list of mitochondrial diseases” (see [30], p. 724) in that mtDNA mutations are associated with a predisposition to prostate cancer should also be received with skepticism. The fact that a known pathogenic mutation such as T8993G can influence the rate of tumor growth cannot alone corroborate this claim—nor can a simple correlation study that contrasts cytochrome oxidase subunit I polymorphisms found in patients with those found in controls. For instance, mutation T6253C that is believed to be associated with prostate cancer [30] belongs to the characteristic mutations for the European haplogroup H15 [21] and both East Asian haplogroups D5 [18] and M13 [31]. To our knowledge it has not been reported yet that, in Japan where D5 and M13 thrive, a considerable number of men suffered from rapidly growing prostate cancer.
Alleged Mutational Hotspots
Methodological procedures can be prone to sequence artifacts that can erroneously be interpreted as mtDNA mutational hotspots. Because of their nature, mutational hotspots emerge frequently in the mtDNA phylogeny and for this reason are well known in human population studies. A familiar example within the clinical literature is the unstable homopolymeric “C” track in the hypervariable segment II (HVS-II) region, from positions 303 to 309, but also positions 146, 150, and 152 [6] in the same segment. Turning it the other way around, rare or stable diagnostic variants are extremely unlikely to be mutational hotspots.
We highlight the unusual findings in Khrapko et al. [32] in this regard. After analyzing a short coding-region fragment of 100 bp by mutational spectrometry, these authors reached the conclusion that different human tissues and cells contained a remarkably similar set of hotspot point mutations. However, we observe that the hotspots they reported are correlated very poorly with the mutational spectra inferred from human populations. For instance, their two most important outstanding “hotspots” (positions 10068 and 10098 in their Figure 3) have never been detected in complete genome sequencing analysis (despite there being more than 1,700 complete mtDNA genomes available in the literature, covering most of the basal worldwide mtDNA phylogeny). To our knowledge, the remaining so-called hotspots have never been found in the human population literature either, with the exception of position 10084, which is associated with J1, K, and L1c lineages [16]. The real causes of this unexpected result remain obscure, since their methodology used to detect sequence variants (mutational spectrometry) is seldom used in the field, so one cannot exclude the possibility that it is strongly affected by artifacts.
The study by Reddy et al. [33] of patients with myelodysplastic syndromes also suggested the presence of novel mutational hotspots. Among them, positions 7264, 7594, and 7595 have never been detected in population studies, whereas others such as 7289 have been found in only a single sequence (Homo sapiens isolate T1–12 mitochondrion, complete genome [19]). Also unrealistic is the fact that 25 out of 52 mutational events listed in their Table III are transversions, whereas another 18 “instabilities” are indels; but only nine transitions (by far the most common type of change in mtDNA) were reported (one of these transitions [A7768G] is diagnostic for haplogroup U5b [20]). The transition:transversion ratio for their data contrasts significantly with very conservative estimates taken from hundreds of human population studies; additionally the extremely high prevalence of indels is certainly unrealistic. For similar and additional reasons, this study has been questioned by others [5]. Unexpected results must be corroborated with standard methodology and by independent studies; in this sense, the commitment of Reddy et al. [33] has not been published yet. In addition, we agree with Gattermann et al. [5] in that the detection of mutations within primer annealing sites is at least unorthodox. The explanations of Reddy et al. [33] are certainly not convincing and, of course, do not explain the most striking fact: why have these hotspots not been detected (even as private variants) anywhere else in population studies?
Contamination and Sample Mix-Up
The recent work of Bandelt et al. [10] analyzed the causes and consequences of artificial recombinants, focusing attention on the forensic and population genetic literature. As predicted, mtDNA analysis in clinics does not escape the problem of contamination and sample mix-up. All the studies that we comment on below have a common denominator: contamination or sample mix-up of the tumor samples under study with exogenous mtDNAs. Typically, these findings usually lead to innocent erroneous interpretations and the concomitant development of a biological explanation or the invocation of a theory that would justify the role of such variants in tumorigenesis. A classical case constitutes the finding of three “somatic” homoplasmic mutations (T710C, T1738C, and T3308C) in colorectal tumor V478 [2]: these rather rare mutations all belong to the sequence motif for haplogroup L1b (see also [34]).
Fliss et al. [3] provide us with a pertinent example in the analysis of mtDNA sequences in tumor studies. Patient 884 (their Table 1; bladder cancer) shows a total of five mutations (T10071C, T10321C, A10792G, C10793T, and C12049T) all of which have been found in a haplogroup L1c2 lineage, namely, no. 173 in Herrnstadt et al. [16], which is related to the African lineage no. 48 in Ingman et al. [15]. Therefore, these mutations are extremely unlikely to have anything to do with tumorigenesis but rather represent an instance of contamination or sample mix-up involving a specific L1c2 mtDNA (Figure 1). Corroborating evidence comes from their Supplemental Table 1 (www.sciencemag.org/feature/data/1048413.shl, which lists the so-called new mtDNA polymorphisms detected at the time as being shared by matched cancerous and normal tissues. As many as 13 mutations recorded for the bladder cancer cases are also seen in lineage no. 173 of Herrnstadt et al. [16], including its seemingly private mutations A633G, A723G, T5580C, and T15672C. Not surprisingly, the five somatic mutations claimed to be somatic in Patient 884 were missed as polymorphisms in the bladder cancer patients. This strongly suggests that two amplicons of normal tissue from Patient 884 (one covering 10071–10793 and the other including 12049) were exchanged with (or contaminated by) tissues stemming from some other patient or patients.
The polymorphisms listed in Supplemental Table 1 of Fliss et al. [3] testify to further problems. Among the mutations that were found in lung cancer patients, four mutations point to haplogroup L1c and three additional ones (A2308G, C11257T, and T11899C) to a very specific branch of subhaplogroup L1c1a [15,35]. This being the case, the three mutations G2758A, C8655T, and A9072G from the evolutionary path between rCRS and haplogroup L1c should have been recorded there as well—but they were not. Neither had G2758A been reported for bladder cancer, although L1c was present there as well. The presence of another mtDNA haplogroup of African ancestry is documented in that table of polymorphisms, namely, the six mutations T1738C, A2768G, T3308C, A8248G, T12519C, and A14769G belong to the characteristic motif of haplogroup L1b. Five of them are associated with lung cancer, four with bladder cancer, and three with head and neck cancer patients. This means that a total of 1 + 2 + 3 = 6 mutations must have been missed in individual patients (as well as C8655T in the head and neck cancer case) since natural back mutations in such great number would be unrealistic. We conclude that sample contamination or massive oversight of mutations must have been the rule rather than the exception in Fliss et al. [3].
Jerónimo et al. [36] claim to have demonstrated the existence of specific patterns of somatic mtDNA mutations in prostate cancer. These authors also reported a spectacular case in the clinical literature: 18 somatic mutations detected in one patient (see their Table 1, patient 1). Surprisingly, many of these alterations conform to the familiar western European mtDNA haplogroup W [20]: A189G, T204C, G207A, A3505G, C11674T, A11947G, T12414C, and C12705T (Figure 1). In addition, their Table 1 suspiciously contains variants such as A3480G that identify haplogroup K, as well as A12308G and G12372A, which are characteristic of the larger haplogroup U, in which, haplogroup K is nested. Although A235G is a good candidate for haplogroup A, it too has been found within haplogroup K (USA.CAU.001306 in the SWGDAM database [37]), while other mutations such as T146C, A16183C (erroneously reported as A16183G in Jerónimo et al. [36]), and T16189C can be found on many haplogroup backgrounds (including haplogroup W). Such an unbelievable departure from random expectation represents a good candidate for cross-contamination between at least two different samples: one from haplogroup W and one from K. Therefore, there is no need to invoke the effect of endogenous factors or catastrophic mutagenic effects of exogenous exposure for this “hypermutated individual”: “Intriguingly, this patient worked for many years at a chemical plant” (see [36], p. 5196).
Similarly, Kirches et al. [38] carried out pairwise comparisons between glioma samples and adjacent brain tissues of 55 patients. Strikingly, patient 2 (their Table 1, p. 536) accumulated a total of 17 homoplasmic transitions, in contrast with the rest of the instabilities reported, which are all length variations of (di-)nucleotide repeats in the control region, except for one homoplasmic change between glioblastoma and normal tissue (namely, at position 72). The somatic mutations reported for patient 2 split into two mutational motifs with respect to rCRS: T195C, T4646C, T5999C, A6047G, A12937G, T13124C, C14620T, C16134T, A16293G, T16356C, and T16519C in the glioblastoma and G185A, T204C, C295T, A5198G, T16126C, and T14798C in the normal tissue (Figure 1). Here we assume that Kirches et al. [38] have misrecorded A12937G as A12936G and misassigned A5198G to the glioblastoma or interchanged the nucleotides A and G at 5198 by mistake. Our interpretation thus posits that 11 of the 13 gliostoma mutations confirm to the motif of a particular branch of haplogroup U4a [6,17,39], thus leaving only A16293G and T13124C as potential private mutations. All six normal tissue mutations point to the (yet unnamed) branch of haplogroup J1c defined by A5198G (compare with Coble et al. [19] and Herrnstadt et al. [16,17]; see Figure 1). Although the authors identify most of these substitutions as known polymorphisms in humans, they failed to recognize the most plausible justification for these results, namely, contamination or sample mix-up: “Whatever the mechanisms of mutation may be, the fast selection of certain tumor mitochondria, as demonstrated by Polyak et al. (1998), offers the only simple explanation for the accumulation of 17 homoplasmic mutations in a single tumor sample…” ([38], p. 537]). As usual, unfortunate results lead to the formulation of unjustifiable conclusions: “The occurrence of mutations in tumors at known polymorphic mtDNA sites and the dominance of transitions suggest a common mechanism generating germline mtDNA polymorphisms and somatic mutations” ([38], p. 536).
Wong et al. ([40], p. 3870) stated that “Kirches et al. (2001) reported 19 somatic mtDNA mutations found in a single glioblastoma. One of our [medulloblastoma] patients harbored 11 somatic mtDNA mutations.” Unfortunately, this study [40] is another example of an artifactual result: leaving aside a length polymorphism of a C-stretch (in HVS-II), the following 11 mutations are recorded in Wong et al.'s patient (their Table 1, case no. 124): C151T, C182T, G246A, A297G, G317A, G7337A, G7521A, G7337A, T7389C, T15904C, and A15937G. Three of these mutations have actually been shifted by one position and should read G247A, G316A, and T15905C instead. Then except for T15905C and A15937G these mutations can be found in the African haplogroup L1c2 (Figure 1) [15,35], and compare with lineages no. 173 and no. 328 of Herrnstadt et al. [16,17]). It is then probably no coincidence that the “novel germ-line variation” reported in their Table 1b ([40], p. 3869) also testifies to two mutations, G10688A and T10810C (previously listed by Fliss et al. [3]), that would be observed with all haplogroup L0 and L1 sequences. In a recent report from the same group [41], patient HE19 was found to harbor 10 differences between DNA in liver cancer and normal tissue. Six of them (A189G, C194T, T195C, T199C, T204C, and G207A) in the tumor tissue point to haplogroup W (see, for instance, USA.CAU.000453 in the SWGDAM database; Figure 1), whereas three mutations, C456T, T489C, and (523–524)del, in the normal tissue could point to the East Asian haplogroup D5b (compare with THA.ASN.000058 in the SWGDAM database).
The study carried out by Liu et al. [42] on ovarian carcinomas includes at least one instance suspicious of sample mix-up or contamination. Namely, the normal tissue of patient OV88 carries mutation A249del (with no mutation at 489) characteristic of haplogroup F, whereas the tumor mtDNA shows the mutations T146C, T199C, T489C (characteristic of haplogroup M7c [43,44] plus T152C (which has also been observed in M7c sequences from East Asia; for example, in CHN.ASN.000337, JPN.ASN.000103, and THA.ASN.000048 from the SWGDAM database). Since this array of homoplasmic HVS-II mutations matches a pathway in the Chinese mtDNA phylogeny, we are led to conclude that the ovarian cancer mtDNA and the serum and normal tissue mtDNA of OV88 likely came from two different individuals. In their Table 1, many mutations point to several East Asian haplogroups (see [18]), and thus these mutations do not have anything to do with ovarian carcinomas (see also Figure 1).
Chen et al. [45] aimed at tracing somatic mutations in 16 cases of prostate cancer, by sequencing the highly hypervariable mtDNA control region in subjects with prostate cancer. Their Table 1 reports a patient (case 1) bearing eight “instabilities,” namely, mutations A16182C, A16183C, T16189C, C16232A, T16249C, G16274A, T16304C, and T16311C, leaving aside (522–523)del. Except for G16274A, which seems to be a private mutation, this is unmistakably a HVS-I haplotype belonging to haplogroup F1b, which is frequently found in China [18,43,44]. Not accidentally, all these variants were detected as heteroplasmies; therefore, this represents a perfect instance of contamination from a biological source carrying this F1b haplotype. Table 1 testifies to yet another highly suspicious example: case 4 carries nine somatic near-homoplasmic mutations (besides a C-stretch length polymorphism), among which A73G, G499A, A16182C, A16183C, T16189C, and T16217C happen to constitute the control-region mutations (relative to rCRS) of a genuine member of the East Asian haplogroup B4b [18], except for position 263, at which the vast majority of mtDNA sequences agree anyway but differ from rCRS. Additional information about case 4 is given in Chen et al. [46], where the B4bd characteristic mutation G15535A is reported. Moreover, among several serial tumor sections, one (C1) confirms haplogroup status B4b with G499A. On the other hand, the mtDNA of section C2 is clearly a member of haplogroup K1a, whereas the other sections would be compatible with haplogroup HV status. Therefore, multiple sample mix-up or contamination events are the most plausible cause underlying the seemingly close relationship among cases 1, 4, and 6, which they have instead interpreted as follows: “The nonrandom distribution of somatic mutations raises the possibility that certain constellations of sequence variation might be prone to somatic mutations”([45], p. 6472). When Chen et al. ([45], p. 6471) claimed that “the somatic mutations cannot be explained by experimental error or by contamination of nuclear mtDNA pseudogenes,” this may have been so, but profuse sample mix-up or cross-contamination perfectly explains the results.
The recent study of mtDNA control-region mutations in patients with esophageal squamous cell carcinoma [47] constitutes another good case for sample crossover. The mtDNA of the normal esophageal tissue of case 21 bears mutations C150T, C16067T, A16164G, A16171G, C16172T, A16182C, A16183C, T16362C, and T16519C and therefore belongs to haplogroup D5a. Interestingly, the blood sample shared all mutations with the normal esophageal tissue, except for having heteroplasmies at positions 16067, 16164, and 16171. In contrast, the tumor mtDNA contains C16184T, T16298C, T16443C, G16470A, G16471A, G16473A, and T16519C, which point to haplogroup M8a. Since only the mutations were reported that distinguish tumor mtDNA from blood mtDNA, we can expect that both actually share the motif A73G, A263G, T489C, T16189C, and C16223T [18]. The contrast between the two different sequences is well reflected by the pair CHN.ASN.000113 and CHN.ASN.000270 in the SWGDAM database. It then seems that one mutation in the blood sample was overlooked (at position 16266) and one in the tumor sample (at position 16319), but this cannot disturb the clear-cut haplogroup allocations. Case 20 [47] testifies to yet another sample mix: here tumor and adjacent normal tissue bear the same mtDNA variants (except for one heteroplasmy at 16266) in HVS-I, namely, C16185T, C16223T, C16260T, and T16298C, whereas the mtDNA sequence from blood is reported to have C16256T, C16270T, and A16399G. We are thus seeing here the clean contrast between a haplogroup Z and a (West Eurasian) haplogroup U5a1 sequence.
Discussion
Somatic mutations, in a heteroplasmic or a homoplasmic state, can occur in all kinds of tissues and body fluids of patients affected by cancer or genetic diseases and in healthy controls [4,48,49]. The problematic findings of Polyak et al. [2] and Fliss et al. [3] have been uncritically accepted [50] and cited in virtually every study of perceived mtDNA alterations in tumors. There seems to be a general expectation that the amount of somatic mutation can be elevated in tumors. The problem then is that instances reported with a whole array of seemingly somatic mutations would confirm this expectation and be taken at face value instead of as a hint at contamination or sample mix-up. The main consequence of such sequencing disasters is that most of these flawed results are not filtered out from the clinical literature, thus adding more noise to the interpretation of the role of mtDNA in the complex tumor process. This eventually leads to a vicious cycle of ill-based interpretations of mutational variation in tumors.
We have detected innumerable deficiencies in the clinical literature related to the analysis and interpretation of mtDNA data in tumor samples. There is no precedent that we know of in the genetics literature for such a high number of flawed papers (most of them published in high-rank journals), which affect a whole subfield of clinical research. Since what we show here is based on the extremely meager information generally available in these published reports, we have every reason to believe that this is only the tip of the iceberg. Note also that the database of coding-region variants in natural populations is still limited (although currently comprising more than 2,100 complete genomes), so some more mistakes in this literature await detection. Moreover, we must keep in mind that the phylogenetic approach used here is certainly not able to detect all errors.
We have found that the vast majority (>80%) of the studies dealing with potential functional implications of the mtDNA molecule in tumorigenesis (and providing data for inspection) are based on faulty data with surreal findings. The present report should lead us to reconsider the role of mtDNA in tumorigenesis. Probably we should abandon the exciting findings unleashed as a result of the many sequencing failures that accumulated during this last decade. A model consisting of basically two main stages [6]—namely, (i) accumulation of homoplasmic mutations in mtDNA-unstable sites during tumorigenesis, and (ii) a consequential effect on the cell physiology—is still valid in order to explain the mtDNA changes occurring during the tumoral process.
The clues to understanding the causes of pitfalls in mtDNA sequencing are extensively discussed [7–12,14,17]. Degraded DNA or extremely low quantities of DNA from old frozen samples or inadequately stored samples (in paraffin, for example) used in many clinical/oncogenetic studies would explain the notoriously low quality of sequence results as well as an elevated risk of contamination. For instance, one can only obtain very small amounts of DNA using the laser-capture microdissection technique employed by Chen et al. ([45]; see above) to retrieve cancerous and noncancerous samples from serial tissue sections. Contamination and sample degradation would greatly affect the quality of DNA during the subsequent processes and finally contribute to rich mtDNA heterogeneity in the sequence. In these situations of limited quantities of endogenous DNA, the clinical geneticist would do well to employ many of the checks for authenticity proposed for ancient DNA studies [51]). In a way, the current situation in the field of carcinogenesis and mtDNA resembles the state of the art of ancient DNA sequencing in those early days where loads of contaminated samples were amplified and claimed to yield “mummy mtDNA.”
It is unfortunate that clinical studies in oncogenetics do not routinely report comprehensive sequencing results. This has two important consequences (see also [6]): first, the phylogenetic interpretation of the spectra of the mtDNA variants found is limited, and second, the phylogenetic proofreading of sequence data cannot be carried out properly by referees and readers.
In short, we advise authors and editors of scientific journals that (i) special care must be taken for sequencing and documentation since conclusions fully depend on the sequencing data; (ii) raw sequence data must be made fully accessible to referees and readers in order to allow a critical evaluation of the results [11,12] and proofreading during the reviewing process. Although this sounds routine, it is striking to observe that in the clinical literature related to tumorigenesis, we have not seen cases—with a very few exceptions (for example, [6])—that provide the complete primary sequencing results; (iii) referring to the mutation lists in MITOMAP is not sufficient; in addition, the complete record of the data from the population genetics field should be consulted as well.
The use of phylogenetic tools is highly recommended for the medical field, not only for the purpose of data analysis but also in the design of appropriate mtDNA studies [52]. In this way, the distinction between neutral polymorphisms in human populations and the mutations associated with the tumor process [6] or with other human disorders [53,54] stands a chance of being realized.
Supporting Information
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession number for Homo sapiens isolate T1–12 mitochondrion complete genome is AY495278.1.
Patient Summary
Background
DNA carries the genetic “blueprint” of all living things, and everyone has slightly different DNA from everyone else. DNA is found in the cell nucleus but also in other parts (organelles) of the cell called mitochondria. Techniques have been developed to analyze the differences between the mitochondrial DNA (mtDNA) of different people. Several scientific and medical uses are now being made of these tests, and the results have been stored on international databases.
Why Was This Study Done?
Many scientists believe that faults (mutations) in mtDNA play a part in the development of cancer and that tests that look for faulty mtDNA are a way of diagnosing cancer in its very earliest stages. (This could be important because cancer treatment is more effective if done early.) Other scientists claim that there have been experimental errors in the study of mtDNA and cancer (for example, samples may have been mixed up, or one sample may have been contaminated with cells from another) and that faulty mtDNA is not a sign of cancer. These authors wanted to investigate whether the previously published studies were correct.
What Did the Researchers Do and Find?
The authors of this paper point out that, if the mtDNA from a cancer tumor and the mtDNA of normal tissue from the same patient are shown by analysis to differ not just a little, but considerably, then this cannot be the result of a fault appearing in the tumor mtDNA. Instead it must mean that a mistake was made in the laboratory by mixing up DNA from different human sources. The researchers looked at international databases of mtDNA and the mtDNA phylogeny (evolutionary reconstruction of the human mtDNA lineages), and their conclusion was that most of the differences between tumor mtDNA and normal mtDNA are the result of experimental error. According to the authors, the experimental process routinely followed in these studies favors this kind of error.
What Do These Finings Mean?
The widely held view of scientists about the role of mtDNA in the development of cancer appears to be wrong. We do not know what role, if any, mtDNA plays.
Where Can I Get More Information Online?
This is a very specialized paper with no immediate implications for change in the way in which cancer is treated. Here are some sources of general information on cancer.
National Cancer Institute (US):
http://www.cancer.gov/
Cancer Research UK:
http://www.cancerresearchuk.org/aboutcancer/?version=1/
This work was supported by grants from the Ministerio de Sanidad y Consumo (Fondo de Investigación Sanitaria; Instituto de Salud Carlos III, PI030893; SCO/3425/2002) and Genoma España (CeGen; Centro Nacional de Genotipado). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Citation: Salas A, Yao YG, Macaulay V, Vega A, Carracedo Á, et al. (2005) A critical reassessment of the role of mitochondria in tumorigenesis. PLoS Med 2(11): e296.
Abbreviations
HCChepatocellular carcinoma
HVShypervariable segment
mtDNAmitochondrial DNA
PTCpapillary thyroid carcinoma
rCRSrevised Cambridge reference sequence
==== Refs
References
Chinnery PF Schon EA Mitochondria J Neurol Neurosurg Psychiatry 2003 74 1188 1199 12933917
Polyak K Li Y Zhu H Lengauer C Willson JK Somatic mutations of the mitochondrial genome in human colorectal tumours Nat Genet 1998 20 291 293 9806551
Fliss MS Usadel H Caballero OL Wu L Buta MR Facile detection of mitochondrial DNA mutations in tumors and bodily fluids Science 2000 287 2017 2019 10720328
Chinnery PF Samuels DC Elson J Turnbull DM Accumulation of mitochondrial DNA mutations in ageing, cancer, and mitochondrial disease: Is there a common mechanism? Lancet 2002 360 1323 1325 12414225
Gattermann N Wulfert M Hofhaus G How frequent is mutation in the mitochondrial cytochrome c oxidase gene in patients with myelodysplastic syndromes? Br J Haematol 2002 119 1139 author reply 1139–1140 12472601
Vega A Salas A Gamborino E Sobrido MJ Macaulay V mtDNA mutations in tumors of the central nervous system reflect the neutral evolution of mtDNA in populations Oncogene 2004 23 1314 1320 14647457
Bandelt HJ Lahermo P Richards M Macaulay V Detecting errors in mtDNA data by phylogenetic analysis Int J Legal Med 2001 115 64 69 11724431
Bandelt HJ Quintana-Murci L Salas A Macaulay V The fingerprint of phantom mutations in mitochondrial DNA data Am J Hum Genet 2002 71 1150 1160 12384858
Bandelt HJ Salas A Bravi C Problems in FBI mtDNA database Science 2004 305 1402 1404
Bandelt HJ Salas A Lutz-Bonengel S Artificial recombination in forensic mtDNA population databases Int J Legal Med 2004 118 267 273 15257464
Yao YG Bravi CM Bandelt HJ A call for mtDNA data quality control in forensic science Forensic Sci Int 2004 141 1 6 15066707
Forster P To err is human Ann Hum Genet 2003 67 2 4 12556229
Hagelberg E Goldman N Lio P Whelan S Schiefenhovel W Evidence for mitochondrial DNA recombination in a human population of island Melanesia Proc R Soc Lond B Biol Sci 1999 266 485 492 erratum 267: 1595–1596
Yao YG Macauley V Kivisild T Zhang YP Bandelt HJ To trust or not to trust an idiosyncratic mitochondrial data set Am J Hum Genet 2003 72 1341 1348 authors' reply, 1346–1348 12772699
Ingman M Kaessmann H Pääbo S Gyllensten U Mitochondrial genome variation and the origin of modern humans Nature 2000 408 708 713 11130070
Herrnstadt C Elson JL Fahy E Preston G Turnbull DM Reduced-median-network analysis of complete mitochondrial DNA coding-region sequences from the major African, Asian, and European haplogroups Am J Hum Genet 2002 70 1152 1171 erratum, 71: 448–449 11938495
Herrnstadt C Preston G Howell N Errors, phantoms and otherwise, in human mtDNA sequences Am J Hum Genet 2003 72 1585 1586 12817588
Kong QP Yao YG Sun C Bandelt HJ Zhu CL Phylogeny of East Asian mitochondrial DNA lineages inferred from complete sequences Am J Hum Genet 2003 73 671 676 erratum, 75: 157 12870132
Coble MD Just RS O'Callaghan JE Letmanyi IH Peterson CT Single nucleotide polymorphisms over the entire mtDNA genome that increase the power of forensic testing in Caucasians Int J Legal Med 2004 118 137 146 14760490
Palanichamy MG Sun C Agrawal S Bandelt HJ Kong QP Phylogeny of mitochondrial DNA macrohaplogroup N in India, based on complete sequencing: Implications for the peopling of South Asia Am J Hum Genet 2004 75 966 978 15467980
Achilli A Rengo C Magri C Battaglia V Olivieri A The molecular dissection of mtDNA haplogroup H confirms that the Franco-Cantabrian glacial refuge was a major source for the European gene pool Am J Hum Genet 2004 75 910 918 15382008
Richards MB Macaulay VA Bandelt HJ Sykes BC Phylogeography of mitochondrial DNA in western Europe Ann Hum Genet 1998 62 241 260 9803269
Hibi K Nakayama H Yamazaki T Takase T Taguchi M Detection of mitochondrial DNA alterations in primary tumors and corresponding serum of colorectal cancer patients Int J Cancer 2001 94 429 431 11745425
Hibi K Nakayama H Yamazaki T Takase T Taguchi M Mitochondrial DNA alteration in esophageal cancer Int J Cancer 2001 92 319 321 11291064
Nishikawa M Nishiguchi S Shiomi S Tamori A Koh N Somatic mutation of mitochondrial DNA in cancerous and noncancerous liver tissue in individuals with hepatocellular carcinoma Cancer Res 2001 61 1843 1845 11280735
Andrews RM Kubacka I Chinnery PF Lightowlers RN Turnbull DM Reanalysis and revision of the Cambridge reference sequence for human mitochondrial DNA Nat Genet 1999 23 147 10508508
Anderson S Bankier AT Barrell BG de Bruijn MHL Coulson AR Sequence and organization of the human mitochondrial genome Nature 1981 290 457 465 7219534
Imaizumi K Parsons TJ Yoshino M Holland MM A new database of mitochondrial DNA hypervariable regions I and II sequences from 162 Japanese individuals Int J Legal Med 2002 116 68 73 erratum, 119: 183 12056523
Yeh JJ Lunetta KL van Orsouw NJ Moore FD Mutter GL Somatic mitochondrial DNA (mtDNA) mutations in papillary thyroid carcinomas and differential mtDNA sequence variants in cases with thyroid tumours Oncogene 2000 19 2060 2066 10803467
Petros JA Baumann AK Ruiz-Pesini E Amin MB Sun CQ mtDNA mutations increase tumorigenicity in prostate cancer Proc Natl Acad Sci U S A 2005 102 719 724 15647368
Yao YG Kong QP Wang CY Zhu CL Zhang YP Different matrilineal contributions to genetic structure of ethnic groups in the Silk Road region in China Mol Biol Evol 2004 21 2265 2280 15317881
Khrapko K Coller HA Andre PC Li XC Hanekamp JS Mitochondrial mutational spectra in human cells and tissues Proc Natl Acad Sci U S A 1997 94 13798 13803 9391107
Reddy PL Shetty VT Dutt D York A Dar S Increased incidence of mitochondrial cytochrome c-oxidase gene mutations in patients with myelodysplastic syndromes Br J Haematol 2002 116 564 575 11849212
Rocha H Flores C Campos Y Arenas J Vilarinho L About the “pathological” role of the mtDNA T3308C mutation Am J Hum Genet 1999 65 1457 1459 10521313
Salas A Richards M De la Fé T Lareu MV Sobrino B The making of the African mtDNA landscape Am J Hum Genet 2002 71 1082 1111 12395296
Jerónimo C Nomoto S Caballero OL Usadel H Henrique R Mitochondrial mutations in early stage prostate cancer and bodily fluids Oncogene 2001 20 5195 5198 11526508
Monson KL Miller KWP Wilson MR DiZinno JA Budowle B The mtDNA population database: An integrated software and database resource for forensic comparison. Forensic Sci Comm 4 2002 Available: http://www.fbi.gov/hq/lab/fsc/backissu/april2002/miller1.htm . Accessed 30 August 2005
Kirches E Krause G Warich-Kirches M Weis S Schneider T High frequency of mitochondrial DNA mutations in glioblastoma multiforme identified by direct sequence comparison to blood samples Int J Cancer 2001 93 534 538 11477557
Achilli A Rengo C Battaglia V Pala M Olivieri A Saami and Berbers—An unexpected mitochondrial DNA link Am J Hum Genet 2005 76 883 886 15791543
Wong LJ Lueth M Li XN Lau CC Vogel H Detection of mitochondrial DNA mutations in the tumor and cerebrospinal fluid of medulloblastoma patients Cancer Res 2003 63 3866 3871 12873974
Wong LJ Tan DJ Bai RK Yeh KT Chang J Molecular alterations in mitochondrial DNA of hepatocellular carcinomas: Is there a correlation with clinicopathological profile? J Med Genet 2004 41 e65 15121793
Liu VW Shi HH Cheung AN Chiu PM Leung TW High incidence of somatic mitochondrial DNA mutations in human ovarian carcinomas Cancer Res 2001 61 5998 6001 11507041
Kivisild T Tolk HV Parik J Wang Y Papiha SS The emerging limbs and twigs of the East Asian mtDNA tree Mol Biol Evol 2002 19 1737 1751 12270900
Yao YG Kong QP Bandelt HJ Kivisild T Zhang YP Phylogeographic differentiation of mitochondrial DNA in Han Chinese Am J Hum Genet 2002 70 635 651 11836649
Chen JZ Gokden N Greene GF Mukunyadzi P Kadlubar FF Extensive somatic mitochondrial mutations in primary prostate cancer using laser capture microdissection Cancer Res 2002 62 6470 6474 12438238
Chen JZ Gokden N Greene GF Green B Kadlubar FF Simultaneous generation of multiple mitochondrial DNA mutations in human prostate tumors suggests mitochondrial hyper-mutagenesis Carcinogenesis 2003 24 1481 1487 12869417
Abnet CC Huppi K Carrera A Armistead D McKenney K Control region mutations and the ‘common deletion' are frequent in the mitochondrial DNA of patients with esophageal squamous cell carcinoma BMC Cancer 2004 4 30 15230979
Salas A Lareu MV Carracedo A Heteroplasmy in mtDNA and the weight of evidence in forensic mtDNA analysis: A case report Int J Legal Med 2001 114 186 190 11296893
Tully G Barritt SM Bender K Brignon E Capelli C Results of a collaborative study of the EDNAP group regarding mitochondrial DNA heteroplasmy and segregation in hair shafts Forensic Sci Int 2004 140 1 11 15013160
Hochhauser D Relevance of mitochondrial DNA in cancer Lancet 2000 356 181 182 10963192
Cooper A Poinar HN Ancient DNA: Do it right or not at all Science 2000 289 1139 10970224
Bandelt HJ Achilli A Kong QP Salas A Lutz-Bonengel S Low “penetrance” of phylogenetic knowledge in mitochondrial disease studies Biochem Biophys Res Commun 2005 333 122 130 15958208
Kong QP Yao YG Sun C Zhu CL Zhong L Phylogeographic analysis of mitochondrial DNA haplogroup F2 in China reveals T12338C in the initiation codon of the ND5 gene not to be pathogenic J Hum Genet 2004 49 414 423 15278763
Bandelt HJ Yao YG Kivisild T Mitochondrial genes and schizophrenia Schizophr Res 2005 72 267 269 15560971
|
16187796
|
PMC1240051
|
CC BY
|
2021-01-05 10:39:18
|
no
|
PLoS Med. 2005 Nov 4; 2(11):e296
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020296
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1618779710.1371/journal.pmed.0020313Research ArticleCancer BiologyCell BiologyOncologyCancer: LungOncologyOncogenic Transformation by Inhibitor-Sensitive and -Resistant EGFR Mutants Oncogenic Transformation by Mutant EGFRGreulich Heidi
1
2
3
*Chen Tzu-Hsiu
1
3
Feng Whei
1
3
Jänne Pasi A
1
2
Alvarez James V
1
2
Zappaterra Mauro
1
Bulmer Sara E
1
Frank David A
1
2
Hahn William C
1
2
3
4
Sellers William R
1
2
3
*Meyerson Matthew
1
3
4
*1Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, Massachusetts, United States of America,2Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts, United States of America,3The Broad Institute of Harvard and MIT, Cambridge, Massachusetts, United States of America,4Department of Pathology, Harvard Medical School, Boston, Massachusetts, United States of AmericaRosen Neal Academic EditorMemorial-Sloan Kettering Cancer CenterUnited States of America*To whom correspondence should be addressed. E-mail: [email protected] (HG); E-mail: [email protected] (WRS); E-mail: [email protected] (MM)
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: HG, THC, WF, JVA, MZ, SEB, DAF, WCH, WRS, and MM designed and executed the study. HG, PJ, WCH, WRS, and MM contributed to writing the paper.
11 2005 4 10 2005 2 11 e31310 5 2005 30 7 2005 Copyright: © 2005 Greulich et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
EGFR Mutations and Lung Cancer
Inhibition of EGFR Signaling: All Mutations are not Created Equal
Background
Somatic mutations in the kinase domain of the epidermal growth factor receptor tyrosine kinase gene EGFR are common in lung adenocarcinoma. The presence of mutations correlates with tumor sensitivity to the EGFR inhibitors erlotinib and gefitinib, but the transforming potential of specific mutations and their relationship to drug sensitivity have not been described.
Methods and Findings
Here, we demonstrate that EGFR active site mutants are oncogenic. Mutant EGFR can transform both fibroblasts and lung epithelial cells in the absence of exogenous epidermal growth factor, as evidenced by anchorage-independent growth, focus formation, and tumor formation in immunocompromised mice. Transformation is associated with constitutive autophosphorylation of EGFR, Shc phosphorylation, and STAT pathway activation. Whereas transformation by most EGFR mutants confers on cells sensitivity to erlotinib and gefitinib, transformation by an exon 20 insertion makes cells resistant to these inhibitors but more sensitive to the irreversible inhibitor CL-387,785.
Conclusion
Oncogenic transformation of cells by different EGFR mutants causes differential sensitivity to gefitinib and erlotinib. Treatment of lung cancers harboring EGFR exon 20 insertions may therefore require the development of alternative kinase inhibition strategies.
Different EGFR mutations are associated with lung cancer. All of the classes can transform fibroblasts and lung epithelial cells, most are sensitive to erlotinib and gefininib, but exon 20 mutations are only sensitive to an irreversible EGFR inhibitor.
==== Body
Introduction
The human epidermal growth factor receptor gene product (EGFR), a member of the ErbB family of receptor tyrosine kinases, is an integral component of signaling in epithelial cell proliferation. Stimulation of the receptor with EGF or other cognate ligands induces receptor dimerization and autophosphorylation, providing docking sites for SH2-containing adaptor proteins that mediate the activation of intracellular signaling pathways [1–3].
Consistent with a role in proliferative signaling, the oncogenic potential of EGFR variants with deletions in the extracellular domain, including the v-erbB oncogene of avian erythroblastosis virus and the vIII mutant found in human cancers, transforms vertebrate cells in the absence of exogenous EGF [4–7]. In contrast, overexpression of the wild-type EGFR gene can transform NIH-3T3 cells only upon EGF addition [8]. Kinase activity is required for ligand-independent transformation by both types of EGFR extracellular domain deletion mutant [9,10].
A series of novel EGFR kinase domain mutations observed in human lung adenocarcinomas has recently been described [11–16]. These mutations arise in four exons: substitutions for G719 in the nucleotide-binding loop of exon 18, in-frame deletions within exon 19, in-frame insertions within exon 20, and substitutions for L858 or L861 in the activation loop in exon 21. Tumors from patients with clinical responses to the EGFR inhibitors gefitinib or erlotinib have been shown to contain EGFR deletion mutations or substitution mutations [11,12,13,15], but no exon 20 insertion mutations have been reported in this group of clinical responders. Although exon 20 mutations were not widely reported at first, recently five large-scale studies that sequenced EGFR exons 18 through 21 reported a total of 18 exon 20 insertions out of 350 EGFR mutations identified in 1,108 non-small-cell lung cancers [14–18]. Patients who responded to gefitinib and subsequently relapsed were found to have T790M secondary mutations, also in exon 20 [19,20].
Although gefitinib treatment and small interfering RNA experiments suggest that cells expressing mutant EGFR are dependent on EGFR function for survival [11,21,22], the direct transforming potential of the mutations observed in lung adenocarcinoma has not been described. Here, we assess the ability of these EGFR kinase domain mutations to constitutively activate EGFR signaling and contribute to tumorigenesis in model cell culture systems.
Methods
Cell Culture
NIH-3T3 cells obtained fromATCC (Manassas, Virginia, United States) were maintained in DMEM (Cellgro/Mediatech, Herndon, Virginia, United States) supplemented with 10% calf serum (Gibco/Invitrogen, Carlsbad, California, United States) and penicillin/streptomycin (Gibco/Invitrogen). NCI-H3255 cells were maintained in ACL-4 media as previously described [11]. Unless otherwise noted, cells were placed in media containing 0.5% calf serum 24 h prior to EGF (Biosource, Camarillo, California, United States) stimulation. hTBE cells expressing SV40 small T and large T antigens and the human telomerase catalytic subunit hTERT were maintained in serum-free, defined medium as described [23]. Neutralizing antibodies were added 3 h prior to EGF stimulation at the following concentrations: 12 μg/ml anti-EGF (R&D Systems, Minneapolis, Minnesota, United States; #MAB636), 12 μg/ml anti-TGFα (R&D Systems; #AF-239-NA), and 12 μg/ml anti-EGFR (Upstate, Waltham, Massachusetts, United States; #05–101). Gefitinib and erlotinib were purchased from WuXi Pharmatech (Shanghai, China) and diluted in DMSO to the indicated concentrations. CL-387,785 was purchased from Calbiochem (San Diego, California, United States) and diluted in DMSO to the indicated concentrations.
Expression Constructs
EGFR was amplified from a cDNA template with the PCR primers 5′-GATGATATCATGCGACCCTCCGGGAC-3′ and 5′-ATCGATATCTCATGCTCCAATAAATTC-3′, digested with EcoRV, and inserted into the SnaB1 site of pBabe-Puro. Point mutations, insertions, and deletions were made using the Quick-Change Mutagenesis XL kit (Stratagene, La Jolla, California, United States) with the following oligonucleotide primers: 5′-AAGATCACAGATTTTGGGAGGGCCAAACTGCTGGGTG-3′ and 5′-CACCCAGCAGTTTGGCCCTCCCAAAATCTGTGATCTT-3′ for L858R; 5′-AAGATCAAAGTGCTGAGCTCCGGTGCGTTCG-3′ and 5′-CGAACGCACCGGAGCTCAGCACTTTGATCTT-3′ for G719S; 5′-GGTGCACCGCGCCCTGGCAGCCA-3′ and 5′-TGGCTGCCAGGGCGCGGTGCACC-3′ for D837A; 5′-GTCGCTATCAAGGAACCAACATCTCCGAAA-3′ and 5′-TTTCGGAGATGTTGGTTCCTTGATAGCGAC-3′ for L747_E749del, A750P; 5′-GGCCAGCGTGGACAACCCCGGCAACCCCCACGT-3′ and 5′-ACGTGGGGGTTGCCGGGGTTGTCCACGCTGGCC-3′ for D770_N771insNPG. All constructs were fully sequenced.
Transfection and Infection
Replication incompetent retroviruses were produced from pBabe-Puro-based vectors either by cotransfection of 293T cells with pCL-Ampho (Imgenex, San Diego, California, United States) or by transfection into the Phoenix 293T packaging cell line (Orbigen, San Diego, California, United States) using Lipofectamine 2000 (Invitrogen). Cells were infected with these retroviruses in the presence of polybrene. Two days after infection, puromycin (2 μg/ml for NIH-3T3s or 0.5 μg/ml for hTBE cells; Sigma, St. Louis, Missouri, United States) was added and pooled stable cell lines were selected, from which clonal cell lines were derived.
Soft Agar Anchorage-Independent Growth Assay
EGFR-expressing NIH-3T3 cells were suspended in a top layer of DMEM containing 10% calf serum and 0.4% Select agar (Gibco/Invitrogen) and plated on a bottom layer of DMEM containing 10% calf serum and 0.5% Select agar. EGF, gefitinib, or erlotinib was added as described to the top agar. NIH-3T3 colonies were counted in triplicate wells from ten fields photographed with a 10× objective. Growth of hTBE cells in soft agar was determined by plating 1 × 105 cells in triplicate in 0.4% Noble agar. Colonies of hTBE cells were counted microscopically 6–8 wk after plating with the MultiImage imaging counter (Alpha Innotech, San Leandro, California, United States).
Focus Formation Assay
NIH-3T3 cells infected with EGFR retrovirus diluted 1:1, 1:10, or 1:100 were split 2 d after infection into 10-cm plates with and without puromycin selection. Cells for the focus assay were maintained without passage as a monolayer in the absence of puromycin for 3 wk, after which foci were stained with crystal violet (Accustain; Sigma) and scored. Numbers of foci were multiplied by the viral dilution factor and normalized to the relative number of infectious units in each viral stock, as determined by a WST assay (Roche, Basel, Switzerland) on cells after 3 d of puromycin selection.
Nude Mouse Injection
In this assay, 2 × 106 cells were injected subcutaneously into immunocompromised mice, three injections per animal, as described [24]. Tumors were counted and tumor diameter was measured after 5 wk. Standard error of the mean is indicated.
Immunoblotting
Cells were lysed in a buffer containing 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 2.5 mM EDTA, 1% Triton X-100, and 0.25% IPEGAL. Protease inhibitors (Roche) and phosphatase inhibitors (Calbiochem) were added prior to use. Samples were normalized for total protein content unless otherwise indicated. Lysates were boiled in sample buffer, separated by SDS-PAGE on 8% or 10% polyacrylamide gels, transferred to nitrocellulose, and probed as described. Antibodies used for immunoblotting were: anti-EGFR (#2232, Cell Signaling Technologies), anti-phospho-EGFR Y1173 (#05–483, Upstate), anti-phospho-EGFR Y1068 (#2234, Cell Signaling Technologies), anti-phospho-EGFR Y845 (Cell Signaling Technologies, Beverly, Massachusetts, United States; #2231), anti-phospho-EGFR Y1045 (Cell Signaling Technologies; #2237), anti-actin (Santa Cruz Biotechnology, Santa Cruz, California, United States; #sc-1615), anti-Shc (Upstate; #06–203), anti-phospho-Shc Y317 (Upstate; #05–668), anti-Stat3 (Cell Signaling Technologies; #9132), and anti-phospho-Stat3 Y705 (Cell Signaling Technologies; #9131), anti-phospho-Akt S473 (Cell Signaling Technologies; #9271), and anti-Akt (Cell Signaling Technologies; #9272).
Immunoprecipitation
Cells were lysed as described above. Anti-EGFR conjugated beads (Santa Cruz Biotechnology; #sc-120AC) were used for immunoprecipitation. Beads (25 μl) were incubated with fresh lysate in 300 μl of lysis buffer for 1 h, washed twice with lysis buffer, eluted in 1% SDS at room temperature for 20 min, and boiled in sample buffer.
Luciferase Assay
The STAT3 (signal transducer and activator of transcription 3) reporter m67 pTATA TK-luc [25] was kindly provided by J. Bromberg, and the Renilla luciferase reporter phRL tk-luc was purchased from Promega (Madison, Wisconsin, United States). NIH-3T3 cells, infected with EGFR or control retroviruses, were plated on 24-well plates and transfected with 2 μg of m67-luc and 0.2 μg phRL tk-luc with Lipofectamine 2000 (Invitrogen). After 48 h, cells were lysed and luminescence was measured using the dual-luciferase reagents from Promega, according to the manufacturer's instructions, using a Luminoskan Ascent luminometer (ThermoLab Systems, Helsinki, Finland). STAT3-dependent luciferase production was normalized to chemiluminescence values from the control Renilla luciferase.
Results
Expression of Mutant EGFR Induces Oncogenic Transformation
To assess the oncogenic potential of EGFR kinase domain mutants, tumor-derived mutations were introduced into the wild-type human EGFR cDNA by site-directed mutagenesis. The resulting wild-type and mutant EGFR cDNAs were then cloned into the pBabe-Puro retroviral vector and transferred into NIH-3T3 cells by retroviral infection. We initially examined two representative substitution mutations: G719S, observed in exon 18, and L858R, observed in exon 21 (Figure 1). The L858R and G719S mutants were able to transform NIH-3T3 cells to anchorage independence in the absence of exogenous EGF, as assayed by colony formation in soft agar (Figure 1A, top photographs). In contrast, as previously described [8], wild-type EGFR transformed only upon EGF addition (Figure 1A, bottom photographs). The kinase-dead D837A mutant [9], included as a negative control, failed to induce colony formation in the presence or absence of EGF. EGFR expression levels were approximately equal for each pooled stably-transfected cell population (Figure 1B). Clonal cell lines derived from the pooled stably-transfected cells expressing the mutant EGFR exhibited profound morphological alterations characterized by a fusiform, refractile phenotype (unpublished data). Levels of L858R EGFR expression necessary to achieve transformation in this model cell culture system were no higher than expression levels observed in the human lung adenocarcinoma cell line H3255 bearing the L858R mutation (Figure 1C).
Figure 1 Mammalian Cells Expressing the Lung Cancer-Derived Mutant EGFR Grow in an Anchorage-Independent Manner
(A) NIH-3T3 cells infected with retroviruses encoding the indicated wild-type or mutant EGFR were selected in the presence of 2.5 μg /ml puromycin for 4 d. In the top photomicrographs, 1 × 105 cells were suspended in soft agar for a colony formation assay and photographed after 3 wk incubation at 37 °C. Expression of lung cancer-derived missense EGFR mutants, but not wild-type or kinase-inactive D837A EGFR, induces colony formation in soft agar. In the bottom photomicrographs, samples were identical, but 20 ng/ml EGF was added to the top agar. Representative photomicrographs are shown.
(B) Anti-EGFR immunoblot analysis of pooled stable NIH-3T3 cells infected as described in (A). All EGFR constructs are expressed at similar levels. pBp, pBabe-Puro vector; wt, wild-type EGFR.
(C) Lysates from 4 × 104 cells from the human lung adenocarcinoma cell line H3255, harboring the L858R mutation in EGFR, or the wild-type or L858R EGFR-overexpressing NIH-3T3 cells, were immunoblotted for total EGFR levels. Although total protein levels per cell are lower for the H3255 than the NIH-3T3 cells, EGFR expression levels are slightly higher in the H3255s.
(D) NIH-3T3 cells infected with retroviruses encoding the mutant EGFR were selected in the presence of 2 μg/ml puromycin for 9 d. Selected cells (1 × 105) were suspended in soft agar for a colony formation assay and photographed after 3 wk incubation at 37 °C. Expression of the deletion and insertion EGFR mutants induced formation of colonies in soft agar with higher efficiency than expression of L858R. Representative photos are shown. Polyoma mT, NIH-3T3 cells infected with positive control pBabe-Puro retrovirus encoding the polyoma middle T antigen.
(E) hTBE cells expressing the SV40 early region and hTERT were infected with control virus pBabe-Puro (pBp) or with viruses encoding the indicated EGFR alleles. Cells were plated in 0.4% Noble agar, and colonies were counted with an automated imager at 6 wk. Mean ± standard deviation is shown for three independent determinations. Control cells (pBp) formed many microscopic colonies, but colonies formed by cells expressing EGFR mutants were more numerous and larger. del, L747_E749del A750P mutated EGFR; ins, D770_N771insNPG mutated EGFR; pBp, pBabe-Puro vector; RasV12, V12 H-Ras; wt, wild-type EGFR.
(F) Anti-EGFR immunoblot analysis of hTBE cells infected as described in (E). All EGFR constructs were expressed at similar levels. pBp, pBabe-Puro vector; wt, wild-type EGFR.
Transformation of NIH-3T3 cells by L858R or G719S EGFR was further assessed in two independent assays. Expression of the EGFR point mutants in NIH-3T3 cells caused loss of contact inhibition, resulting in focus formation on an unselected monolayer, whereas the wild-type and D837A kinase-inactive controls did not (Table 1). In addition, injection of clonal, transformed NIH-3T3 fibroblasts expressing L858R and G719S EGFR into immunocompromised mice led to the formation of tumors (Table 2). No tumor formation was observed upon injection of cells infected with vector, wild-type, and kinase-dead controls.
Table 1 NIH-3T3 Cells Expressing the Lung Cancer-Derived Mutant EGFR Display Loss of Contact Inhibition
Table 2 Clonal NIH-3T3 Cell Lines Expressing the Lung Cancer-Derived Mutant EGFR Form Tumors in Immunocompromised Mice
Representative exon 19 deletion and exon 20 insertion mutations of EGFR were then assessed for transforming activity. Mutants L747_E749del, A750P [11] and D770_N771insNPG (K. Naoki and M. M., unpublished data) were introduced into NIH-3T3 cells by retrovirus-mediated gene transfer as described above. Cells expressing the EGFR deletion and insertion mutants formed colonies in soft agar with a higher efficiency than that of cells expressing the missense mutants, comparable to the colony formation efficiency of cells expressing polyoma middle T antigen (Figure 1D). Expression of the deletion mutant was comparable to that of L858R EGFR, whereas expression of the insertion mutant was lower, as reflected in the EGFR expression levels of the clonal cell lines (Figure 2A and unpublished data). Cells expressing the L747_E749del A750P and D770_N771insNPG EGFR mutants also exhibited a greater degree of loss of contact inhibition than was observed in cells expressing the L858R or G719S EGFR mutants in a primary focus formation assay (unpublished data).
Figure 2 Ligand-Independent Activation of the Mutant EGFR
(A) Cells expressing the wild-type or mutant EGFR were lysed and immunoblotted with antibodies to total EGFR or antibodies that recognize specific phosphorylation sites in the EGFR C-terminal tail as labeled. All four mutant EGFR proteins, representative of the four classes of EGFR mutations observed in lung adenocarcinoma tumor DNA, exhibited constitutive phosphorylation on the indicated C-terminal autophosphorylation sites. Note that the nomenclature for the anti-phospho-EGFR antibodies reflects elimination of the 24-amino acid signal peptide. Due to difficulties in isolating clonal cell lines with the same levels of mutant EGFR expression, G719S is expressed at higher levels and D770_N771ins NPG at lower levels than the other mutant EGFR. del, L747_E749del A750P; ins, D770_N771insNPG; pBp, pBabe-Puro vector control; wt, wild-type EGFR.
(B) Cells expressing the wild-type or L858R EGFR were placed in media containing 0.5% CS for 24 h. A combination of three neutralizing antibodies (anti-EGF, anti-TGFα, and anti-EGFR) was added 3 h prior to EGF stimulation and lysis. Upper row of blots show the anti-phospho-EGFR Y1068 immunoblots. The lower row shows anti-EGFR immunoblots. No inhibition of L858R EGFR autophosphorylation was observed upon treatment with a combination of three neutralizing antibodies (“neutr Ab”) sufficient to prevent EGF stimulation of autophosphorylation of the wild-type EGFR.
To determine the ability of mutant EGFR to transform a more physiologically relevant cell type, retroviruses expressing the L858R and G719S mutant forms of EGFR were introduced into hTBE cells expressing the SV40 early region T antigens and the human telomerase catalytic subunit hTERT [23]. We previously showed that such cells are fully transformed by the additional expression of oncogenic alleles of H- or K-RAS [23]. Similarly, the expression of L858R and G719S mutant EGFR genes conferred enhanced anchorage-independent growth upon such hTBE cells, with colony numbers approximately 15-fold above the background level of microscopic colonies observed in hTBE cells expressing wild-type EGFR or a control vector (Figure 1E). The representative deletion and insertion mutants, L747_E749del A750P and D770_N771insNPG, formed colonies in soft agar with even greater efficiency, with the caveat that the deletion mutant is expressed at higher levels than the other mutants in this assay (Figure 1F). Similar to hTBE cells expressing H-RAS V12, expression of these EGFR mutants did not increase the rate of cell proliferation in defined medium (unpublished data).
Multiple tumor-derived mutants of EGFR therefore contribute to oncogenic transformation as shown by three different assays: anchorage-independent cell growth, focus formation, and in vivo tumor formation.
Mutant EGFR Proteins Are Constitutively Active
To determine whether transformation by mutant EGFR is associated with constitutive receptor activation in the absence of exogenous EGF, tyrosine autophosphorylation in the C-terminal tail of EGFR was examined by immunoblotting of cell lysates. Constitutive tyrosine phosphorylation of the mutant EGFR molecules was observed at several C-terminal sites, including Y845, Y1068, and Y1173 (Figure 2A). High-level phosphorylation of the insertion mutant at Y1045, the docking site for the Cbl E3 ubiquitin ligase [26], is correlated with decreased abundance of this protein (Figure 2A), but whether the differential protein levels are a result of Cbl activity has not been confirmed.
The constitutive increase in EGFR activity appears to be ligand-independent, as a combination of neutralizing antibodies against EGF, TGFα, and EGFR failed to inhibit elevated basal levels of L858R autophosphorylation (Figure 2B). However, tyrosine phosphorylation on the EGFR mutants could be further increased by EGF stimulation (Figure 2B), suggesting that the mutant EGFRs exhibit both ligand-independent and ligand-dependent activation, similar to that observed upon EGF stimulation of the L858R mutant H3255 lung adenocarcinoma cell line [21]. Ligand-independent activation of EGFR with lung cancer-derived kinase domain mutations has not been observed by other groups working with transient transfection systems [22,27]. We too have failed to detect constitutive elevation of mutant receptor autophosphorylation when transiently expressed in NIH-3T3 and HeLa cells (unpublished data). The reason for this phenotypic difference remains unclear.
Expression of Mutant EGFR Results in Activation of Shc, STAT3, and Akt
We next asked whether constitutive activation of mutant EGFR is associated with alterations in downstream signaling pathways. Because Y1173, a docking site for the adaptor protein Shc [28], is constitutively phosphorylated on mutant EGFR (Figure 2A), we analyzed Shc-EGFR complex formation in cells expressing wild-type and mutant EGFR. Coimmunoprecipitation studies revealed a low level of constitutive Shc binding to the L858R EGFR, further augmented by EGF stimulation (Figure 3A), whereas Shc complexed with the wild-type EGFR only upon EGF stimulation. Immunoblotting of whole cell lysates with an antibody specific for tyrosine-phosphorylated Shc revealed constitutive phosphorylation on Shc in cells expressing the L858R EGFR, consistent with the known phosphorylation of EGFR-bound Shc [29]; in contrast, in cells expressing wild-type EGFR, Shc was phosphorylated only in response to EGF stimulation (Figure 3B). Similar to the situation with receptor autophosphorylation, constitutive phosphorylation of Shc in mutant EGFR-expressing cells has not been observed in transient expression systems [27].
Figure 3 Shc and STAT3 Signaling Pathways Are Constitutively Activated in Cells Expressing the Mutant EGFR
(A) Cells expressing the wild-type or L858R mutant EGFR were placed in 0.5% calf serum for 24 h and left unstimulated or stimulated with 20 ng/ml EGF (“+EGF”) for 8 min. EGFR was immunoprecipitated from 200 μg of cell lysate, and eluted immune complexes were separated by SDS-PAGE and immunoblotted with anti-Shc. Shc constitutively coimmunoprecipitates with the L858R EGFR but not the wild-type EGFR. IP, immunoprecipitation; NL, no-lysate control immunoprecipitation.
(B) Anti-phospho-Shc immunoblots (upper row of blots) and anti-Shc immunoblots (lower row) of whole cell lysates from the experiment in (A). All three Shc isoforms are constitutively phosphorylated in cells expressing the L858R EGFR.
(C) Immunoblots of whole cell lysates with anti-phospho-STAT3 Y705 (upper row), total STAT3 (middle row), or actin as a loading control (lower row). STAT3 is constitutively phosphorylated in cells expressing any of the four mutant EGFR proteins, representative of the four classes of EGFR mutations observed in lung adenocarcinoma tumor DNA. del, L747_E749del A750P; ins, D770_N771insNPG; pBp, pBabe-Puro vector control; wt, wild-type EGFR.
(D) M67 STAT3 luciferase reporter assay. NIH-3T3 cells expressing the indicated EGFR were transfected with a STAT-dependent reporter (m67-firefly luciferase) [25] and a control reporter expressing Renilla luciferase. STAT-dependent luciferase production was measured after 48 h and normalized to Renilla luciferase. The normalized luciferase values were divided by the values for cells expressing the wild-type EGFR to produce relative luciferase units. NIH-3T3 expressing mutant forms of EGFR exhibited elevated levels of STAT-dependent transcriptional activity relative to wild-type. ins, D770_N771insNPG EGFR; wt, wild-type EGFR.
(E) Immunoblots of whole cell lysates with anti-phospho-Akt S473 (upper row), total Akt (middle row), or actin as a loading control (lower row). Akt is constitutively phosphorylated in cells expressing mutant EGFR. del, L747_E749del A750P; ins, D770_N771insNPG; wt, wild-type EGFR.
Consistent with previous reports on L858R mutant EGFR [22], STAT signaling pathways are constitutively activated in the transformed NIH-3T3 cells. Immunoblotting with antibodies specific for phosphorylated Y705, the tyrosine responsible for STAT3 dimerization [30], revealed constitutive phosphorylation in cells expressing the lung cancer-derived mutant EGFR but not wild-type EGFR (Figure 3C). Increased STAT3-dependent gene expression in cells expressing the mutant EGFRs was confirmed in a reporter assay (Figure 3D) using a STAT-dependent luciferase construct [25].
Constitutive phosphorylation of mutant EGFR on Y1068 (see Figure 2A), the binding site for the phosphatidylinositol 3-kinase interacting protein Gab1 [31], indicated that signaling pathways downstream of phosphatidylinositol 3-kinase might be constitutively activated as well. One such pathway is controlled by the serine/threonine kinase Akt, which is involved in promotion of cell survival. Western blotting with anti-phospho-Akt confirmed that Akt is constitutively activated in cells expressing the mutant EGFR (Figure 3E). We therefore conclude that at least a subset of physiological EGFR signaling pathways is activated by stable expression of mutant EGFR.
Transformation by the Exon 20 Insertion Mutant Is Not Sensitive to Gefitinib or Erlotinib
Given the association between the presence of activating EGFR mutations and clinical responses to gefitinib or erlotinib in lung adenocarcinoma patients [11,12,13,15], we assessed the ability of these EGFR inhibitors to inhibit anchorage-independent growth of clonal NIH-3T3 cell lines expressing wild-type or mutant EGFR. Consistent with the increased sensitivity to gefitinib and erlotinib of patient tumors harboring the missense mutations or exon 19 deletions, anchorage-independent growth of cells expressing L858R, G719S, or L747_E749del A750P was inhibited by 100 nM erlotinib (Figure 4A and 4B) or gefitinib (Figure 4B and unpublished data), although the G719S mutant may be somewhat more resistant to gefitinib (Figure 4A and unpublished data), consistent with other in vitro studies [32]. In contrast, 1 μM erlotinib (Figure 4A) or gefitinib (unpublished data) did not inhibit anchorage-independent growth of EGF-treated cells overexpressing the wild-type EGFR.
Figure 4 Sensitivity of Cell Transformation Induced by Expression of Mutant EGFR Characterized by Missense Mutation or Exon 19 Deletion, but not Exon 20 Insertion, to Gefitinib and Erlotinib
(A) Anchorage-independent growth of clonal NIH-3T3 cells transformed with mutant EGFR or EGF-stimulated wild-type EGFR treated with the indicated concentrations of erlotinib immediately prior to suspension in soft agar. Transformation induced by expression of L858R, G719S, and L747_E749del A750P EGFR, but not EGF-stimulated wild-type EGFR or D770_N771insNPG EGFR, was inhibited by 0.1 μM erlotinib. Representative photographs are shown.
(B) Number of colonies formed in soft agar by clonal NIH-3T3 cells expressing L858R EGFR and D770_N771insNPG EGFR treated with the indicated concentrations of gefitinib or erlotinib immediately prior to suspension in soft agar. Transformation by cells expressing the L858R EGFR was inhibited by 0.1 μM gefitinib or erlotinib, whereas transformation by cells expressing the insertion mutant was resistant to low concentrations of these inhibitors. Colonies were quantitated by counting ten fields each of triplicate wells photographed with a 10× objective; mean ± standard deviation is shown. Ins, D770_N771insNPG EGFR.
(C) Transformation induced by expression of D770_N771insNPG EGFR is inhibited 10-fold more efficiently by the irreversible EGFR inhibitor CL-387,785 [35]. Clonal NIH-3T3 cells expressing the insertion mutant were treated with the indicated concentrations of gefitinib, erlotinib, or CL-387,785 immediately prior to suspension in soft agar. This assay was not done in triplicate, but the results are representative of two independent experiments. The number of colonies was normalized to maximum colony formation for each treatment, and sigmoidal dose response curves were fitted to the data using Prism Graphpad software to determine IC50s.
Transformation by the D770_N771insNPG EGFR mutant was remarkably insensitive to gefitinib and erlotinib, as inhibition of colony growth in soft agar required exposure to 100-fold higher concentrations (>1 μM) of these agents than was required to inhibit colony formation by cells expressing the EGFR missense mutants or deletion mutant (Figure 4A). In fact, no significant inhibition of anchorage-independent growth of cells expressing D770_N771insNPG EGFR was observed at 3 μM gefitinib or erlotinib (Figure 4B). The concentrations of gefitinib or erlotinib necessary to reverse insertion mutant transformation are therefore higher than the achievable serum concentrations of gefitinib (0.5–1.0 μM) and possibly higher than the achievable serum concentrations of erlotinib (2.8–4.0 μM) [33,34].
Consistent with this result, all three lung adenocarcinoma patients with known exon 20 insertion mutants of EGFR have failed to show a clinical response to treatment and have instead achieved only stable disease with erlotinib alone (n = 1; L. Sequist and T. Lynch, personal communication), or in combination with chemotherapy (n = 2; D. Eberhard and K. Hillan, personal communication). These results suggest that cancers harboring distinct activating kinase domain mutations of EGFR may exhibit a differential sensitivity to specific EGFR inhibitors.
Interestingly, the irreversible EGFR inhibitor CL-387,785 [35] is more effective than gefitinib or erlotinib for inhibition of colony formation by cells expressing the exon 20 insertion mutant (Figure 4C). Calculated IC50 values for gefitinib, erlotinib, and CL-387,785 against D770_N771insNPG were 2.6 μM, 2.5 μM, and 0.2 μM, respectively. CL-387,785 had an even greater effect on colony formation by cells expressing L858R EGFR, completely inhibiting transformation at 0.003 μM (unpublished data). These effects are also observed upon assessment of receptor autophosphorylation (Figure 5). Although the inhibitory concentrations do not exactly correlate with the results of the colony formation assay, probably due to the difference in duration of the assays, the trends are the same. Insertion mutant autophosphorylation is less sensitive to inhibition by gefitinib than that of L858R, but CL-387,785 is more effective than gefitinib at inhibiting insertion mutant (and L858R) autophosphorylation.
Figure 5 Sensitivity of Mutant EGFR Autophosphorylation to EGFR Inhibitors Reflects Inhibition of Anchorage-Independent Growth
Cells expressing wild-type, L858R, or D770_N771insNPG EGFR were treated for 2 h with the indicated concentrations of gefitinib or CL-387,785. Cells expressing the wild-type EGFR were then stimulated for 10 min with 7 ng/ml EGF, and all plates were lysed. Whole-cell lysates were immunoblotted for phospho-EGFR Y1068 (upper row of blots), total EGFR (middle row), and actin as a loading control (lower row). Although compound concentrations necessary for inhibition of autophosphorylation do not exactly correspond to inhibition of anchorage-independent growth, the relative sensitivity of autophosphorylation of the wild-type and mutant EGFR to gefitinib or CL-387,785 mirrors the relative sensitivity of colony formation to these inhibitors.
Discussion
Treatment with the EGFR inhibitors gefitinib and erlotinib has led to dramatic responses in many lung cancer patients, predominantly for those cancers in which EGFR mutations can be detected. However, there has been a subset of lung cancer patients with these mutations who do not respond to the EGFR inhibitors in current clinical use.
By demonstrating that lung-cancer derived kinase domain mutants of EGFR are constitutively activated and that they can transform cultured mammalian cells, we have provided an in vitro system with which to study EGFR-dependent oncogenesis in a genetically homogeneous background. Although the anchorage-independent growth assay measures only one of many phenotypes of transformation and does not, for example, recapitulate tumor microenvironment or account for the influence of the immune system on tumor formation, this system will be useful for dissecting inhibitor response and downstream signaling pathways, particularly for those mutants not found in existing cancer-derived cell lines.
Using the NIH-3T3 transformation system, we have found that transformation by an exon 20 insertion mutant is resistant to inhibition by gefitinib and erlotinib. Strikingly, transformation by this EGFR exon 20 insertion mutant is more sensitive to treatment with an irreversible inhibitor, CL-387,785. This compound was previously found to be active against EGFR containing the exon 20 point mutation T790M, associated with resistance to gefitinib and erlotinib [19].
These results indicate a need for the use of novel EGFR inhibitors in primary treatment of lung cancers harboring the exon 20 insertion mutations. Furthermore, the distinct inhibitor sensitivity of various EGFR mutants argues that therapies may need to be targeted against specific mutant forms of a protein, whereas generalized inhibition of a particular oncogenic target may not be sufficient.
Patient Summary
Background
While lung cancer is still one of the deadliest cancers, a new class of drugs called epidermal growth factor receptor (EGFR) inhibitors have shown promising results in some patients. As the name suggests, these drugs inhibit a protein called EGFR, which is altered in a subset of lung cancers.
What Did the Researchers Do and Find?
The aim of this study was to understand in more detail what role the different EGFR alterations played in the tumor and which ones made the tumor responsive to EGFR inhibitor treatment. First, the researchers introduced different altered EGFR versions into human cells and studied their behavior. The EGFR protein, which can stimulate cell growth, is normally tightly regulated and active only when a cell receives a signal from its neighbors. However, the alterations made the EGFR protein always active, regardless of a stimulating signal, and caused them to “grow out of control.” They then treated the cells expressing the various alterations with different EGFR inhibitor drugs and showed that the specific alteration determined whether cell growth could be stopped by a specific drug.
What Do the Results Mean for Patients?
EGFR inhibitors are still considered to be experimental treatment. Researchers are making progress in understanding how genetic alterations in EGFR cause abnormal growth in some lung cancers and also which specific alterations cause the tumor to be responsive to a particular drug. The goal is to match the tumor and the drug to maximize anti-tumor response and avoid giving a drug that doesn't work with a particular tumor.
Where Can I Get More Information?
The following Web sites contain information on lung cancer, the role of EGFR mutations, and the EGFR class of drugs.
Lung Cancer Online (follow links for experimental treatments and EGFR inhibitors):
http://www.lungcanceronline.org/index.htm
US National Cancer Institute information page on lung cancer:
http://www.nci.nih.gov/cancertopics/types/lung
Ongoing clinical trials of EGFR inhibitors:
http://www.clinicaltrials.gov
We thank Tom Roberts, Ben Neel, Guillermo Paez, Jeff Lee, Forest White, Yi Zhang, Orit Rosen, Sue-Ann Woo, Kwok Wong, Hongbin Ji, Suzanne Gaudet, Paul Jasper, Birgit Schoeberl, Jane Jiang, and David Hill for helpful discussions. MZ is supported by the National Institutes of Health-sponsored MD-PhD program. WCH is supported by the Doris Duke Charitable Foundation and the National Cancer Institute. WRS and MM are supported by the Novartis Research Foundation, the Claudia Adams Barr Foundation, and the Charles A. Dana Human Cancer Genetics Program of the Dana-Farber Cancer Institute, the Poduska Family Foundation, the Damon-Runyon Cancer Research Foundation, Joan's Legacy, the American Cancer Society, the Flight Attendant Medical Research Institute, and the National Cancer Institute. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Citation: Greulich H, Chen TH, Feng W, Jänne PA, Alvarez JV, et al. (2005) Oncogenic transformation by inhibitor-sensitive and -resistant EGFR mutants. PLoS Med 2(11): e313.
Abbreviations
EGFRepidermal growth factor receptor
hTBEhuman tracheobronchial epithelial
STATsignal transducer and activator of transcription
==== Refs
References
Olayioye MA Neve RM Lane HA Hynes NE The ErbB signaling network: Receptor heterodimerization in development and cancer EMBO J 2000 19 3159 3167 10880430
Yarden Y The EGFR family and its ligands in human cancer. Signalling mechanisms and therapeutic opportunities Eur J Cancer 2001 37 Suppl 4 S3 8
Jorissen RN Walker F Pouliot N Garrett TP Ward CW Epidermal growth factor receptor: Mechanisms of activation and signalling Exp Cell Res 2003 284 31 53 12648464
Gamett DC Tracy SE Robinson HL Differences in sequences encoding the carboxyl-terminal domain of the epidermal growth factor receptor correlate with differences in the disease potential of viral erbB genes Proc Natl Acad Sci U S A 1986 83 6053 6057 3016739
Boerner JL Danielsen A Maihle NJ Ligand-independent oncogenic signaling by the epidermal growth factor receptor: v-ErbB as a paradigm Exp Cell Res 2003 284 111 121 12648470
Garcia de Palazzo IE Adams GP Sundareshan P Wong AJ Testa JR Expression of mutated epidermal growth factor receptor by non-small cell lung carcinomas Cancer Res 1993 53 3217 3220 8391918
Moscatello DK Holgado-Madruga M Godwin AK Ramirez G Gunn G Frequent expression of a mutant epidermal growth factor receptor in multiple human tumors Cancer Res 1995 55 5536 5539 7585629
Velu TJ Beguinot L Vass WC Willingham MC Merlino GT Epidermal-growth-factor-dependent transformation by a human EGF receptor proto-oncogene Science 1987 238 1408 1410 3500513
Danielsen AJ Maihle NJ Ligand-independent oncogenic transformation by the EGF receptor requires kinase domain catalytic activity Exp Cell Res 2002 275 9 16 11925101
Huang HS Nagane M Klingbeil CK Lin H Nishikawa R The enhanced tumorigenic activity of a mutant epidermal growth factor receptor common in human cancers is mediated by threshold levels of constitutive tyrosine phosphorylation and unattenuated signaling J Biol Chem 1997 272 2927 2935 9006938
Paez JG Janne PA Lee JC Tracy S Greulich H EGFR mutations in lung cancer: Correlation with clinical response to gefitinib therapy Science 2004 304 1497 1500 15118125
Lynch TJ Bell DW Sordella R Gurubhagavatula S Okimoto RA Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib N Engl J Med 2004 350 2129 2139 15118073
Pao W Miller V Zakowski M Doherty J Politi K EGF receptor gene mutations are common in lung cancers from “never smokers” and are associated with sensitivity of tumors to gefitinib and erlotinib Proc Natl Acad Sci U S A 2004 101 13306 13311 15329413
Kosaka T Yatabe Y Endoh H Kuwano H Takahashi T Mutations of the epidermal growth factor receptor gene in lung cancer: Biological and clinical implications Cancer Res 2004 64 8919 8923 15604253
Mitsudomi T Kosaka T Endoh H Horio Y Hida T Mutations of the epidermal growth factor receptor gene predict prolonged survival after gefitinib treatment in patients with non-small-cell lung cancer with postoperative recurrence J Clin Oncol 2005 23 2513 2520 15738541
Shigematsu H Lin L Takahashi T Nomura M Suzuki M Clinical and biological features associated with epidermal growth factor receptor gene mutations in lung cancers J Natl Cancer Inst 2005 97 339 346 15741570
Chou TY Chiu CH Li LH Hsiao CY Tzen CY Mutation in the tyrosine kinase domain of epidermal growth factor receptor is a predictive and prognostic factor for gefitinib treatment in patients with non-small cell lung cancer Clin Cancer Res 2005 11 3750 3757 15897572
Huang SF Liu HP Li LH Ku YC Fu YN High frequency of epidermal growth factor receptor mutations with complex patterns in non-small cell lung cancers related to gefitinib responsiveness in Taiwan Clin Cancer Res 2004 10 8195 8203 15623594
Kobayashi S Boggon TJ Dayaram T Janne PA Kocher O EGFR mutation and resistance of non-small-cell lung cancer to gefitinib N Engl J Med 2005 352 786 792 15728811
Pao W Miller VA Politi KA Riely GJ Somwar R Acquired resistance of lung adenocarcinomas to gefitinib or erlotinib is associated with a second mutation in the EGFR kinase domain PLoS Med 2005 2 e73 15737014
Tracy S Mukohara T Hansen M Meyerson M Johnson BE Gefitinib induces apoptosis in the EGFRL858R non-small-cell lung cancer cell line H3255 Cancer Res 2004 64 7241 7244 15492241
Sordella R Bell DW Haber DA Settleman J Gefitinib-sensitizing EGFR mutations in lung cancer activate anti-apoptotic pathways Science 2004 305 1163 1167 15284455
Lundberg AS Randell SH Stewart SA Elenbaas B Hartwell KA Immortalization and transformation of primary human airway epithelial cells by gene transfer Oncogene 2002 21 4577 4586 12085236
Hahn WC Counter CM Lundberg AS Beijersbergen RL Brooks MW Creation of human tumour cells with defined genetic elements Nature 1999 400 464 468 10440377
Besser D Bromberg JF Darnell JE Hanafusa H A single amino acid substitution in the v-Eyk intracellular domain results in activation of Stat3 and enhances cellular transformation Mol Cell Biol 1999 19 1401 1409 9891073
Levkowitz G Waterman H Ettenberg SA Katz M Tsygankov AY Ubiquitin ligase activity and tyrosine phosphorylation underlie suppression of growth factor signaling by c-Cbl/Sli-1 Mol Cell 1999 4 1029 1040 10635327
Amann J Kalyankrishna S Massion PP Ohm JE Girard L Aberrant epidermal growth factor receptor signaling and enhanced sensitivity to EGFR inhibitors in lung cancer Cancer Res 2005 65 226 235 15665299
Okabayashi Y Kido Y Okutani T Sugimoto Y Sakaguchi K Tyrosines 1148 and 1173 of activated human epidermal growth factor receptors are binding sites of Shc in intact cells J Biol Chem 1994 269 18674 18678 8034616
Pelicci G Lanfrancone L Grignani F McGlade J Cavallo F A novel transforming protein (SHC) with an SH2 domain is implicated in mitogenic signal transduction Cell 1992 70 93 104 1623525
Zhong Z Wen Z Darnell JE Stat3: A STAT family member activated by tyrosine phosphorylation in response to epidermal growth factor and interleukin-6 Science 1994 264 95 98 8140422
Rodrigues GA Falasca M Zhang Z Ong SH Schlessinger J A novel positive feedback loop mediated by the docking protein Gab1 and phosphatidylinositol 3-kinase in epidermal growth factor receptor signaling Mol Cell Biol 2000 20 1448 1459 10648629
Jiang J Greulich G Sellers WR Meyerson M Griffin J EGF-independent transformation of Ba/F3 cells with cancer-derived EGFR mutants induces gefitinib-sensitive cell cycle progression Cancer Res 2005 In Press
Cohen MH Williams GA Sridhara R Chen G McGuinn WD United States Food and Drug Administration Drug approval summary: Gefitinib (ZD1839; Iressa) tablets Clin Cancer Res 2004 10 1212 1218 14977817
Hidalgo M Siu LL Nemunaitis J Rizzo J Hammond LA Phase I and pharmacologic study of OSI-774, an epidermal growth factor receptor tyrosine kinase inhibitor, in patients with advanced solid malignancies J Clin Oncol 2001 19 3267 3279 11432895
Discafani CM Carroll ML Floyd MB Hollander IJ Husain Z Irreversible inhibition of epidermal growth factor receptor tyrosine kinase with in vivo activity by N-[4-[(3-bromophenyl)amino]-6-quinazolinyl]-2-butynamide (CL-387,785) Biochem Pharmacol 1999 57 917 925 10086326
|
16187797
|
PMC1240052
|
CC BY
|
2021-01-05 10:39:18
|
no
|
PLoS Med. 2005 Nov 4; 2(11):e313
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020313
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1618773410.1371/journal.pmed.0020327PerspectivesImmunologyInfectious DiseasesOtherAllergy/ImmunologyEpidemiology/Public HealthVaccinesParasitologyInfectious DiseasesMicrobiologyMedicine in Developing CountriesImmunology and allergyInternational healthThe End of the Line for Hookworm? An Update on Vaccine Development PerspectivesDevaney Eileen Eileen Devaney is Professor of Parasite Immunobiology, Parasitology Group, Institute of Comparative Medicine, Faculty of Veterinary Medicine, University of Glasgow, Glasgow, United Kingdom. E-mail: [email protected]
Competing Interests: The author is on the Board of Directors of the Moredun Research Institute (http://www.moredun.org.uk), a grant-aided public body. A group at the institute is working on vaccine development against nematodes of livestock.
10 2005 4 10 2005 2 10 e327Copyright: © 2005 Eileen Devaney.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Vaccination with Recombinant Aspartic Hemoglobinase Reduces Parasite Load and Blood Loss after Hookworm Infection
Gut-associated molecules of human hookworms have been the focus of recent interest as potential vaccine candidates.
==== Body
Human hookworms are parasitic nematodes infecting about 700 million individuals, largely in tropical regions of the world [1]. In endemic areas, most infected people carry a mixed worm burden, including Ascaris lumbricoides (roundworms), Trichuris trichuria (whipworms), and Ancylostoma duodenale and/or Necator americanus (both hookworms). Of these soil-transmitted helminths, hookworms are the most pathogenic because of their propensity to feed on blood, resulting in anaemia, particularly in those with low iron reserves such as children and women of reproductive age.
The Pathogenesis of Hookworm Infection
Hookworms' blood-feeding (hematophagous) habits cause pathology in humans and animals. The worms attach to the wall of the small intestine using their mouthparts and feed on blood from ruptured capillaries. Each female worm is estimated to ingest a minimal 0.1 ml of blood per day. However, actual blood loss can be significantly greater; the worms change their feeding sites several times a day, and the secretion of anti-coagulants means that the vacated sites continue to bleed, contributing greatly to blood loss.
Hookworms do not kill, but they can cause subclinical disease, most notably anaemia and impaired physical and cognitive development in children. As hookworm infection is associated with low socioeconomic status, it adds significantly to the burden of disease in such areas [1].
Natural Infection Elicits Poor Immunity
Hookworms have a simple life cycle in which the third-stage larvae (L3) infect humans, generally by skin penetration, although some species are also infective via oral ingestion. The parasites enter the bloodstream and migrate to the lungs; from there, they are coughed and swallowed to the small intestine (Figure 1). The adult parasites mature in the intestine, and following mating, the female worm produces many thousands of eggs that pass out in the faeces and develop on the ground to infective L3.
Figure 1 Life Cycle of the Human Hookworm N. americanus
(Illustration: Sapna Khandwala, reproduced from [13])
Hookworms can be treated using anthelmintic drugs such as albendazole, but treated people soon become reinfected. Additionally, recent epidemiological studies from China and Brazil show the highest worm burden and the highest prevalence of infection in the elderly [2], contrasting with the intensity/prevalence curves for other soil-transmitted helminths, which typically peak in mid-to-late childhood. These data suggest that under natural conditions of exposure, little immunity is evoked. Given this immuno-epidemiological picture, developing a vaccine is a significant challenge.
The Search for a Vaccine
Recent studies suggest that the hematophagous lifestyle of hookworms may prove their downfall. Hookworms are armed with an array of molecules that are essential for blood feeding and digestion; these include anticoagulants and a variety of proteases that digest haemoglobin (Hb) and other serum proteins. In a new article in PLoS Medicine, Loukas et al. [3] now describe a vaccination schedule using one such protease—an aspartic haemoglobinase—from the hookworm Ancylostoma caninum (Ac-APR-1). In their study, the schedule protected against blood loss in an animal model of hookworm infection.
As with other parasitic nematodes, hookworms are complex multicellular organisms that have evolved an array of mechanisms for suppressing or avoiding host immune responses. The only commercially available vaccine against a parasitic nematode is Huskvac (an oral lungworm vaccine for calves), a preparation of radiation-attenuated L3 of Dictyocaulus viviparus, which protects against parasitic bronchitis [4]. In the 1960s, a similar approach was adopted to controlling hookworm infection in dogs using irradiated L3 of A. caninum [5]. While this vaccine was efficacious, it was a commercial failure. However, these older vaccines established the principle that protective immunity can be elicited. The challenge now is to identify protective antigens and present them to the immune system in an appropriate manner.
Proteases as Potential Vaccine Candidates
Hookworms express a range of proteases, including cysteine, aspartic, and metallo-proteases, several of which have been characterised in detail. These enzymes are localised to the brush border of the worm intestine and have been shown to function in a multi-enzyme cascade to digest Hb and other serum proteins [6]. Some of these molecules show an exquisite specificity; for example, Na-APR-2, an aspartic protease from N. americanus, cleaves Hb from the permissive host (human) with twice the efficiency of Hb from a non-permissive host (dog) [7].
These gut-associated molecules of the parasite have been the focus of recent interest as potential vaccine candidates. The rationale behind this approach is that the induction of antibodies to molecules that function in parasite feeding will neutralise their activity and effectively “starve” the worm. A similar approach has previously been trialled against nematode parasites of livestock such as Haemonchus contortus [8], and some important lessons have been learned. One of these is the requirement for expression of recombinant molecules in a eukaryotic system [9]. In general, bacterial-expressed antigens do not stimulate protection, presumably because they are improperly folded and/or modified and catalytically inactive.
Under the auspices of the Human Hookworm Vaccine Initiative and the Sabin Vaccine Institute (http://www.sabin.org/hookworm.htm), which supported Loukas and colleagues' study [3], the drive to develop a hookworm vaccine has gained significant momentum. A number of candidate antigens have been tested in the A. caninum dog model with varying degrees of success. Ideally, vaccination would protect against infection with L3. The same team of researchers have previously identified one such secreted molecule of the L3 of N. americanus, Na-ASP-2 (for abundant secreted protein-2), and have shown it to partially protect dogs against infection [10]. Phase 1 safety trials are now underway with this antigen.
Other candidate molecules are the cysteine and aspartic haemoglobinases from the adult worm. In another study by Loukas and colleagues, vaccination with a catalytically active cathepsin-B-like protease from A. caninum (Ac-CP-2) produced in the yeast Pischia pastoris resulted in worms that were stunted and produced fewer eggs but did not produce a reduction in number of worms or protect against anaemia [11]. The current study from the same group tested an active aspartic haemoglobinase (Ac-APR-1) in the same model [3]. A modest reduction in worm burden was observed in immunised animals, but a highly significant reduction in worm fecundity was observed (up to 85% reduction in mean egg output between vaccinated and adjuvant-only controls), emphasising the nutritional demand of the parasite for egg laying and presumably reflecting an accumulation of neutralising antibodies. Most importantly, four of five vaccinated animals showed a reduction in Hb loss. Thus Ac-APR-1 could represent a pathology-limiting component of a future multivalent vaccine [3]. In addition, by restricting worm fecundity it would, in essence, also act as a “transmission blocking” vaccine.
The Challenges Ahead
Despite these achievements, significant challenges remain, such as the selection of appropriate adjuvants for use in humans, the production of the vaccine at low cost and high yield, its distribution in the tropics, and the possible requirement to de-worm individuals prior to vaccination. However, the Human Hookworm Vaccine Initiative is progressing on many of these fronts and is an excellent example of what can be achieved with proper funding and good collaborations. It has the potential to become a 21st-century paradigm for a control programme aimed at a neglected tropical disease and should also provide renewed impetus to control programmes aimed at vaccine development against other hematophagous helminths of humans and domestic animals.
Finally, it is somewhat ironic to note that as attempts are made to eradicate worm infection in tropical regions of the world, worms are being used in the developed countries to regulate pathogenic proinflammatory immune responses. For example, patients with ulcerative colitis have shown an encouraging amelioration of pathology following infection with the pig whipworm Trichuris suis [12]. These studies demonstrate the capacity of helminth parasites to induce regulatory immune networks in their hosts, and they emphasise the scale of the challenge facing the development of vaccines against worms.
Citation: Devaney E (2005) The end of the line for hookworm? An update on vaccine development. PLoS Med 2(10): e327.
Abbreviations
Hbhaemoglobin
L3third-stage larvae
==== Refs
References
de Silva NR Brooker S Hotez PJ Montresor A Engels D Soil-transmitted helminth infections: Updating the global picture Trends Parasitol 2003 19 547 551 14642761
Bethony J Chen J Lin S Xiao S Zhan B Emerging patterns of hookworm infection: Influence of aging on the intensity of Necator infection in Hainan province, People's Republic of China Clin Infect Dis 2002 35 1336 1344 12439796
Loukas A Bethony JM Mendez S Fujiwara RT Goud GN Vaccination with recombinant aspartic hamoglobinase reduces parasite load and blood loss after hookworm infection PLoS Med 2005 2 e295 10.1371/journal.pmed.0020295 16231975
Jarrett WFH Jennings FW McIntyre WIM Mulligan W Sharp NCC A pasture trial using two immunising doeses of a parasitic bronchitis vaccine Am J Vet Res 1961 22 492 494 13789594
Miller TA Industrial development and field use of the canine hookworm vaccine Adv Parasitol 1978 16 333 332 364958
Williamson AL Lecchi P Turk BE Choe Y Hotez PJ A multi-enzyme cascade of hemoglobin proteolysis in the intestine of blood-feeding hookworms J Biol Chem 2004 279 35950 35957 15199048
Williamson AL Brindley PJ Abbenante G Prociv P Berry C Cleavage of hemoglobin by hookworm cathepsin D aspartic proteases and its potential contribution to host specificity FASEB J 2002 16 1458 1460 12205047
Knox DP Smith WD Vaccination against gastro-intestinal nematode parasites of ruminants using gut-expressed antigens Vet Parasitol 2001 100 21 32 11522403
Hotez PJ Zhan B Bethony JM Loukas A Williamson A Progress in the development of a recombinant vaccine for human hookworm disease: The Human Hookworm Vaccine Initiative Int J Parasitol 2003 33 1245 1258 13678639
Bethony J Loukas A Smout M Brooker S Mendez S Antibodies against a secreted protein from hookworm larvae reduce the intensity of hookworm infection in humans and vaccinated laboratory animals 2005 FASEB J. E-pub ahead of print
Loukas A Bethony JM Williamson AL Goud GN Mendez S Vaccination of dogs with a recombinant cysteine protease from the intestine of canine hookworms diminishes the fecundity and growth of worms J Infect Dis 2004 189 1952 1961 15122534
Summers RW Elliott DE Urban JF Thompson RA Weinstock JV
Trichuris suis therapy for active ulcerative colitis: A randomized controlled trial Gastroenterology 2005 128 825 832 15825065
Hotez PJ Bethony J Bottazzi ME Brooker S Buss P Hookworm: “The great infection of mankind” PLoS Med 2005 2 e67 10.1371/journal.pmed.0020067 15783256
|
16187734
|
PMC1240053
|
CC BY
|
2021-01-05 10:38:16
|
no
|
PLoS Med. 2005 Oct 4; 2(10):e327
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020327
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020331SynopsisGenetics/Genomics/Gene TherapyPharmacology/Drug DiscoveryCardiology/Cardiac SurgeryPathologyCardiovascular MedicineGeneticsPathologyADMA in Vascular Disease: More than a Marker? Synopsis10 2005 4 10 2005 2 10 e331Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Effects of ADMA upon Gene Expression: An Insight into the Pathophysiological Significance of Raised Plasma ADMA
==== Body
The endothelium plays a crucial role in the maintenance of vascular tone and structure. Endothelial dysfunction is associated with cardiovascular risk factors, metabolic diseases, and systemic or local inflammation. One proposed mechanism for the development of endothelial dysfunction is the presence of elevated blood levels of asymmetric dimethylarginine (ADMA), an analogue of the amino acid L-arginine. The concentration of ADMA in the plasma of healthy adults varies between 0.4 µM and 1 µM, but it may increase to the range 1.45–4.0 µM in certain diseases.
Elevated ADMA is now widely recognized as a risk marker for vascular disease. Circulating concentrations of ADMA are increased in patients with renal failure, pulmonary hypertension, heart failure, hypercholesterolemia, and a range of other conditions associated with cardiovascular disease. In patients with end-stage renal failure, plasma levels of ADMA have been shown to predict mortality and cardiovascular outcome, and in a cohort of otherwise healthy men, those with the highest levels of ADMA had increased risk of acute coronary events. Increased circulating ADMA in pregnant women predicts increased risk of pre-eclampsia and intrauterine growth retardation.
But is ADMA just a risk marker, or does it have a causal role in the pathophysiology of cardiovascular disease? ADMA inhibits the formation of nitric oxide, a major endothelium-derived vasoactive mediator, and the most potent endogenous vasodilator known. An elevated level of ADMA could, thus, impair vascular function. However, some have argued that the concentration of ADMA found in plasma, even in disease states, is too low to be an effective inhibitor of nitric oxide synthase, and that the usual concentrations of arginine in cells should overcome any inhibitory effects of ADMA on nitric oxide synthase. In order to determine how ADMA might exert effects on endothelial cells and produce pathology, Caroline Smith and colleagues assessed the effects of ADMA on gene expression.
ADMA's effects on endothelial gene expression
The researchers treated human coronary artery endothelial cells with ADMA, and used microarrays to measure the effects on gene expression. They detected substantial changes in gene expression in these cells after 24 hours of exposure to concentrations of ADMA similar to those reported in pathophysiological states. Changes in several genes were confirmed by Northern blotting, quantitative PCR, and in some instances, at the protein level, by Western blotting. To determine whether such changes also occur in vivo, the team examined tissues from mice with elevated ADMA levels. Some of the genes exhibiting consistent changes pointed to pathways known to be associated with cardiovascular risk and pulmonary hypertension.
Smith and colleagues concluded that the concentrations of ADMA found in disease states affect the transcriptional profile of endothelial cells. Moreover, their results suggest novel mechanisms by which ADMA might contribute to or cause disease. Changes in bone morphogenetic protein signaling, and in enzymes involved in arginine methylation, may be particularly relevant to understanding the pathophysiological significance of raised ADMA levels. The effects on bone morphogenetic protein signaling may be important in renal disease and in the link between raised ADMA and pulmonary hypertension. The hope is that in the long-term understanding the mechanisms by which increased ADMA contributes to cardiovascular diseases might suggest new therapeutic strategies.
|
0
|
PMC1240054
|
CC BY
|
2021-01-05 11:13:37
|
no
|
PLoS Med. 2005 Oct 4; 2(10):e331
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020331
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1618773510.1371/journal.pmed.0020338EssayHIV/AIDSInfectious DiseasesHIV Infection/AIDSResource allocation and rationingAdherence to Antiretroviral Therapy: Merging the Clinical and Social Course of AIDS EssayCastro Arachu Arachu Castro is Assistant Professor of Social Medicine in the Program in Infectious Disease and Social Change, Department of Social Medicine, Harvard Medical School, Boston, Massachusetts, United States of America; Director at the Institute for Health and Social Justice, Partners In Health, Boston, Massachusetts, United States of America; and Medical Anthropologist in the Division of Social Medicine and Health Inequalities, Department of Medicine, Brigham and Women's Hospital, Boston, Massachusetts, United States of America. E-mail: [email protected].
Competing Interests: The author declares that no competing interests exist.
12 2005 4 10 2005 2 12 e338Copyright: © 2005 Arachu Castro.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Compliance with HIV treatment is affected by many issues and social factors are as important as biological ones.
==== Body
The survival of people diagnosed with HIV/AIDS dramatically improves with access to highly active antiretroviral therapy (HAART). Such therapy employs a combination of antiretroviral agents—protease inhibitors (PIs), nucleoside reverse transcriptase inhibitors, non-nucleoside reverse transcriptase inhibitors, nucleotide reverse transcriptase inhibitors, and fusion inhibitors—to suppress viral replication, and, thus, reduces the likelihood of developing HIV mutations that could lead to the development of drug-resistant viral strains. HAART also prevents further viral destruction of the cellular immune system, thereby, allowing for increases in the level of CD4+ cells, which improves the immunologic response to opportunistic infections.
A physician in Havana, Cuba, holds a bottle of antiretroviral therapy
(Photo: Arachu Castro)
However, suboptimal treatment adherence has been associated with virologic, immunologic, and clinical failure. In this essay, I look critically at the issue of adherence, and argue that, to address causes of incomplete adherence, we need to combine both quantitative and qualitative methodologies. These methodologies must be grounded in an understanding of adherence as a biological and social process that changes with time, and must be framed within an analysis of access to health care and medications.
Adherence and Drug Resistance
A consensus exists that in order to achieve an undetectable viral load and prevent the development of drug resistance, a person on HAART needs to take at least 95% of the prescribed doses on time [1]. For many people, this means taking a regimen of three antiretrovirals twice per day—on both occasions, they are usually taking several pills [2]. An increasing number of studies show that the relationship between adherence and resistance is drug specific [3]. Although the suggestion that this relationship follows a bell-shaped curve has existed since just after the introduction of HAART [4], there is increasing evidence that drug resistance is highest among those taking 70%–80% of regimens containing a nonboosted PI (i.e., regimens with no combined ritonavir), and among those with intermittent or single-dose regimens of non-nucleoside reverse transcriptase inhibitors (including when nevirapine is used once to prevent mother-to-child transmission of HIV) or with poor adherence to these types of antiretrovirals. Ritonavir-boosted PIs (a full dose of a PI combined with ritonavir to increase the blood levels of the former) confer limited resistance, regardless of one's level of adherence [3].
Given the relevance of treatment adherence to improving life expectancy and preventing the spread of drug-resistant strains, many studies have attempted to predict causes of nonadherence in order to design strategies that reduce the number of missed doses. Except for some factors that have been associated with incomplete adherence in various settings, such as depression [5,6] or illegal drug use [7], study results are often inconclusive and do not yield comparable results—often due to conceptual and methodological differences among research protocols. Methodologically, there is growing agreement that patients' self-assessments of adherence—through interviews or self-administered questionnaires—show significant correlation with viral load tests [8–10], whereas estimations by their health-care providers often lead to invalid results [11].
Despite the current shortcomings in predicting who is more likely to miss doses of antiretroviral therapy (ART), the issue of adherence has become extremely important when setting priorities for allocating resources to fight AIDS in poor countries, where the majority of people who are HIV positive live. Some commentators have argued that adherence barriers are insurmountable in poor settings, so we should be cautious in delivering ART to these populations—a claim that is not grounded in evidence. In fact, adherence in poor settings is proving to be equal to, or even higher than, adherence in developed countries [12–17]. Furthermore, the argument about insurmountable adherence barriers in poor settings has also been challenged because it creates an unjustifiable double standard [11]—ART is not withheld from wealthy settings on the basis that many patients will skip doses.
Adherence as a Biosocial and Dynamic Phenomenon
A biosocial approach to adherence relies on the dynamic analysis of the clinical and social course of disease and the continuous interaction of biological and social processes over time [18–21]. The study of pathology embedded in social experience captures a series of distal and proximal factors acting together—such as not seeking treatment for undesirable side effects due to lack of money to travel to a health center or purposely missing doses when one is asymptomatic to pretend that AIDS is not a concern. These factors shape the everyday life of patients, while patients internalize them and use them to provide meaning to their disease experience.
Years ago, before the advent of AIDS, anthropologists had noted that the introduction of effective therapy for a particular disease may profoundly alter the social interpretations of that disease [18,22]. Exposure to AIDS in Haiti, in the 1980s, generated cultural models of its etiology and expected course [19,23,24], which aimed to provide meaning to otherwise unknown phenomena—often locally interpreted in terms of jealousy and curse. Likewise, the social experience of AIDS in rural Haiti is also deeply affected by the advent of effective therapy, as preliminary data suggest that the introduction of quality HIV care can lead to a rapid reduction in stigma, with resulting increased uptake of testing [21].
Within the changing context in which disease may take place, adherence level is likely to change as biological and social circumstances, and interpretations of them, unfold. In Brazil, a boy and a girl living in a support house for children and adolescents orphaned by AIDS or living with HIV explained why, upon feeling well, they had started to delay the morning dose of antiretrovirals until the afternoon: “[The schedule] doesn't matter; you can take them any time, as long as you take them” [25]. Until they started HAART, they had both suffered several opportunistic infections and, at times, had been at the brink of death. Their incomplete adherence to the regimen was, in part, a lack of understanding about the importance of not missing doses—but, fundamentally, it was a strategy they had developed as an act of defiance against the rules of the support house, and probably to feel more like the children who were HIV negative who also lived there and were not taking daily medications [25]. By delaying the morning dose, they were providing their lives with a sense of normalcy, which had otherwise been characterized by orphanhood and chronic disease since early age, while acting out in protest.
By any account, these children would be classified as having incomplete adherence or, plainly, as nonadherent. Yet to improve their adherence, how helpful is it to know that they are not taking all their medications on time without understanding why? Could their nonadherence really be understood without analyzing the life trajectories of these children and their social context? Some studies approaching ART adherence within a dynamic framework have relied on a biosocial approach, providing a rich context in which taking medications occurs and evolves [7,9,26–28]. Other studies, while having observed that the level of adherence is not a static value, have remained mostly biomedical, examining the impact of side effects [29] or substance abuse [30], and have been devoid of the complexity of biosocial interactions and their changing nature, as in the case of the Brazilian children.
The study of life histories of patients—a standard qualitative method of ethnographic research—and of the interactions of social experience with illness episodes allows us to generate associations between the clinical and the social course of disease, including such themes as stigma, health-seeking behavior, or adherence to therapy. These associations, which are often drawn from a small sample of patients, can be validated by larger, statistically representative, epidemiological studies designed to include variables reflecting the social context of patients. For example, the effect of adipose tissue alterations (lipodystrophy)—a common side effect of PIs—on adherence to ART has often been studied without considering local patterns of ideal weight and body shape for women and men at different ages, and existing variations of these ideals related to the social position of patients. This lack of consideration for local patterns may explain why some studies arrive at opposite results—dystrophic weight gain as a barrier to or as an enhancer of adherence [29,31]. The effect of adherence on weight gain seen in patients on ART, as a result of a reversal of the disease process and general clinical improvements, should also be analyzed in relation to these social ideals. In Senegal, for example, where weight gain is a symbol of good health, such weight gain has been shown to increase adherence [26].
Most research studies on adherence to ART share the basic understanding that patients are adherent when they, after agreeing to the recommendations of a health-care provider [32], take the prescribed medications in a timely manner. However, an overemphasis on pill counting as a sum of discrete events limits our understanding of adherence as a complex process embedded in the clinical and social course of AIDS, as the case of the Brazilian children shows. An approach to adherence that combines both biological and social knowledge—a biosocial approach—and that relies on qualitative and quantitative methodologies is more likely to move us closer to a better understanding of adherence and, eventually, to improving adherence to ART.
Poverty and Adherence
Partly because the introduction of ART in poor settings is recent, and partly because biomedical research rarely examines the social context in which patients live, there is a dearth of information on the direct effect of poverty on ART adherence. Most studies conducted in poor settings overlook how direct and indirect economic burdens borne by patients affect their ability to access a steady supply of antiretrovirals and take them on time. Such burdens may include the cost of missing work, the cost of elder or child care during medical visits, the cost of transportation to a health center, the cost of user fees, or the cost of tests and supplies. Although these costs may seem minimal to health professionals and decision makers, bearing these costs often translates into difficult household decisions about who eats, who works, or who goes to school. Taking medications in a timely manner may also require the challenging tasks of obtaining food and safe water, or of readjusting food intake to fit the drug-regimen schedule.
Despite the difficulties in overcoming these obstacles, the inability of a person living in poverty to obtain and take medications after initiating therapy is often labeled “noncompliance” or “nonadherence”—as has often occurred with tuberculosis patients [33,34]—and categorized as patient-related characteristics, ignoring social and economic causes or failures on the part of public-health interventions to address those causes [35]. Some studies conducted in Côte d'Ivoire [36], Senegal [28,37,38], and Botswana [39,40] show that user fees not only deter people from accessing AIDS care but also create an obstacle to treatment adherence. In other contexts where ART is free, such as Costa Rica, transportation costs have been associated with lower adherence [41]. The argument that patients would not value free drugs, or that free treatment might be humiliating for patients, are not borne out by higher adherence rates when drugs are free, such as in a comprehensive AIDS program in rural Haiti [14] or in Cuba [17], or when user fees are lowered, such as in Senegal [37,42]. Indeed, variations of directly observed therapy (DOT) for the delivery of ART (known as DOT-HAART) have proven useful in introducing complex multidrug regimens in poor settings lacking health-care infrastructures. In rural Haiti, for example, support for patients receiving DOT-HAART from community health workers improves rates of adherence [14].
As adequate and equitable access to comprehensive AIDS prevention and care are introduced, optimal adherence could be achieved if the multiple causes that shape patients' adherence are analyzed within their social context—including those related to the financing of health-care systems, and particularly cost-recovery mechanisms.
A Biosocial Approach to Causes of Nonadherence: The Way Forward
The use of a biosocial framework grounded in the lived experience of people diagnosed with AIDS is essential to understanding adherence, the way adherence changes over time, and the reasons for nonadherence. Often times, particularly in poor settings, these reasons will be found outside the individual responsibility of patients. Addressing adherence may require providing social support to patients, lowering or eliminating user fees, bringing health-care workers closer to patients, opening health centers focused on patients' competing demands to survive, improving drug procurement strategies, or creating mechanisms for lowering the cost of drugs and lab tests. In many cases it will mean improving and investing in primary health care, public hospitals, and referral networks, or in incentives to recruit and retain health-care workers committed to serving their patients. Given this complexity, the possibility of advancing the understanding of the multifaceted causes of nonadherence needs to be analyzed within its larger social, economic, and political context.
Box 1 shows the main biosocial variables needed to analyze adherence to ART defined within eight broad categories: socioeconomic factors, health-care system, social capital, cultural models of health and disease, personal characteristics, psychological factors, clinical factors, and antiretroviral regimen. Most clinical studies have focused on the last four areas, which are easier to measure quantitatively but which do not account for the larger social context. Although clinical epidemiological studies are essential to finding associations between drug regimens and adherence—and, depending on the method chosen, to establishing causality—a biosocial approach that combines quantitative and qualitative methodologies is necessary to bridge the current gap in knowledge on adherence to ART. Only by understanding the complicated interplay between the clinical and social factors that affect adherence to ART can we hope to overcome the real causes of nonadherence.
Box 1. Biosocial Variables for Understanding Adherence to ART
Socioeconomic Factors
Poverty and inequality
War, political violence
Cost of medications
Cost of CD4+ counts, viral load
Transportation costs
Cost of missing days from work
Cost of food and safe water
Costs associated to changes in lifestyle
Other costs
Health-Care System
Health-care infrastructure
Drug stock shortages
Financing mechanisms (including user fees)
Quality of relationship with health-care providers
Social Capital
Kinship patterns
Networks of social support
Social status
Homelessness, incarceration
Cultural Models of Health and Disease
On etiology and transmission
On health-care providers and healers
On cure
On efficacy and toxicity of drugs
On sick role
Personal Characteristics
Age
Sex and gender
Ethnic group
Education
Occupation
Household composition
Substance abuse
Physical disabilities
Psychological Factors
Self-esteem and motivation
Mental health conditions
Clinical Factors
Immunological or clinical stage of HIV disease
Occurrence and severity of opportunistic infections
Side effects (desirable and undesirable)
Symptomatology at onset of treatment
Effect of pregnancy or lactation
Antiretroviral Regimen
Number of drug regimens per day
Number of pills per regimen
Therapeutic class composition of drug regimen
I am indebted to Alice Wilson, Yasmin Khawja, and Susan Mathai for their research assistance in the preparation of this paper, and to Carlos Aragonés, Jorge Pérez, César Abadía-Barrero, and Philip Onyebujoh for our fruitful conversations on adherence to therapy. I am also thankful to David Bangsberg for his constructive comments on an earlier draft.
Citation: Castro A (2005) Adherence to antiretroviral therapy: Merging the clinical and social course of AIDS. PLoS Med 2(12): e338.
Abbreviations
ARTantiretroviral therapy
DOTdirectly observed therapy
HAARThighly active antiretroviral therapy
PIprotease inhibitor
==== Refs
References
Paterson D Swindells S Mohr J Brester M Vergis EN Adherence to protease inhibitor therapy and outcomes in patients with HIV infection Ann Intern Med 2000 133 21 30 10877736
Partners In Health The PIH guide to the community-based treatment of HIV in resource-poor settings 2004 Boston Partners In Health Available: http://www.pih.org/library/aids/PIH_HIV_Handbook_Bangkok_edition.pdf . Accessed 25 August 2005
Bangsberg D Moss A Deeks S Paradoxes of adherence and drug resistance to HIV antiretroviral therapy J Antimicrob Chemother 1999 53 696 699
Friedland G Williams A Attaining higher goals in HIV treatment: The central importance of adherence AIDS 1999 13 S61 S72
Carrieri MP Chesney MA Spire B Loundou A Sobel A Failure to maintain adherence to HAART in a cohort of French HIV-positive injecting drug users Int J Behav Med 2003 10 1 14 12581944
Starace F Ammassari A Trotta MP Murri R De Longis P Depression is a risk factor for suboptimal adherence to highly active antiretroviral therapy J Acquir Immune Defic Syndr 2002 31 S136 S139 12562037
Ware N Wyatt M Tugenberg T Adherence, stereotyping and unequal HIV treatment for active users of illegal drugs Soc Sci Med 2005 61 565 576 15899316
Sethi AK Celentano DD Gange SJ Moore RD Gallant JE Association between adherence to antiretroviral therapy and human immunodeficiency virus drug resistance Clin Infect Dis 2003 37 1112 1118 14523777
Spire B Duran S Souville M Leport C Raffi F Adherence to highly active antiretroviral therapies (HAART) in HIV-infected patients: From a predictive to a dynamic approach Soc Sci Med 2002 54 1481 1496 12061483
Chesney M Morin M Sherr L Adherence to HIV combination therapy Soc Sci Med 2000 50 1599 1605 10795966
Moatti JP Spire B Kazatchkine M Drug resistance and adherence to HIV/AIDS antiretroviral treatment: Against a double standard between the north and the south AIDS 2004 18 S55 S61
Coetzee D Boulle A Hildebrand K Asselman V Van Cutsem G Promoting adherence to antiretroviral therapy: The experience from a primary care setting in Khayelitsha, South Africa AIDS 2004 18 S27 S31
Katzenstein D Laga M Moatti JP The evaluation of the HIV/AIDS drug access inititiatives in Côte d'Ivoire, Senegal and Uganda; how access to antiretroviral treatment can become feasible in Africa AIDS 2003 17 S1 S4
Koenig S Léandre F Farmer P Scaling-up HIV treatment programmes in resource-limited settings: The rural Haiti experience AIDS 2004 18 S21 S25 15322480
Lanièce I Ciss M Desclaux A Diop K Mbodj F Adherence to HAART and its principal determinants in a cohort of Senegalese adults AIDS 2003 17 S103 S108
Nemes M Carvalho H Souza M Antiretroviral therapy adherence in Brazil AIDS 2004 18 S15 S20 15322479
Pérez J Pérez D González I Jidy MD Orta M Approaches to the Management of HIV/AIDS in Cuba. Perspectives and Practice in Antiretroviral Treatment (Case Study) 2004 Geneva World Health Organization Available: http://www.who.int/hiv/pub/prev_care/en/cuba.pdf . Accessed 25 August 2005
Kleinman A Wang WZ Li SC Cheng XM Dai XY The social course of epilepsy: Chronic illness as social experience in interior China Soc Sci Med 1995 40 1319 1330 7638642
Farmer P Sending sickness: Sorcery, politics, and changing concepts of AIDS in rural Haiti Med Anthropol Q 1990 4 6 27
Singer M Clair S Syndemics and public health: Reconceptualizing disease in bio-social context Med Anthropol Q 2003 17 423 441 14716917
Castro A Farmer P Understanding and addressing AIDS-related stigma: From anthropological theory to clinical practice in Haiti Am J Public Health 2005 95 53 59 15623859
Lévi-Strauss C Anthropologie structurale 1958 Paris Plon 480
Farmer P AIDS and accusation: Haiti and the geography of blame 1992 Berkeley University of California Press 338
Farmer P AIDS-talk and the constitution of cultural models Soc Sci Med 1994 38 801 809 8184331
Abadía-Barrero C Castro A Experiences of stigma and access to HAART in children and adolescents living with HIV/AIDS in Brazil Soc Sci Med 2005 Epub ahead of print
Desclaux A Les antiretroviraux en Afrique: de la culture dans une économie mondialisée Anthropol Soc 2003 27 41 58
Ickovics JR Meade CS Adherence to antiretroviral therapy among patients with HIV: A critical link between behavioral and biomedical sciences J Acquir Immune Defic Syndr 2002 31 S98 S102 12562029
Lanièce I Desclaux A Ciss M Diop K Ndiaye B Desclaux A Lanièce I Ndoye I Taverne B L'observance des traitements antirétroviraux et ses déterminants Analyse quantitative 2002 Paris ANRS 97 107 L'Initiative sénégalaise d'accès aux médicaments antirétroviraux: Analyses économiques, sociales, comportementales et médicales
Ammassari A Antinori A Cozzi-Lepri A Trotta MP Nasti G Relationship between HAART adherence and adipose tissue alterations J Acquir Immune Defic Syndr 2002 31 S140 S144 12562038
Tesoriero J French T Weiss L Waters M Finkelstein R Stability of adherence to highly active antiretroviral therapy over time among clients enrolled in the treatment adherence demonstration project J Acquir Immune Defic Syndr 2003 33 484 493 12869837
Guaraldi G Murri R Orlando G Orlandi E Sterrantino G Morphologic alterations in HIV-infected people with lipodystrophy are associated with good adherence to HAART HIV Clin Trials 2003 4 99 106 12671777
World Health Organization Adherence to long-term therapies. Evidence for action 2003 Geneva World Health Organization 110
Castro A Farmer P Violence structurelle, mondialisation et tuberculose multirésistante Anthropol Soc 2003 27 23 40
Greene JA An Ethnography of Nonadherence: Culture, poverty, and tuberculosis in urban Bolivia Cult Med Psychiatry 2004 28 401 425 15600119
Castro A Singer M Unhealthy health policy: A critical anthropological examination 2004 Walnut Creek (California) Altamira Press 408
Delaunay K Vidal L Msellati P Moatti JP Msellati P Vidal L Moatti JP La mise sous traitement antirétroviral dans l'Initiative: l'explicite et l'implicite d'un processus de sélection 2001 Paris ANRS 87 113 L'accès aux traitements du VIH/sida en Côte d'Ivoire. Aspects économiques, sociaux et comportementaux
Desclaux A Castro A Singer M Equality in access to AIDS treatment in Africa: Pitfalls among achievements 2004 Walnut Creek (California) Altamira Press 408 Unhealthy health policy: A critical anthropological examination
Lanièce I Diop K Desclaux A Sow K Ciss B Determinants of adherence among adults receiving antiretroviral drugs in Senegal (ANRS 1215 Cohort Study) Antivir Ther 2003 8 S501 S502
Nwokike JI Baseline data and predictors of adherence to antiretroviral therapy in Maun General Hospital (MGH), Maun, Botswana Antivir Ther 2003 8 S396
Weiser S Wolfe W Bangsberg D Thior I Gilbert P Barriers to antiretroviral adherence for patients living with HIV infection and AIDS in Botswana J Acquire Immune Defic Syndr 2003 34 281 288
Stout B Leon M Niccolai L Determining risk factors associated with non-adherence in HIV patients in Costa Rica [abstract] 2003 2nd International AIDS Society Conference on HIV Pathogenesis and Treatment; 2003 July 13–16 July; Paris, France. Abstract no. 675. Available: http://www.iasociety.org/abstract/show.asp?abstract_id=9974 . Accessed 29 August 2005
Desclaux A Souteyrand Y Morin M L'observance en Afrique: Question de culture ou vieux problèmes de santé publique? 2001 Paris ANRS 57 66 L'observance aux antirétroviraux
|
16187735
|
PMC1240055
|
CC BY
|
2021-01-05 10:40:26
|
no
|
PLoS Med. 2005 Dec 4; 2(12):e338
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020338
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020369SynopsisInfectious DiseasesMicrobiologyAllergy/ImmunologyVaccinesParasitologyImmunology and allergyInfectious DiseasesGetting Closer to a Vaccine for Hookworm Synopsis10 2005 4 10 2005 2 10 e369Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Vaccination with Recombinant Aspartic Hemoglobinase Reduces Parasite Load and Blood Loss after Hookworm Infection
==== Body
Hookworms are intestinal parasites of mammals, including humans, dogs, and cats; in humans, these infections are a leading cause of intestinal blood loss and iron-deficiency anemia. These infections occur mostly in tropical and subtropical climates, and are estimated to infect about 1 billion people worldwide—about one-fifth of the world's population. People who have direct contact with soil that contains human feces in areas where hookworm is common are at high risk of infection; because children play in dirt and often go barefoot, they are at highest risk.
However, since transmission of hookworm infection requires development of the larvae in soil, hookworm cannot be spread person to person. Anthelminthic chemotherapy with benzimidazole drugs is effective at eliminating existing adult parasites. But since reinfection occurs rapidly after treatment, making a vaccine against hookworm disease is a public health priority. Previous animal vaccine studies have had mixed results. Dogs have been successfully vaccinated against infection with the dog hookworm Ancylostoma caninum by immunization with attenuated third-stage infective larvae (L3). Varying levels of efficacy have been reported for vaccination against the major antigens secreted by the same larval stage in hamsters and dogs. However, only partial reductions in parasite load have been reported. In addition, protective antigens from the larval stage are only expressed in larvae, not in adult worms; hence, antibodies against L3 secretions are useless against adult stage parasites in the gut.
In this month's PLoS Medicine, Alex Loukas and colleagues suggest that the ideal hookworm vaccine would be a mixture of two recombinant proteins, targeting both the infective larva and the blood-feeding adult stage of the parasite. Such a vaccine would limit the amount of blood loss caused by feeding worms and maintain normal levels of hemoglobin, said the authors. This outcome is particularly important in young children and women of childbearing age, where menstrual and, particularly, fetal hemoglobin demands are high.
Hookworms secrete proteins that are being used as vaccines in animal models
Of the different proteins expressed by blood-feeding parasitic helminths, proteolytic enzymes have shown promise as intervention targets for vaccine development. A previous study in which dogs were vaccinated with a catalytically active recombinant cysteine hemoglobinase, Ac-CP-2, induced antibodies that neutralized proteolytic activity, and provided partial protection to vaccinated dogs by reducing egg output and worm size, but there were not significant reductions of adult worm burdens or blood loss.
In the present study, the researchers found that vaccination of dogs with recombinant Ac-APR-1, an aspartic hemoglobinase that initiates the hemoglobin digestion cascade in hookworms, induced antibody and cellular responses, and resulted in significantly reduced hookworm burdens and fecal egg counts in vaccinated dogs compared to control dogs after challenge with infective larvae of A. caninum. Most importantly, vaccinated dogs were protected against blood loss and most did not develop anemia, the major pathologic sequelae of hookworm disease.
The authors went on to show that IgG from vaccinated animals decreased the catalytic activity of the recombinant enzyme in vitro, and the antibody bound in situ to the intestines of worms recovered from vaccinated dogs, implying that the vaccine interfered with the parasite's ability to digest blood.
This result of vaccination against APR-1 shows the best efficacy so far reported for a recombinant vaccine aimed at reducing hookworm egg counts, intestinal worm burdens, and hookworm-induced blood loss, say the authors. They suggest that vaccination with APR-1 damaged the parasite's intestine and resulted in decreased blood intake by feeding worms, and, hence, reduced blood loss from the dogs.
The authors go on to suggest that the optimal hookworm vaccine would combine two elements: one to prevent L3 from developing into adult blood-feeding hookworms, and one to block the establishment, survival, and fecundity of the adult parasites in the intestine. Achieving both goals would require a vaccine comprised of an L3 antigen, such as ASP-2, which is now under clinical development, and an adult gut protease, such as APR-1.
These results have implications for human hookworm vaccine development; the authors finish by saying that there is now enough evidence to conclude that the counterpart vaccine for the major human hookworm Necator americanus (Na-APR-1) should be developed and entered into human clinical trials.
|
0
|
PMC1240056
|
CC BY
|
2021-01-05 10:39:17
|
no
|
PLoS Med. 2005 Oct 4; 2(10):e369
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020369
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020375SynopsisAllergy/ImmunologyEpidemiology/Public HealthAsthmaImmunology and allergyRespiratory MedicineTeasing Out the Effects of Latitude and Birth Date on Allergy Synopsis10 2005 4 10 2005 2 10 e375Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Latitude, Birth Date, and Allergy
==== Body
The prevalence of asthma and allergy, defined as immunologically mediated hypersensitivity, is increasing. It is estimated that more than 20% of the world's population has IgE-mediated allergic diseases. The scale of the clinical problem is immense. The World Health Organization estimates that asthma affects nearly 150 million people worldwide, and more than 180,000 deaths each year are due to asthma. Approximately US$20 billion is spent globally each year on allergic rhinitis, including medications, time off work, and clinician consultations. The cost of allergy drugs alone is estimated to be US$8 billion per annum.
While allergy prevalence is increasing, the causal risk factors are still unknown. Matthias Wjst and colleagues investigated whether the spatial (latitude) and temporal (birth month) distribution of risk factors might offer insight into the mechanism of disease.
Previous studies have already shown that birth month is a risk factor associated with allergy; as the authors point out, birth month, used as a proxy for early allergen exposure, might be associated with upper respiratory infections during winter months. Studies have also associated geographical latitude with allergy. But some experts have noted that latitude, a proxy for ultraviolet solar exposure, might also reflect climatic differences, genetic influences, or even cultural differences in the raising of children.
In this study, Wjst and colleagues tried to further understand the effects of latitude and birth date on the prevalence of allergy defined by markers such as allergic rhinitis, sensitization to grass or dust, and total IgE levels. They distributed a questionnaire to 20– to 44–year-old individuals in 54 centers across Europe, North Africa, India, North America, Australia, and New Zealand. Altogether, data from 200,682 participants were analyzed.
World and European distribution of study centres
The median prevalence of allergic rhinitis was 22%, but with a substantial variation across centers. They found allergic rhinitis decreased with geographical latitude, but there were many exceptions. There was no increase in prevalence during certain winters, and no altered risk by birth month, except borderline reduced risks in September or October. Altogether, the authors concluded that there was no major risk by being born in a particular month or during a particular season. There may be relevant birth month effects in single centers, but a global effect was questionable.
Previous research on the effect of birth month has also shown mixed results: differing studies have found associations that were positive, negative, or simply unclear.
But the authors noted that one difference of their study compared with others was the higher age of the subjects—interviewees were born between 1945 and 1973—and suggested that it might be possible that there are more marked symptoms in children that were being lost in adulthood.
Most previous studies have shown an association with allergic sensitization, indicating subclinical effects that gained importance only when occurring in combination with additional risk factors. One of the main advantages of this study—a standardized allergen test protocol—might, thus, be a disadvantage since the effects of local allergens might have been missed. Another methodological restriction might have been the use of self-reported “hay fever.” This term may be used in a different way across Europe, said the authors.
Data on the geographical distribution of allergic diseases are rare, and, hence, this study is valuable. However, a previous meta-analysis has shown negative association of latitude and symptoms of allergic rhinitis, with a −0.05% decrease per degree. In this study, symptoms of allergic rhinitis decreased with geographical latitude on a worldwide scale, but not when the analysis was restricted to Europe alone. One intriguing possibility, which needs further work, is that a risk factor within language borders might be more relevant than geographical latitude alone in determining the distribution of allergic diseases.
|
0
|
PMC1240057
|
CC BY
|
2021-01-05 10:38:16
|
no
|
PLoS Med. 2005 Oct 4; 2(10):e375
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020375
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1618779810.1371/journal.pmed.0020330Research ArticleInfectious DiseasesClinical PharmacologyEpidemiology/Public HealthHealth PolicyParasitologyHealth PolicyInfectious DiseasesMalariaPublic HealthEffect of Artemether-Lumefantrine Policy and Improved Vector Control on Malaria Burden in KwaZulu–Natal, South Africa Effect of AL and IRS on Malaria BurdenBarnes Karen I
1
*Durrheim David N
2
Little Francesca
1
3
Jackson Amanda
4
Mehta Ushma
1
Allen Elizabeth
1
Dlamini Sicelo S
4
Tsoka Joyce
4
Bredenkamp Barry
4
Mthembu D. Jotham
5
White Nicholas J
6
7
Sharp Brian L
4
1Division of Clinical Pharmacology, University of Cape Town, Cape Town, South Africa,2Health Protection, Hunter New England Population Health, Newcastle, New South Wales, Australia,3Department of Statistical Sciences, University of Cape Town, Cape Town, South Africa,4Malaria Research Lead Programme, Medical Research Council, Durban, South Africa,5Malaria Control Programme, KwaZulu–Natal Provincial Department of Health, South Africa,6Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand,7Centre of Vaccinology and Tropical Medicine, Churchill Hospital, Oxford, United KingdomGreenwood Brian Academic EditorUniversity of LondonUnited Kingdom*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: NJW is chairman of the World Health Organization malaria treatment guidelines committee and is on the editorial board of PLoS medicine. The authors have no other conflict of interest to declare.
Author Contributions: See Acknowledgments.
11 2005 4 10 2005 2 11 e33012 4 2005 11 8 2005 Copyright: © 2005 Barnes et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Rolling Back a Malaria Epidemic in South Africa
KwaZulu-Natal's Successful Fight against Malaria
Background
Between 1995 and 2000, KwaZulu–Natal province, South Africa, experienced a marked increase in Plasmodium falciparum malaria, fuelled by pyrethroid and sulfadoxine-pyrimethamine resistance. In response, vector control was strengthened and artemether-lumefantrine (AL) was deployed in the first Ministry of Health artemisinin-based combination treatment policy in Africa. In South Africa, effective vector and parasite control had historically ensured low-intensity malaria transmission. Malaria is diagnosed definitively and treatment is provided free of charge in reasonably accessible public-sector health-care facilities.
Methods and Findings
We reviewed four years of malaria morbidity and mortality data at four sentinel health-care facilities within KwaZulu–Natal's malaria-endemic area. In the year following improved vector control and implementation of AL treatment, malaria-related admissions and deaths both declined by 89%, and outpatient visits decreased by 85% at the sentinel facilities. By 2003, malaria-related outpatient cases and admissions had fallen by 99%, and malaria-related deaths had decreased by 97%. There was a concomitant marked and sustained decline in notified malaria throughout the province. No serious adverse events were associated causally with AL treatment in an active sentinel pharmacovigilance survey. In a prospective study with 42 d follow up, AL cured 97/98 (99%) and prevented gametocyte developing in all patients. Consistent with the findings of focus group discussions, a household survey found self-reported adherence to the six-dose AL regimen was 96%.
Conclusion
Together with concurrent strengthening of vector control measures, the antimalarial treatment policy change to AL in KwaZulu–Natal contributed to a marked and sustained decrease in malaria cases, admissions, and deaths, by greatly improving clinical and parasitological cure rates and reducing gametocyte carriage.
In KwaZulu-Natal strengthening of vector control and a change in antimalarial treatment policy to use of artemether-lumefantrine has been associated with a decrease in malaria cases, admissions, and deaths.
==== Body
Introduction
Malaria morbidity and mortality in Africa has risen, principally because of increasing resistance to chloroquine and sulfadoxine-pyrimethamine (SP) in Plasmodium falciparum [1,2]. Highly successful malaria control in South Africa, before the mid–1990s, had been achieved through high coverage with effective indoor residual spraying (IRS) and early access to effective antimalarial treatment. In KwaZulu–Natal in 1988, SP officially replaced chloroquine (CQ) as first-line treatment of uncomplicated malaria. In 1996, rapid immunochromatographic card tests were implemented to ensure definitive malaria diagnosis in all public-sector health-care facilities. Between 1995 and 2000 there was a dramatic increase in malaria morbidity and mortality in KwaZulu–Natal (Figure 1). Reinvasion of the area by the highly anthropophilic vector Anopheles funestus, which was resistant to pyrethroids, and a rapid increase in SP resistance in P. falciparum were considered the main contributors to this epidemic [3–5].
Figure 1 Number of Notified Malaria Cases in KwaZulu–Natal by Month (January 1993–December 2003)
The number of cases is given in relation to season (peak transmission from January to May, inclusive) and timing of significant malaria control interventions: A indicates reintroduction of DDT for IRS of traditional structures in KwaZulu–Natal in March 2000; B indicates introduction of IRS in southern Mozambique in October 2000; and C indicates implementation of AL as first-line treatment of uncomplicated falciparum malaria in KwaZulu–Natal in January 2001.
To improve control of the A. funestus vector, an effective insecticide, dichlorodiphenyltrichloroethane (DDT), was reintroduced in March 2000 to replace failing pyrethroids for IRS of traditional (mud, reed, or wood) homesteads. However, western style structures, which constitute at least 40% of homesteads in the study area, continued to be sprayed with the pyrethroids because the residue left by DDT on painted or cement-plastered surfaces is aesthetically unacceptable. By 2003, A. funestus s.s. was identified in only three sites in northern KwaZulu–Natal [6], whereas A. arabiensis continues to be found sporadically in window exit traps throughout northern KwaZulu–Natal.
By 2000 the 42-d cure rate after antimalarial treatment with SP had fallen to 11% [5]. In response to this drug-resistant malaria epidemic, KwaZulu–Natal was the first Ministry of Health in Africa to implement an artemisinin-based combination therapy (ACT) as first-line treatment of uncomplicated P. falciparum malaria. The decision to adopt an ACT policy was influenced by the sustained high cure rates, decreased malaria transmission, and decreased antimalarial resistance that had been documented on the western border of Thailand [7]. It was supported by a growing international consensus that wide-scale systematic implementation of ACT is one of few effective measures that could enable malaria-endemic countries to achieve the ambitious goals set in Abuja to “Roll Back Malaria,” particularly the halving of malaria morbidity and mortality by 2010 [8]. Improved cure rates and decreased gametocyte carriage had been confirmed in recent large randomised controlled clinical trials conducted across Africa [9,10]. Decreased gametocyte carriage after AL treatment has been shown to reduce post-treatment transmission of P. falciparum to Anopheles mosquitoes [11].
Artemether-lumefantrine (AL; Coartem, Novartis, Kempton Park, South Africa) was the only ACT available in 2000 with sufficient data to support fast-track registration by the South African drug regulatory authority. High levels of resistance in P. falciparum to CQ, SP, and amodiaquine in KwaZulu–Natal precluded their use in artemisinin-based combinations [5,12].
Although at least 14 African countries have recently adopted an ACT malaria treatment policy and an increasing number are in the process of changing to ACTs because of high levels of resistance to CQ and SP (the traditional first-line antimalarials) [13], concern has been expressed that the benefits of ACT observed in Asia have not yet been proven in Africa and may be influenced by coverage, adherence, quality, affordability, and access issues [14–17]. International subsidy of ACT costs has been widely recommended to address affordability constraints, although supply and quality issues currently remain substantial obstacles [18–20].
This study is a comprehensive evaluation of the first programme-wide implementation of ACT in Africa, describing changes in clinical and parasitological cure rates, gametocyte carriage, community perspectives on malaria treatment, and the impact of this ACT deployment and the strengthening of vector control on the number of malaria cases, admissions, and deaths. The factors considered to have contributed to the observed changes in morbidity and mortality are described.
Methods
Study Site
KwaZulu–Natal province, which has approximately 600,000 people living in malaria-risk areas, experienced the highest intensity malaria transmission in South Africa before 2001. Malaria risk and distribution is monitored through the provincial malaria geographic information systems that record all notified malaria cases, and provincial records are collated nationally [21]. Malaria transmission in South Africa has been restricted over the past four decades to the northeastern border areas with Mozambique, Swaziland, and Zimbabwe, primarily as a result of annual widespread IRS with insecticides by the provincial malaria control programmes, just prior to the malaria transmission season [22] and early effective treatment. Historically, the main mosquito vectors were the A. funestus group, which was considered to have been eradicated by IRS, and A. arabiensis, an efficient vector that displays both indoor and outdoor biting and resting behaviour, and thus is less readily controlled by IRS [23]. Following the identification of A. funestus resistant to pyrethroids in 1999 [3], traditional structures were sprayed with DDT (2 g/m2) although a pyrethroid (deltamethrin 0.02 g/m2) had to be applied to western-style structures.
Malaria transmission in KwaZulu–Natal is low (annual entomological inoculation rate <1; B. Sharp, unpublished data) and seasonal (Figure 1) [22]. The four sentinel health-care facilities studied, namely Ndumo clinic and Mosvold, Manguzi, and Bethesda rural district hospitals, are all located in Umkhanyakude district, which bears the heaviest malaria burden in KwaZulu–Natal (Figure 2). P. falciparum accounts for the majority (85%–100%) of infections.
Figure 2 Map of Umkhanyakude District, Northern KwaZulu–Natal, South Africa
The map indicates the following: the malaria risk by section and the four sentinel facilities for malaria morbidity and mortality review (Ndumo clinic, and Mosvold, Manguzi, and Bethesda rural district hospitals); the communities selected for the household (HH) survey and FGDs; and the Manguzi district hospital where sentinel safety surveillance and Ndumo Clinic where the SP (2000) and AL (2002) in vivo therapeutic efficacy studies were conducted.
South Africa has a decentralized health-care system. Antimalarial resistance patterns vary geographically. Malaria treatment policies have differed between provinces since 1988, when KwaZulu–Natal introduced SP to replace failing CQ, nine years before SP was introduced in the two other South African provinces with malaria transmission.
AL was implemented officially as first-line treatment of uncomplicated malaria in KwaZulu–Natal's public-health sector during January 2001 [24]. Malaria treatment in South Africa is administered on the basis of a P. falciparum–positive malaria smear or rapid immunodiagnostic card test. These tests are performed on all patients in whom malaria is clinically suspected; diagnostic quality control is regularly assessed by the Department of Health and the National Health Laboratory Services. Malaria treatment is provided free of charge in public-sector health-care facilities. AL was implemented 6 mo after the high SP failure rates were detected, and this delay consisted of a 3-mo policymaking process and a further 3-mo implementation preparation phase. AL was distributed to all public-sector fixed clinics (n = 50) and district hospitals (n = 5) in Umkhanyakude district in 24-tablet individual patient blister packs, for administration according to age–weight categories in a six-dose regimen over 3 d. Quinine remained the recommended treatment for severe malaria and for uncomplicated malaria in pregnant women and infants under 1 y of age. The implementation of AL included face-to-face training of public-sector health-care providers and distribution of specific malaria treatment guidelines and wall charts, and was followed by systematic withdrawal of SP.
Facility Review of Malaria Cases, Admissions, and Deaths
Malaria morbidity and mortality data were collected retrospectively by reviewing hospital records provided by the medical superintendents between 2000 and 2003 at the sentinel clinic, Ndumo clinic, and the three sentinel hospitals, namely Manguzi, Mosvold, and Bethesda hospitals that serve Umkhanyakude district where the vast majority of malaria cases occur in KwaZulu–Natal (Figure 2). Malaria cases were defined as patients with clinical features of malaria in whom Plasmodium parasites were detected on malaria smear or rapid diagnostic test. Those malaria cases requiring admission for management of severe disease or that were considered a high-risk group, including infants and pregnant women, were classified as malaria-related hospital admissions. Malaria-related deaths were patients in whom malaria was included as a cause of death on the death certificate. Death-register data included those deaths that occurred outside the hospitals but were registered at a hospital for the purpose of issuing a death certificate required for burial.
In Vivo Therapeutic Efficacy Study
Between January and June 2002, an open-label in vivo study was conducted to determine the therapeutic efficacy of a six-dose regimen of AL administered over 3 d (total adult dose 480 mg artemether/2,880 mg lumefantrine; Coartem, Novartis, Kempton Park, South Africa). Patients presenting sequentially to Ndumo clinic with an axillary temperature ≥ 37.5 °C or a history of fever, who were older than 12 mo, weighed more than 10 kg, and lived close enough to the clinic to allow reliable follow-up, were screened for P. falciparum infection using a rapid immunochromatographic card test for detecting histidine-rich protein-2 (Malaria PF/PV ICTML02;SA Scientific Products, Midrand, South Africa). P. falciparum infection was confirmed by a positive Giemsa-stained thick blood smear. Informed consent was sought from all patients (or their guardians) with proven uncomplicated falciparum malaria and a parasite density between 1,000 and 500,000 asexual parasites/μl blood. Patients who reported receiving antimalarial treatment in the previous 7 d, pregnant women, or those with severe malaria [25] or danger signs (e.g. prostrate, repeated vomiting, dehydrated) were excluded. A pregnancy test (Human Chorionic Gonadotrophin Combo; Abbotts, Johannesburg, South Africa) was used to exclude pregnancy in any woman unsure of her pregnancy status.
Eligible subjects received the six-dose AL treatment according to body weight, and this was co-administered with 250 ml “amahewu” (a non-alcoholic fermented gruel-like drink made from maize and containing 0.3 g fat/100 g). The first, third, and fifth doses were administered under observation at the clinic. Patients were observed for 1 h following treatment and then allowed to return home. If vomiting occurred within 30 min of treatment, the patient was re-treated with a full dose. Vomiting between 30 and 60 min after treatment resulted in re-treatment with half the dose. Patients were advised to take the second, fourth, and sixth doses at home. Self-reported adherence with home treatment and timing of administration were recorded, together with pill counts, at each subsequent clinic visit. Concomitant medication without known antimalarial activity was administered at the investigator's discretion to relieve malaria symptoms or treat concurrent disease.
Clinical and parasitological response to treatment, and occurrence of adverse events were monitored on days 0, 1, 2, 3, 7, 14, 21, 28, and 42 according to modified World Health Organization (WHO) procedures [26]. Parasite clearance time was defined as the interval between initiation of treatment and the first of two consecutive negative thick blood smears. Filter-paper blood spots were collected for polymerase chain reaction (PCR) differentiation of re-infection from recrudescence using nested PCR amplifications of blocks within the polymorphic genes encoding glutamate-rich protein and merozoite surface proteins I and II [27]. Results were compared with those from a preceding open-label in vivo study of the therapeutic efficacy of a single dose of SP monotherapy (at a dose of 1.25 mg/kg pyrimethamine; Fansidar; Roche, Isando, South Africa), which followed a similar protocol and was conducted at the same site in 2000 with 98 patients completing 42-day follow-up [5].
Subjects were withdrawn from the study if they developed any danger signs of severe malaria [25], parasitological failure, a serious adverse event requiring withdrawal, or if the patient or their guardian requested withdrawal. “Rescue therapy” with quinine (quinine sulfate; Lennon Limited, Port Elizabeth, South Africa) was administered at a dose of 10 mg/kg three times daily for 7 d to patients with parasitological or clinical failure.
Sentinel Pharmacovigilance
Intensive active sentinel surveillance for serious adverse events associated with antimalarial drug treatment was conducted at Manguzi hospital during and immediately after the malaria season (March–June 2002). This hospital was selected because it was the rural district hospital in the high-risk malaria transmission area with adequate resources to support intensive monitoring. The surveillance programme included any person presenting to Manguzi district hospital with a suspected serious adverse reaction after being treated with AL, or who was admitted for malaria, or had been treated for malaria in the 4 wk prior to hospital presentation. A standardised questionnaire was used to collect details of symptoms, clinical signs, and the results of special investigations for each patient. An international multidisciplinary panel, consisting of experienced clinical, laboratory, and public-health specialists (listed under acknowledgements), then reviewed the details of each case with any adverse effect to identify possible causal association with AL therapy. The total number of treatment courses distributed in the study area (Manguzi subdistrict) was used as a denominator for calculating the rates of adverse events.
In addition, all adverse events that occurred during the 42 d of follow-up in the in vivo therapeutic efficacy study were assessed to determine whether AL was a possible or probable cause according to guidelines of the International Committee for Harmonisation and the Medicines Control Council, the South African Drug Regulatory Authority [28,29]. These guidelines differ, in that lack of efficacy is classified as an adverse drug reaction in the Medicines Control Council guidelines, but not in the International Committee for Harmonisation guidelines.
Household Surveys
Care-seeking behaviour for fever and malaria, and patient adherence with prescribed antimalarial treatment, were assessed using a survey questionnaire administered in the local language, isiZulu, to household members by specifically trained field teams between 6 to 12 wk after the implementation of AL. The highest risk communities within KwaZulu–Natal, with malaria notification rates ranging from 250 to 800 cases of malaria per 1,000 population in 1998–1999, were identified using the Malaria Control Programme Geographic Malaria Information System platform [21]. However, Muzi 2 section could not be accessed because of severe floods, and was replaced by Ngutshana 6, the section with the next highest malaria notification rates recorded in 2000–2001. All 439 households in the seven selected malaria sections were included in the survey (Figure 2).
Care-seeking behaviour was explored for three groups of respondents: those who reported being diagnosed with malaria in the previous 4 wk, those who reported an episode of fever (not considered by the respondent to be malaria) in the previous 4 wk, and those who reported “ever having had” malaria. Malaria diagnosis is generally definitively confirmed in patients seeking care at formal health-care facilities (rapid diagnostic testing is provided free of charge to both public- and private-sector facilities by the malaria control programme), but would be diagnosed clinically in those seeking treatment from traditional healers or not seeking treatment outside the home.
Patient adherence to antimalarial treatment was assessed by asking three main questions deliberately interspersed between other interview topics: (1) Do you have any treatment for malaria at home? (2) Do you still have any of this treatment (from this most recent malaria episode) remaining? and (3) Did you complete the malaria treatment course for your most recent infection?
Focus Group Discussions
Four focus group discussions (FGDs) were conducted with 8–16 female household heads or caregivers in each of the areas selected for the household survey, to facilitate a greater understanding of community perspectives on treatment seeking and adherence to malaria treatment. The FGDs were conducted by an independent isiZulu–speaking qualitative researcher, in the home of a volunteering participant. Discussion topics included distinguishing between fever and malaria, management of children and adults with fever or suspected malaria, and factors influencing adherence with medication.
Analysis
Data were double entered, verified, and, for the household survey, analysed using Microsoft Access 2000 (Microsoft, Redmond, Washington, United States). EpiInfo 3.01 (Centers for Disease Control, Atlanta, Georgia, United States) was used for analysing sentinel hospital and safety surveillance data. SPSS Version 8.0 (SPSS, Chicago, Illinois, United States) was used for analysing the in vivo study. Normally distributed continuous variables were compared using the t-test for independent samples, and proportions were compared using chi square tests (Yates' corrected or Fisher's exact, where appropriate), or odds ratios.
Stata 8.0 (College Station, Texas, United States) was used to analyse the gametocyte data. The cumulative transmissibility measure, the area under the gametocyte time curve (AUC), was calculated using the linear trapezoidal rule. Gametocyte density was compared between groups using the ratio of arithmetic means. This was done by fitting generalized linear mixture models with a logarithmic link function to model the logarithm of the (arithmetic) mean gametocyte density and a logistic link function to model gametocyte prevalence using a zero-inflated negative binomial distribution. This accounted for the typically skewed distributions and excess variance, and allowed inclusion of all data points (including zeroes), modified from the method of Sutherland et al. [11]. This model generated parallel results that provided estimates of the relative risk of a zero gametocyte density and estimates of the incidence rate ratio of the mean gametocyte densities in the two treatment groups [30].
FGDs were independently analysed by two authors (KIB and SSD) who identified consistent themes or analytical categories in the data using an iterative approach, based on an adaptation of the grounded theory approach [31]. FGD findings were confirmed by a third independent reviewer. The findings of the household survey and FGDs were compared for consistency, thus reducing the possibility of systematic distortions inherent in using only one method [32].
Ethical Considerations
This study was approved by the Research Ethics Committees of the University of Cape Town and the South African Medical Research Council, and was planned and conducted in full partnership with the KwaZulu–Natal Ministry of Health. In the in vivo studies, written informed consent was obtained in the patients' local language, isiZulu, from all literate patients or guardians, and an independent literate witness confirmed verbal consent for illiterate patients or guardians who also recorded their consent as an “X” on the consent form. Verbal informed consent was obtained from participants in the household survey and FGDs. Confidentiality of patient identity was maintained for all records.
Results
Review of Malaria Cases and Deaths
The catchment areas of the three sentinel hospitals studied include almost 285,000 (47%) of the estimated 600,000 persons at risk of malaria in KwaZulu–Natal in 2000; these facilities carry the heaviest malaria burden because they are in the highest risk area in the far northeast of the province. There was a very marked reduction in the number of malaria cases, hospital admissions, and deaths in all these sentinel health-care facilities between 2000 and 2001, that were sustained (Table 1). Between 2000 and 2001 the number of malaria deaths and admissions both decreased by 89%, and malaria outpatient cases decreased by 85%. These initial reductions in malaria cases (96% versus 86%), admissions (92% versus 82%). and deaths (93% versus 78%) were greatest at Mosvold hospital and lowest at Manguzi hospital. Because Manguzi hospital borders on southern Mozambique, the migrant proportion of their patients may not have benefited to the same extent from the change in the first-line treatment policy in KwaZulu–Natal. The KwaZulu–Natal Department of Health estimates that Manguzi hospital serves a catchment population of approximately 20,000 people from southern Mozambique.
Table 1 Confirmed Malaria Cases and Malaria-Related Hospital Admissions and Deaths in the Sentinel Health-Care Facilities with the Highest Incidence of Malaria in Northern KwaZulu–Natal, South Africa, between 2000 and 2003
The marked reduction in malaria morbidity and mortality continued at all sentinel facilities. By 2003 the number of malaria-related outpatient cases and admissions had decreased by 99%, and malaria-related deaths had decreased by 97%. Trend analysis showed a highly significant decrease at all sentinel hospitals (p < 0.001).
In Vivo Study of Therapeutic Efficacy
Between January and June 2002, 100 patients (53 female) with uncomplicated malaria who sequentially met inclusion criteria were enrolled in this study. Patients had a median age of 14 y (interquartile range [IQR] 9–28 y) with 5% under 5 y of age. The geometric mean parasite density at baseline was 26,705/μl (95% confidence interval [CI]: 20,385 to 34,985). The median duration of symptoms at presentation was 3 d (range 0–14 d). Paracetamol (approximately 15 mg/kg) was administered at least once to 87% of patients. Antibiotics (without antimalarial activity) were administered to three patients in addition to AL. Two patients were lost to follow-up on day 28, and one patient missed his follow-up visit on day 28 but attended the day 42 visit. Parasites reappeared in one patient on day 28. Because the PCR results for this patient were indeterminate, this was assumed conservatively to be a recrudescence. There were no other parasitological failures and no clinical treatment failures. Of the 100 patients, two carried gametocytes after treatment; both these patients had gametocytes present in their blood smears at enrolment. No gametocytes developed in any patient in whom gametocytes were absent on day 0.
These results were compared with the in vivo study of the therapeutic efficacy of SP conducted in 2000 at the same sentinel study clinic, in a study population with a median age of 13 y (IQR 7–18 y) with 17% under 5 y of age.[6] The baseline geometric mean parasite density was significantly higher in the SP monotherapy study (62,114/μl [95% CI: 47,570–81,104]); p < 0.05. Pretreatment gametocyte carriage was 1/98 in 2000 and 3/100 in 2002 (p = 0.621). Parasite clearance time was significantly more rapid after AL (54 h [CI: 47–60]) than after SP (125 h [CI: 70–180]). Parasitological cure rates at 42 d of follow-up (corrected for PCR-confirmed reinfections) was 11/98 (11%) with SP monotherapy compared with 97/98 (99%) with AL in the present study (p < 0.001). Survival analysis confirmed markedly superior clinical and parasitological responses following AL, when compared with SP (log-rank test p < 0.001) (Figure 3). Early treatment failures decreased from 15/98 (15%) with SP to 0/98 with AL (p < 0.001) and late parasitological failures decreased from 72/98 (73%) to 1/98 (1%) (p < 0.001). Reinfections during follow-up decreased from 4/98 in 2000 to 0/98 in 2002 (p = 0.121).
Figure 3 Kaplan-Meier Survival Analysis of Time to Clinical or Parasitological Failure
Following treatment with SP in 2000 (n = 98) and artemether-lumefantrine in 2001 (n = 100), the proportion of patients with an adequate clinical and parasitological response to treatment at each day of follow up is shown.
During the 2000 SP in vivo therapeutic efficacy study, 52 (57%) of the 91 study subjects for whom gametocyte densities were recorded were found to carry gametocytes after treatment with SP compared to 2/100 (2%) after AL treatment (relative risk [RR]: 28.5 [95% CI: 7.16 to 113.97]); p < 0.001] (Figure 4). Gametocyte carriage peaked on day 7 following SP treatment, and on day 0 (prior to starting treatment) for AL. The median gametocyte area under the time curve (AUC) decreased significantly from 420 (IQR: 0–2,797) gametocytes/μl per person-week after SP to 0 (IQR: 0–0) gametocytes/μl per person-week after AL (Kruskal-Wallis p < 0.001). Regression analysis of the cumulative area under the gametocyte time curve found that the RR of a zero gametocyte density following treatment was 44-fold higher (95% CI: 13–148) after AL than SP, representing a 97.7% relative reduction in gametocyte prevalence (p < 0.001). In addition, AL was associated with a 95% decrease in gametocyte density among those carrying gametocytes (incidence rate ratio = 0.050; 95% CI: 0.009–0.274; p = 0.001).
Figure 4 Scatter Plot of Individual Patient Gametocyte Densities
The gametocyte densities (per microlitre) are given over time following treatment with SP (2000) and AL (2002).
Safety Monitoring and Adverse Events
Of 17 adverse events reported by patients recruited into the in vivo study, four were considered probably related to AL, but none of these was considered serious. A case of urticaria on day 1 resolved with administration of an antihistamine (without discontinuation of AL). There were two cases of vomiting and one possible treatment failure on day 28. Four patients demonstrated a sustained decrease in haemoglobin (Hb), although this was considered of potential clinical significance in only one case (Hb = 93 g/l on day 0, Hb = 70 g/l on day 7, and Hb = 74 g/l on day 42). The Hb level remained above 100 g/l, despite a decrease of at least 20 g/l, in the other three patients. Mouth ulcers that affected three patients after AL, and pre-existing vomiting that persisted in two cases after AL, were considered possibly related to treatment. The remaining adverse events were considered more likely due to malaria (body pains, n = 2) or inter-current illness (chest pain which resolved with amoxicillin, n = 1; ear infection, n = 1).
From April to June 2002, when approximately 310 blister packs of 24 tablets were dispensed in this hospital's catchment area, a total of 44 patients admitted to Manguzi hospital met the inclusion criteria of the intensive hospital-based safety monitoring study. Two of these patients did not provide consent and were thus excluded from the study. The remaining 42 patients, of whom 12 (29%) were under 2 years of age, were admitted either for malaria (n = 36), or for persistent (n = 5) or new (n = 1) symptoms following antimalarial treatment.
A total of 13 patients were treated with AL before (n = 5) or during (n = 8) their hospital stay. The five patients who had recently been treated with AL were admitted because of a poor response to AL, although it was not confirmed whether or not these patients had adhered fully to this treatment. These would not be considered adverse drug reactions according to International Committee for Harmonisation criteria. One 14-y-old female patient died. She had apparently received AL 3–4 d prior to her death, and was then treated with an unidentified herbal mixture by a traditional healer when she did not improve. She was admitted with a diagnosis of possible cerebral malaria with herbal intoxication and died within 24 h of admission. On admission her malaria smear was negative and a cerebrospinal fluid examination confirmed bacterial meningitis; the multidisciplinary expert review team concluded that her death was unlikely to be related to AL. One 24-y-old female patient treated during hospital stay with AL and doxycycline experienced vomiting after receiving the first and subsequent dose; this was considered a possible adverse reaction to AL and/or doxycycline.
Community Perspectives on Malaria Treatment Seeking and Adherence
Two of the 439 households selected for the household survey elected not to participate, and one respondent was unable to complete the interview due to illness (data from the portion of her interview completed was included in this analysis). Data were collected on a total of 2,506 household members of whom 55% were female. The average estimated distance from study households to the nearest public healthcare facility was 6.5 km, and this took an average of 90 min to reach. Overall 68% of household members had previously suffered from malaria.
Respondents reported that 239 (10%) household members had suffered from malaria in the 4 wk immediately preceding the household survey, which was conducted during February and March 2001 at the peak of the malaria transmission season, and 4 to 6 wk after the implementation of AL. Of these, only four (1.7%) reported not seeking treatment outside the home. Most had first sought treatment at either a Malaria Control Programme field camp (n = 101; 42%) or a public sector clinic (n = 127; 53%). Five patients reported initially seeking treatment at a public hospital, whereas one patient sought treatment from a private doctor, and one indicated first going to a traditional healer. Respondents reported that 226 (96%) had completed recent treatment (i.e., taken all prescribed tablets) but were uncertain whether a further seven (3%) household members had completed treatment. These responses were consistent with those regarding whether there was any antimalarial treatment remaining at home, with only two patients (1%) admitting that they still had antimalarial tablets (both AL) remaining from the recent treatment course.
Amongst household members that reported never having had malaria (“malaleveva”), 26% (204/785) reported experiencing fever (“umkhuhlane”) in the 4 wk prior to the household survey. Only 88 (43%) had sought care outside the home. Of those seeking fever treatment, 79 (90%) had presented to a public sector clinic, five at a public sector hospital, one to a private doctor, two to a Malaria Control Programme field camp, and one to a traditional healer. A significantly greater proportion of patients with recent perceived malaria (235/239; 98%) than those with a recent febrile disease not considered to be malaria (88/204; 43%) sought treatment at health-care facilities (RR: 2.28; CI: 1.95–2.67; p < 0.001).
Participants in all four FGDs agreed that if home treatment was not successful in rapidly alleviating malaria symptoms, the affected household member would be taken to the nearest public health-care facility. The FGDs revealed that people generally first attempted treatment of “just” fever with herbal medicines and enemas. Participants mentioned weakness and headache more often than fever as a feature they associate with malaria; participants commented that malaria, unlike “just” fever, was not self limiting. When asked about sources of treatment for malaria fever, all FGDs emphasized that there was no effective alternative to public health-care facilities and that malaria treatment should be sought urgently: “There's nothing else, you rush to the clinic” and “I have never heard of someone who cleared malaria at home, without going to the clinic.” Delaying treatment seeking was considered potentially fatal: “Once you don't go to the clinic, it kills…. You will die from malaria” and “Once you waste time at the sangoma [traditional healer], you die.”
The change in treatment policy from SP to AL appeared well accepted, because SP was regarded as ineffective: “These tablets (SP) which were used before the present ones [AL]—most of us didn't like taking them.…You would feel as if the malaria has become more severe,” “Every year I used to have malaria. This year I heard that new pills were coming. I was very sick with malaria. They gave me the [new] pills…. The [next] morning I was very fine…. I have cultivated and harvested and I've never had malaria [again] this year.” Although FGD participants generally reported completing full malaria treatment courses (“We finish the malaria treatment because we have seen that it [malaria] finishes [kills] us”), a few individuals indicated terminating treatment once symptoms resolved and saving remaining medication for future use.
Discussion
Reversing the alarming increase in malaria associated mortality in Africa is possible with use of existing effective methods of vector control and widespread use of highly effective ACT. KwaZulu–Natal Province was experiencing an epidemic of malaria fuelled by re-emergence of an insecticide-resistant mosquito vector and spiralling resistance to SP. There was a dramatic response to strengthening the vector control programme and wide-scale implementation of an ACT, AL, in this low-intensity malaria transmission setting; the number of malaria cases fell by 99%. The implementation of these dual interventions in KwaZulu–Natal was found to be cost effective and resulted in substantial cost savings [33].
Malaria transmission is multifactorial: In KwaZulu–Natal, where malaria control operations are intense, exploration of monthly malaria case data between 1971 and 2001 found that climatic factors only influenced interannual variation in malaria transmission but not the medium to long term trends in case totals, which were significantly associated only with antimalarial resistance and HIV prevalence [34,35]. Despite the continued increase in antenatal HIV seroprevalence in KwaZulu–Natal from 32.5% in 1999 to 40.7% in 2004 (http://www.doh.gov.za/aids/index, we observed a 99% reduction in confirmed malaria cases. However, HIV coinfection may have precluded a greater reduction in malaria-related admissions and deaths because this coinfection has been shown to increase the risk of severe malaria and malaria-related deaths in the non-immune KwaZulu–Natal population [36]. There were no other substantial social, political, and health-care changes likely to affect malaria burden during the study period, although small effects cannot be excluded.
Notification of communicable diseases is constrained by under-reporting and this is particularly severe during epidemics in resource poor areas, when health workers probably prioritise patient management over notification [37]. The increased demand placed on health-care providers by the large-scale malaria epidemic in KwaZulu–Natal Province that peaked in 2000 suggests that under-notification may have been particularly prevalent during that year. It is expected that any delay in changing the SP malaria treatment policy and in strengthening vector control would have resulted in increased malaria transmission and further increases in malaria morbidity and mortality. Thus the remarkable reduction in malaria case notification documented here is probably an underestimate of the true effect of implementing effective treatment and improving vector control.
The decreases in malaria morbidity and mortality observed in KwaZulu–Natal Province reflect both the therapeutic effects of AL and the enhanced malaria vector control (following the reintroduction of DDT and extension of IRS to neighbouring southern Mozambique). The reliable and rapid therapeutic response to ACTs, combined with their effect of reducing gametocyte carriage, make them ideal treatments. The dramatic reductions in gametocyte prevalence and density observed would substantially contribute to reducing malaria transmission [11]. The relative contributions of the two strategies, deliberately introduced in short succession to optimise public health impact, cannot be accurately apportioned. The extent of the public health impact observed following the policy change to ACTs is unlikely to have been achieved in the absence of effective vector control. Equally, we believe that the benefits of reintroducing DDT for IRS of traditional (but not western-style) structures would have had less benefit without the replacement of highly ineffective SP with an effective acceptable ACT, given the 89% treatment failures and high gametocyte carriage rates observed following SP monotherapy.
On the northwestern border area of Thailand, a 47% reduction in the incidence of P. falciparum infections was observed in the 12 mo following the introduction of artesunate plus mefloquine, when no changes were made to vector control [38]. This improved further to a 6-fold reduction over a 10-y period of use [7]. The similar experience of marked public health benefits in northwestern Thailand suggests that the results in KwaZulu–Natal Province reflect the benefits of effective ACT, rather than being specific to AL. The area studied on the western border of Thailand is similar to KwaZulu–Natal in terms of a low intensity of seasonal malaria transmission (annual entomologic inoculation rate < 1) and reasonably high levels of access to health-care facilities providing definitive diagnosis of malaria and relatively reliable, well-regulated drug supply. The reduction in malaria mortality by ACTs reflects both the reduced incidence of malaria following decreased gametocyte carriage, and the reliable and rapid antimalarial activity of the ACT. General deployment of artemisinin in Vietnam was also associated with a marked and sustained reduction of malaria mortality [39].
The improvement in clinical and parasitological cure rate from 11% to 99% is particularly important in KwaZulu–Natal, because the low intensity of malaria transmission and, consequently, the low levels of acquired immunity mean that a substantial proportion of patients with parasitological failure would develop recrudescent P. falciparum infections or even progress to severe malaria. The high AL cure rate, similar to the 97%–98% published recently in large randomised controlled trials reported from East Africa [40,41], was achieved despite co-adminstration at the clinic with a relatively low-fat content drink, which had been selected for its widespread availability. Previous studies have shown that lumefantrine absorption is highly dependent on coadministration with fat, although the minimum fat content required for adequate absorption has not yet been defined [42].
Similarly to previous studies [43,44], we documented that AL was well tolerated both in the sentinel surveillance programme at Manguzi hospital, and in patients enrolled in the in vivo study who were closely monitored. This probably contributed to the generally good adherence reported. However, the markedly decreased malaria incidence in KwaZulu–Natal Province limited our ability to detect uncommon serious adverse effects of AL.
In South Africa, orthodox medicines are relatively strictly regulated, and AL treatment requires prescription by a registered health-care provider. ACT coverage depends on treatment seeking at health-care facilities where AL is available. Household survey results demonstrated that the majority (97%) of people with recent symptoms thought to result from malaria initially sought treatment at public health-care facilities. For recent fevers not considered by respondents to be related to malaria, however, seeking treatment outside the home was significantly less frequent, although the public clinics remained the most popular source of health care. These malaria-specific treatment-seeking patterns are likely to have been enforced by the regular communication between the provincial Malaria Control Programme and the community. The general preference for public health-care facilities may be associated with the provision of free health care for notifiable diseases in public sector clinics and a relatively high level of access to these facilities, with 81% of the population living within 10 km of a public clinic in northern KwaZulu–Natal [21]. This pattern of treatment-seeking behaviour was confirmed during the FGDs, in which fear of malaria and perceived effectiveness of treatment served as key factors motivating patients to seek treatment from public health-care facilities. High levels of AL coverage are thus achievable in northern KwaZulu–Natal through public-sector implementation alone, provided that patients recognise malaria symptoms. Because ACT coverage is considered an important determinant of community benefit in terms of reducing malaria transmission and slowing antimalarial resistance, ensuring ready access to this treatment (and definitive diagnosis) is a key component of effective malaria control. This cannot be achieved without addressing local health-care infrastructure needs and ensuring a high level of health literacy in the local community.
Adherence with prescribed antimalarial treatment regimens is essential to optimise cure rates and prevent resistance. There were some inconsistencies between household survey and FGD findings on patient adherence, with a few FGD participants indicating that they would stop treatment once symptoms resolved and save medication for future episodes. These FGD findings confirm, but cannot quantify, the general trend that self-reporting tends to overestimate levels of adherence, because patients who report poor adherence are generally accurate, whereas those who deny poor adherence may not be [45]. Although the consistently high levels of adherence reported in the answers to three different questions, and the dramatic reductions in malaria morbidity and mortality are reassuring findings, it has been suggested that the public health consequences of partial adherence are delayed until cure rates start to decline [46]. There is a clear need for better measures of adherence in general and, for AL in particular, for ongoing efforts by health staff to encourage patients to complete all six doses and to co-administer it with a fat-containing drink.
There are a number of factors to be considered before assuming direct generalisability of our findings to other countries, including: whether the current malaria treatment policy is already highly effective, whether baseline resistance to the non-artemisinin component of the ACT is higher, whether home treatment and treatment seeking in the informal or private sector is more prevalent, whether malaria diagnosis is clinical or confirmed, and perhaps most importantly, the intensity of malaria transmission intensity. Recent data suggest fewer than half of the populations at risk of malaria globally live in areas of high-intensity malaria transmission, with 25.4% of those at risk living in low-intensity transmission areas and a further 31.3% living in areas of moderate-intensity malaria transmission [47].
Decreasing malaria morbidity and mortality substantially in high-transmission areas, as in much of sub-Saharan Africa, is expected to be considerably more challenging. The higher risk of new malaria infections could impact on the overall effectiveness of ACTs, and ACT coverage and compliance may be lower in semi-immune individuals because a substantial proportion of infections would be asymptomatic and symptoms would more likely resolve with incomplete treatment. There are fewer examples of successful large-scale vector control programmes in high-intensity transmission areas.
Between the 1940s and 1960s, pilot malaria eradication projects across sub-Saharan Africa recorded significant reductions in malaria [48]. Subsequently, systematic indoor residual insecticide spraying programmes have been highly successful in reducing malaria transmission, particularly in southern Africa and island states [49]. Similarly, widespread coverage with insecticide-treated bed nets in areas of high-intensity malaria transmission has resulted in sustained protection of both individuals and communities against malaria [50]. These findings suggest that effective sustainable vector control may be achievable across all levels of malaria endemicity and could limit the number of malaria cases requiring ACT treatment, particularly if malaria is definitively diagnosed. This would increase the affordability of ACTs and adequacy of ACT supply, and be expected to optimise the effect of ACTs on malaria transmission.
Conclusions
We consider the ready access to treatment in a relatively well-developed rural primary health-care infrastructure, coupled with an effective vector control programme, important factors for deriving the greatest benefit from ACT implementation. Equally important are the strong community perceptions that malaria diagnosis and treatment should be sought urgently at public health-care facilities and treatment then completed. These factors deserve consideration by those responsible for reducing malaria morbidity and mortality, because wide-scale implementation and rational use of effective vector control and ACT are cornerstones of combating the enormous health and economic burden of malaria.
Patient Summary
Background
Malaria is caused by a parasite transmitted by some types of mosquito; it kills about a million people every year, especially children in Africa. The disease has become more common in recent years because the parasites have become resistant to many malaria drugs and the mosquitoes have developed resistance to insecticides. Between 1995 and 2000, malaria increased dramatically in South Africa's KwaZulu–Natal province. The main reasons for the increase are believed to be resistance to the drug sulfadoxine-pyrimethamine (SP) and to pyrethroid types of insecticide. In 2000, in response, a new drug called artemether-lumefantrine (AL) was introduced to treat people with malaria. (AL is a combination drug that includes artemisinin, which has been shown to work in control programs in Asia, but this the first time a drug of this type has been used in a program in Africa.) Mosquito control efforts were also increased and, although pyrethroids were still used in some situations, the older insecticide DDT was also reintroduced.
What Did the Researchers Do and Find?
They measured the success of the action taken in KwaZulu–Natal. In the year following the changes, hospital admissions for malaria declined by 89% and so did deaths; outpatient cases decreased by 85%. By 2003, outpatient cases and admissions had both fallen by 99% and deaths had decreased by 97%. AL treatment cured 99% of cases, compared with the 11% cure rate previously found with SP. Most patients (96%) said they completed the six-dose course of AL, and no serious drug side effects were reported.
What Does This Mean?
The study shows an encouraging and important example of how malaria can be effectively fought. Because both better mosquito control and better drugs were introduced around the same time, the researchers could not say how much each of the two measures contributed to the overall success. However, it is likely that both contributed, and it shows that the two together can be very effective. When applying the lessons from this success story to other parts of Africa and to other continents, one needs to keep in mind that the starting conditions in KwaZulu–Natal were quite favorable. For example, there is a low level of background immunity against malaria, which means that infected people usually get quite sick and seek treatment rather than acting as reservoirs for further transmission. In addition, most patients can get prompt diagnosis and treatment because the province has a reasonable health-care infrastructure.
Where Can I Get More Information?
The Roll Back Malaria Department of the World Health Organization has a Web site with information about the disease and the global efforts to fight it (this site also has links to other organizations and to useful publications):
http://www.who.int/malaria
The South African Department of Health Web site includes guidelines on the treatment of malaria (2002) and prevention of malaria (2003):
http://www.doh.gov.za/docs/index.html
The authors gratefully acknowledge the staff and patients of Ndumo clinic and Mosvold, Manguzi, and Bethesda hospitals who participated in this study, and particularly the respective medical superintendents Hervey Vaughan Williams, Etienne Immelman, and Andrew Grant for providing records of malaria morbidity and mortality at each sentinel health-care facility. Bheki Qwabe and David Mthembu assisted in the collection of data in the in vivo studies. Chris le Cock, KwaZulu–Natal Department of Health (DOH), and Nicros Mngomezulu, Mpumalanga DOH, both conducted quality assurance checks on all malaria smears taken in the in vivo study. Charlotte Muheki, University of Cape Town Health Economics Unit, assisted in drafting the questionnaire and training of interviewers for the household survey. Karen Daniels and Judy Dick, Medical Research Council, provided technical support in the development of the methodology and analysis of the household survey and focus group discussions. Bernice Harris and Stefano Fieremans (Mpumalanga DOH), Lucille Blumberg (National Institute of Communicable Diseases) and Isabela Ribeiro (WHO) participated in the review of serious adverse events. NJ White is supported by the Wellcome Trust as part of the Wellcome Trust-Mahidol University-Oxford Tropical Medicine Research Program. The South East African Combination Antimalarial Therapy [SEACAT] evaluation, within which this study was nested, received core financial support from the United Nations Development Programme/ World Bank/WHO Special Programme for Research and Training in Tropical Diseases (WHO TDR). The funders had no role in data collection and analysis, decision to publish, or preparation of the manuscript. Technical support was provided by Piero Olliaro (WHO TDR) in the design of this study and by Francois Nosten throughout the policy change in KwaZulu–Natal; both critically reviewed this manuscript.
Author Contributions: KIB is the principal investigator of the South East African Combination Antimalarial Therapy (SEACAT) evaluation, within which this study was nested. KIB conceptualised this study and analysed the data, drafted this manuscript, and together with DND, NJW, and BLS designed and interpreted results of all study components. FL supervised the statistical analyses performed. AJ managed data entry and together with SSD conducted preliminary analysis and interpretation of the household survey. UM conducted preliminary analysis and interpretation of sentinel safety data. JT conducted preliminary analysis and interpretation of sentinel hospital morbidity and mortality data. EA and BB assisted in the design, and together with DJM, assisted in the coordination of the in vivo study. All authors revised manuscript critically for substantial intellectual content and have access to all data in the studies reported.
Citation: Barnes KI, Durrheim DN, Little F, Jackson A, Mehta U, et al. (2005) Effect of artemether-lumefantrine policy and improved vector control on malaria burden in KwaZulu–Natal, South Africa. PLoS Med 2(11): e330.
Abbreviations
ACTartemisinin-based combination therapy
ALartemether-lumefantrine
CIconfidence interval
CQchloroquine
DDTdichlorodiphenyltrichloroethane
FGDfocus group discussion
Hbhaemoglobin
IQRinterquartile range
IRSindoor residual spraying
RRrelative risk
SPsulfadoxine-pyrimethamine
WHOWorld Health Organization
==== Refs
References
Trape JF Pison G Preziosi MP Desgrees du Lou A Delaunay V Impact of chloroquine resistance on malaria mortality C R Acad Sci III 1998 321 689 697 9769862
Snow W Trape J Marsh K The past, present and future of childhood malaria mortality in Africa Trends in Parasitol 2001 17 593 597
Hargreaves K Koekemoer LL Brooke BD Runt RH Mthembu J
Anopholes funestus resistant to pyrethroid insecticides in South Africa Med Vet Entomol 2000 14 181 189 10872862
Vaughan-Williams CH Success of insecticide spraying in controlling malaria [letter] S Afr Med J 2003 93 160
Bredenkamp BLF Sharp BL Mthembu SD Durrheim DN Barnes KI Failure of sulfadoxine-pyrimethamine in treating Plasmodium falciparum malaria in KwaZulu–Natal S Afr Med J 2001 91 970 972 11847918
Maharaj R Mthembu DJ Sharp BL The impact of DDT reintroduction on malaria transmission in KwaZulu Natal, South Africa S Afr Med J 2005 In press
Nosten F van Vugt M Price R Luxemburger C Thway KL Effects of artesunate mefloquine combination on incidence of Plasmodium falciparum malaria and mefloquine resistance in Western Thailand: A prospective study Lancet 2000 356 297 302 11071185
White NJ Nosten F Looareesuwan S Watkins WM Marsh K Averting a malaria disaster Lancet 1999 353 1965 1967 10371589
Adjuik M Agnamey P Babiker A Borrmann S Brasseur P Amodiaquine-artesunate versus amodiaquine for uncomplicated Plasmodium falciparum malaria in African children: A randomised, multicentre trial Lancet 2002 359 1365 1372 11978332
International Artemisinin Study Group Artesunate combinations for treatment of malaria: Meta-analysis Lancet 2004 363 9 17 14723987
Sutherland CJ Ord R Jawara M Dunyo S Drakeley CJ A six-dose regimen of co-artemether prevents Gambian children with falciparum malaria from becoming infectious to Anopheles mosquitoes PLoS Med 2005 2 4 e92 15839740
Freese JA Sharp BL Rossouw EJ Gouws E Fay SA The in vitro sensitivity of southern African isolates of Plasmodium falciparum to amodiaquine, chloroquine, mefloquine, quinine and sulphadoxine/pyrimethamine S Afr J Sci 1994 90 417 420
Roll Back Malaria 2004 Roll Back Malaria (2004) Facts on ACTs An update on recent progress in policy and access to treatment Available: http://www.rbm.who.int/cmc_upload/0/000/015/364/RBMInfosheet_9.htm Accessed 17 February 2005
Dorsey G Njama D Kamya MR Cattamanchi A Kyabayinze D Sulfadoxine/pyrimethamine alone or with amodiaquine or artesunate for treatment of uncomplicated malaria: A longitudinal randomised trial Lancet 2002 360 2031 2038 12504399
Bloland PB A contrarian view of malaria therapy policy in Africa Am J Trop Med Hyg 2003 68 125 126 12641398
Duffy PE Mutabingwa TK Drug combinations for malaria: Time to ACT? Lancet 2004 363 3 4 14723982
Bloland PB Ettling M Meek S Combination therapy for malaria in Africa: Hype or hope? Bull World Health Organ 2000 78 1378 1388 11196485
Panosian CB Economic access to effective drugs for falciparum malaria Clin Infect Dis 2005 40 713 717 15714418
Roll Back Malaria 2005 Changing malaria treatment policy to artemisinin-based combinations An implementation guide Available: http://rbm.who.int/rbm/Attachment/20050418/malariaTreatmentPolicyMarch2005.pdf , Accessed 11 September 2005
Newton PN Dondorp A Green M Mayxay M White NJ Counterfeit artesunate antimalarials in southeast Asia Lancet 2005 362 169
Martin C Curtis B Fraser C Sharp B The use of a GIS–based malaria information system for malaria research and control in South Africa Health Place 2002 8 227 236 12399212
Sharp BL le Sueur D Malaria in South Africa—The past, the present and selected implications for the future S Afr Med J 1996 86 83 89 8685790
Sharp BL le Sueur D Behavioural variation of Anopheles arabiensis (Diptera: Culicidae) populations in Natal, South Africa Bull Entomol Res 1991 81 107 110
Baker L Barnes K Baker L, Barnes K (2001) New antimalarial treatment in KwaZulu Natal S Afr Med J 2001 91 358 359 11455793
World Health Organization Severe falciparum malaria. World Health Organization, Communicable Diseases Cluster Trans R Soc Trop Med Hyg 2000 94 Supp 1 62
World Health Organization Assessment and monitoring of antimalarial drug efficacy for the treatment of uncomplicated falciparum malaria WHO/HTM/RBM/2003.50 Geneva, Switzerland World Health Organization Available: http://www.cdc.gov/malaria/pdf/WHO2003_monitoring.pdf . Accessed 7 September 2005
Farnert A Arez AP Babiker HA Beck HP Benito A Genotyping of Plasmodium falciparum infections by PCR: A comparative multicentre study Trans R Soc Trop Med Hyg 2001 95 225 232 11355566
Medicines Control Council General regulations made in terms of the medicines and related substances act, 1965 (Act No. 101 of 1965), as amended 2003 Available: http://www.mccza.com/showdocument.asp?Cat=27&Desc=Acts%20and%20Regulations . Accessed 11 September 2005
ICH E2A: Clinical Data Safety Management Definitions and Standards for Expedited Reporting 1995 Available: http://www.emea.eu.int/pdfs/human/ich/394503en.pdf . Accessed 11 September 2005
Greene WH Accounting for excess zeros and sample selection in Poisson and negative binomial regression models. Working paper no 1994 EC-94–10, New York New York University, Department of Economics, Stern School of Business
Strauss A Corbin J Denzin NK Lincoln YS Grounded theory methodology. An overview Handbook of qualitative research 1994 Thousand Oaks (California) Sage Publications 273 285
Maxwell JA Bickman L Rogers DJ Designing a qualitative study Handbook of applied social research methods 1998 Thousand Oaks (California) Sage Publications 69 100
Muheki C McIntyre D Barnes KI Artemisinin-based combination therapy reduces expenditure on malaria treatment in KwaZulu Natal, South Africa Trop Med Int Health 2004 9 959 966 15361108
Craig MH Kleinschmidt I Nawn JB Le Sueur D Sharp BL Exploring 30 years of malaria case data in KwaZulu Natal, South Africa: Part I: The impact of climatic factors Trop Med Int Health 2004 9 1247 1257 15598256
Craig MH Kleinschmidt I Le Sueur D Sharp BL Exploring 30 years of malaria case data in KwaZulu–Natal, South Africa: Part II. The impact of non-climatic factors Trop Med Int Health 2004 9 1258 1266 15598257
Grimwade K French N Mbatha DD Zungu DD Dedicoat M HIV infection as a cofactor for severe falciparum malaria in adults living in a region of unstable malaria transmission in South Africa AIDS 2004 18 547 554 15090809
Durrheim DN Knight S Notification—Completing the cycle S Afr Med J 1996 86 1434 1435
Price RN Nosten F Luxemburger C ter Kuile FO Paiphun L Effects of artemisinin derivatives on malaria transmissibility Lancet 1996 347 1654 1658 8642959
White NJ Cook GC Zumla A Malaria Elsevier Science 2003 Edinburgh Sage Publications 1243
Piola P Fogg C Bajunirwe F Biraro S Grandesso F Supervised versus unsupervised intake of six-dose artemether-lumefantrine for treatment of acute, uncomplicated Plasmodium falciparum malaria in Mbarara, Uganda: A randomised trial Lancet 2005 365 1467 1473 15850630
Mutabingwa TK Anthony D Heller A Hallett R Ahmed J Amodiaquine alone, amodiaquine+sulfadoxine-pyrimethamine, amodiaquine+artesunate, and artemether-lumefantrine for outpatient treatment of malaria in Tanzanian children: A four-arm randomised effectiveness trial Lancet 2005 365 1474 1480 15850631
White NJ van Vugt M Ezzet F Clinical pharmacokinetics and pharmacodynamics of artemether-lumefantrine Clin Pharmacokinet 1999 37 105 125 10496300
Ribeiro IR Olliaro P Safety of artemisinin and its derivatives. A review of published and unpublished clinical trials Med Trop (Mars) 1998 58 50 53 10212898
van Vugt M Looareesuwan S Wilairatana P McGready R Villegas L Artemether-lumefantrine for the treatment of multidrug-resistant falciparum malaria Trans R Soc Trop Med Hyg 2000 94 545 548 11132386
Farmer KC Methods for measuring and monitoring medication regimen adherence in clinical trials and clinical practice Clin Ther 1999 21 1074 1090 10440628
Yeung S White NJ How do patients use antimalarial drugs? A review of the evidence Trop Med Int Health 2005 10 121 138 15679555
Hay SI Guerra CA Tatem AJ Noor AM Snow RW The global distribution and population at risk of malaria: Past, present and future Lancet Inf Dis 2004 4 327 336
Kouznetsov RL Malaria control by application of indoor spraying of residual insecticides in tropical Africa and its impact on community health Trop Doctor 1977 7 81 91
Mabaso ML Sharp B Lengeler C Historical review of malaria in southern Africa with emphasis on the use of indoor residual house spraying Trop Med Int Health 2004 9 846 856 15303988
Lengeler C Insecticide-treated bed nets and curtains for preventing malaria 2004 Cochrane Database of Syst Rev CD000363
|
16187798
|
PMC1240068
|
CC BY
|
2021-01-05 10:39:19
|
no
|
PLoS Med. 2005 Nov 4; 2(11):e330
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020330
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020371SynopsisInfectious DiseasesClinical PharmacologyEpidemiology/Public HealthHealth PolicyParasitologyHealth PolicyInfectious DiseasesMalariaPublic HealthKwaZulu–Natal's Successful Fight against Malaria Synopsis11 2005 4 10 2005 2 11 e371Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Public Domain declaration, which stipulates that, once placed in the public domain, this work may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose.
Effect of Artemether-Lumefantrine Policy and Improved Vector Control on Malaria Burden in KwaZulu-Natal, South Africa
Rolling Back a Malaria Epidemic in South Africa
==== Body
The resurgence of malaria remains a major global concern. Artemisinin-based drugs are increasingly seen as one of the best hopes for, at last, making progress in the battle against malaria. Trials of artemisinin-based combination therapy (ACT) in control programs in Southeast Asia have been very encouraging. However, we need to know whether similar levels of effectiveness are achievable in Africa, where the majority of the world's cases of malaria are found.
One part of Africa that has seen increases both in the number of malaria cases and in drug resistance is South Africa's KwaZulu–Natal province. The rise in malaria in this area has been dramatic, with a 15–fold increase in cases taking place during the 1990s. Control efforts during this period involved mosquito control with pyrethroid insecticides (which had replaced DDT) and sulfadoxine-pyrimethamine (SP) as a first-line treatment. (SP was introduced in 1988 in response to high levels of chloroquine resistance.)
In the year 2000, new measures were put in place to address KwaZulu–Natal's malaria crisis. The key elements of this new strategy were the introduction of an ACT drug, artemether-lumefantrine (AL), and an intensification of mosquito control efforts. While pyrethroids were retained for indoor residual spraying of western-style structures, DDT was also reintroduced for spraying traditional homesteads. Karen Barnes and colleagues now present the first comprehensive description and evaluation of the program.
The researchers reviewed four years of malaria morbidity and mortality data at four representative health-care facilities within KwaZulu–Natal's malaria-endemic area. They found that, in the year following improved vector control and implementation of AL treatment, malaria-related admissions and deaths declined by 89%, and outpatient visits decreased by 85%. By 2003, malaria-related outpatient cases and admissions had fallen by 99%, and malaria-related deaths had decreased by 97%. There was a marked and sustained decline in malaria throughout the province. AL cured 99% of those study patients who were followed up for 42 days. Consistent with the findings of focus group discussions, a household survey found that self-reported adherence to the six-dose AL regimen was 96%. Two surveys in subsets of patients receiving AL revealed no serious adverse events resulting from the treatment.
These are impressive results, but they are not solely due to the introduction of ACT. As the authors say, “the ready access to treatment in a relatively well-developed rural primary health-care infrastructure, coupled with an effective vector control programme [are] important factors for deriving the greatest benefit from ACT implementation. Equally important are the strong community perceptions that malaria diagnosis and treatment should be sought urgently at public health-care facilities and treatment then completed.”
Artemisinin is extracted from Artemisia annua, the annual wormwood (Photo: Scott Bauer)
In an accompanying Perspective (DOI: 10.1371/journal.pmed.0020368), Patrick Duffy and Theonest Mutabingwa highlight lessons learned from the “notable success” in KwaZulu–Natal “amid the dire statistics showing a deadly resurgence of malaria.” They also discuss how economic and noneconomic conditions in other parts of sub-Saharan Africa differ from KwaZulu–Natal in ways that are likely to affect the influence of ACT.
|
0
|
PMC1240069
|
CC0
|
2021-01-05 10:39:45
|
no
|
PLoS Med. 2005 Nov 4; 2(11):e371
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020371
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020373SynopsisCancer BiologyGenetics/Genomics/Gene TherapyOncologyOncologyCasting Doubt on the Role of Mitochondria in Tumorigenesis Synopsis11 2005 4 10 2005 2 11 e373Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Public Domain declaration, which stipulates that, once placed in the public domain, this work may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose.
A Critical Reassessment of the Role of Mitochondria in Tumorigenesis
==== Body
Mitochondrial DNA (mtDNA) has been intensively studied over the past two decades, and point mutations, more commonly known as deletions, of this DNA are known to be involved in several syndromes. Unlike nuclear DNA, with 46 chromosomes, half from each parent, mtDNA is just one piece of genome of which there are many copies, but all copies come only from the mother. Mitochondrial disease syndromes, such as MELAS, have a range of different clinical manifestations depending on how many copies of the abnormal mtDNA are present in affected cells.
There are several international resources of mtDNA sequences. From these sequences, it has been possible to show that different population groups have different patterns of substitutions in the mtDNA—so-called haplogroups; this information has been used, for example, in the investigation of the origin and migration patterns of human populations, and some investigators have even suggested that it could be used to trace back to earliest human history the founding mothers of humanity.
More recently, however, attention has turned to the question of whether mtDNA is involved in tumor formation. However, deciding whether mutations are harmful or innocuous has been difficult. One concern is that isolation of mtDNA from any tissue is not simple, and may be particularly difficult from tumor samples, which are often contaminated with exogenous DNA.
In a hard-hitting paper, Antonio Salas and colleagues cast more doubt over a causal role for mtDNA alterations in tumors; they have now reassessed many of the studies that have examined the role of mtDNA in tumorigenesis, and concluded that much of the data are at best questionable. The group used a phylogenetic approach to analyze the reported work, which compared the sequences under consideration with the current database of complete sequences of mtDNA. The authors believe that such an approach is essential to look at the overall picture of the mtDNA rather than assessing each substitution in the mtDNA independently.
Salas and colleagues conclude that more than 80% of published mtDNA sequencing studies contain obvious errors, and that many of the published results that implicate mutations in tumorigenesis are in fact part of normal population variability, and their presence must be due to contamination of the tumor sample. Salas does not hold back in his criticisms. He states that the result of such sequencing “disasters” is that flawed results are not filtered out from the clinical literature, which makes the task of interpreting the role of mtDNA in the tumor process very difficult.
The Role of Mitochondria in Tumorigenesis is not yet clear
The researchers draw analogies to the time when ancient DNA sequencing was beginning, and many contaminated samples were amplified and claimed to yield prehistoric DNA. They urge clinical geneticists to use checks similar to those proposed to assess authenticity for ancient DNA studies: special care should be taken in sequencing and documentation, raw sequence data should be made fully accessible to referees and readers, and the complete record of data from the population genetics field should be used to put the results in context.
The group's findings will undoubtedly cause much debate in this research community. With their conclusion that “there is no precedent that we know of in the genetics literature for such a high number of flawed papers (most of them published in high-rank journals), which affect a whole subfield of clinical research,” they urge reconsidering the role of mtDNA in tumorigenesis.
|
0
|
PMC1240070
|
CC0
|
2021-01-05 10:39:17
|
no
|
PLoS Med. 2005 Nov 4; 2(11):e373
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020373
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020374SynopsisCancer BiologyOncologyOncologyCancer: LungEGFR Mutations and Lung Cancer Synopsis11 2005 4 10 2005 2 11 e374Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Public Domain declaration, which stipulates that, once placed in the public domain, this work may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose.
Oncogenic Transformation by Inhibitor-Sensitive and -Resistant EGFR Mutants
==== Body
Tyrosine kinases of the epidermal growth factor receptor (EGFR) family are frequently mutated in human cancers. Mutations in the tyrosine kinase domain of EGFR (encoded by exons 18–24) have mostly been found in lung cancers. Some, but not all, lung cancers carrying such mutations are responsive to treatment with small-molecule EGFR inhibitors, including the two FDA-approved drugs erlotinib and gefitinib.
Matthew Meyerson and colleagues undertook a systematic study of the different classes of EGFR mutations found in lung cancers in order to understand their roles in tumorigenesis on one hand, and their relation to drug sensitivity on the other. They introduced different altered EGFR versions into fibroblast and lung epithelial cells, and found that all of the mutant proteins transformed both cell types in an EGF-independent manner. Transformation was associated with constitutive kinase activity and with the activation of known downstream signaling pathways.
While the various mutant receptors had similar transforming capabilities, cells expressing them differed in their response to EGFR inhibitors. Transformation of cells expressing mutations in exons 18, 19, and 21 was inhibited by 100 nM erlotinib or gefitinib, whereas no significant inhibitory effect on cells expressing an exon 20 insertion mutation was seen even at much higher concentrations of either drug. This result is consistent with the lack of clinical responses to erlotinib or gefitinib in three lung cancer patients with exon 20 mutations. In contrast, when the researchers tested another experimental EGFR inhibitor called CL-387,785, they found cells expressing the exon 20 insertion mutation to be sensitive, consistent with previous studies that had found similar patterns with other EGFR exon 20 mutations.
Morphology of fibroblasts expressing mutant EGFR
These results highlight the problems and the possibilities of individualized cancer therapy. One drug is unlikely to fit all tumors, not even all tumors with mutations in a specific oncogene. On the other hand, having a collection of drugs against a particular target increases chances that one of them will prove effective, and that alternatives exist when tumors develop resistance. Developing such a collection and selecting the right drug for the right patient is a challenge not only scientifically but also economically.
|
0
|
PMC1240071
|
CC0
|
2021-01-05 10:39:17
|
no
|
PLoS Med. 2005 Nov 4; 2(11):e374
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020374
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1618773610.1371/journal.pmed.0020368PerspectivesInfectious DiseasesEpidemiology/Public HealthHealth PolicyInfectious DiseasesMalariaHealth PolicyPublic HealthRolling Back a Malaria Epidemic in South Africa PerspectiveDuffy Patrick E *Mutabingwa Theonest K Patrick Duffy is at the Seattle Biomedical Research Institute, Seattle, Washington, United States of America, and Walter Reed Army Institute of Research, Silver Spring, Maryland, United States of America. Theonest K. Mutabingwa is at the London School of Hygiene and tropical Medicine, London, United Kingdom, and the National Institute for Medical Research, Dar es Salaam, Tanzania.
Competing Interests: The authors declare that no competing interests exist.
*To whom correspondence should be addressed. E-mail: [email protected] 2005 4 10 2005 2 11 e368Copyright: © 2005 Duffy et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.
Effect of Artemether-Lumefantrine Policy and Improved Vector Control on Malaria Burden in KwaZulu-Natal, South Africa
KwaZulu-Natal's Successful Fight against Malaria
The authors discuss the success in malaria control in KwaZulu-Natal (reported by Barnes and colleagues), and its implications for the rest of Africa.
==== Body
A mid the dire statistics showing a deadly resurgence of malaria, a notable success has been scored in South Africa. In KwaZulu–Natal province, malaria cases increased from about 600 in 1991–1992 to more than 30,000 by 1999–2000 [1]. Then, after household spraying with DDT was implemented, and the new antimalarial combination artemether-lumefantrine (AL) was widely deployed (Figure 1), cases declined by more than 99% over the next three years. A paper in PLoS Medicine by Barnes et al. [2] examines the implementation and efficacy of AL during the KwaZulu–Natal crisis. They conclude that vector control and widespread use of artemisinin-based combination therapy (ACT) such as AL may confer similar benefits in other African countries. Could the adoption of these policies salvage the Roll Back Malaria Initiative that was formed in 1999 to halve malaria deaths by 2010 [3], but which was recently lamented as “dysfunctional” for its inaction in the face of rising malaria morbidity and mortality rates [4]?
Figure 1 Mosquito Control and ACT Are Both Likely Contributors to the Reduction of Malaria in KwaZulu–Natal
(Photo: Karen Barnes and Atis Muelenbachs)
Artemisinins to the Rescue
Artemisinin derivatives such as artemether have several advantages—they act rapidly, cause few side effects, and have not yet acquired resistant parasites [5]. Artemisinins also prevent parasite transmission by inactivating or killing gametocytes [6]. In northwestern Thailand, malaria incidence declined after the ACT artesunate-mefloquine was introduced, and its effectiveness has been sustained over several years, possibly due to gametocytocidal effects [7]. In a study recently published in PLoS Medicine [6], gametocytes from Gambian children treated with AL were less likely than those from children treated with a chloroquine and sulfadoxine-pyrimethamine (SP) combination to infect mosquitoes. These reports have led many to expect that ACTs will dramatically improve case management, and reduce malaria transmission in Africa.
KwaZulu–Natal—A Special Case?
Caution is warranted, however. KwaZulu–Natal is more similar to Thailand than to most sub-Saharan countries in ways that may affect the influence of ACTs. The economic strength of South Africa supported effective vector control measures and a health-care infrastructure that facilitated prompt diagnosis and treatment. Poorer sub-Saharan countries are unable to support similar programs in the absence of additional financial resources.
Noneconomic issues may also limit the effect of ACTs on malaria incidence and case management in Africa. Malaria transmission and, therefore, immunity are low in both Thailand and KwaZulu–Natal. Thais and South Africans typically get sick when they are infected, and seek treatment. In African countries where malaria transmission is high, semi-immune individuals often do not feel sick enough to seek treatment, and act as a reservoir for continued transmission. Additionally, African children often present with high-density parasitemia, making it more likely that parasites will be temporarily suppressed but then recrudesce after artemisinin therapy [8,9].
ACTs will improve treatment outcomes in areas of sub-Saharan Africa where resistant parasites have rendered the current first-line drugs nearly useless. Drug resistance has probably played a key role in the rising malaria mortality rates among African children [10], so ministries of health are optimistic that ACTs will reverse this awful trend. However, the long-term effectiveness of ACTs in high endemicity areas has not been proven, and many operational questions remain unanswered.
ACT Alternatives
Are ACTs the most effective new antimalarial combination? In the July issue of PLoS Medicine, Dorsey and colleagues reported that the nonartemisinin combination of SP and amodiaquine (AQ) was as effective or better (and cheaper) than the combination of artesunate and AQ for treating Ugandan children, when both recrudescent and new infections were considered [11]. Resistance to both SP and AQ is spreading in Africa, and this will limit the sustainability of the combination. Furthermore, because recrudescent parasites are more likely to be drug-resistant [12], and recrudescences were more common after SP-AQ, this combination may accelerate the spread of resistant parasites. Nevertheless, the combination should be considered as a short-term strategy in areas where the parasite remains sensitive. The results also caution that the benefits of ACTs may be limited in high endemicity areas unless reinfections are promptly treated.
Is AL the Best ACT for Africa?
AL is the only co-formulated ACT, which improves compliance. Furthermore, a dramatic rollback of malaria has been achieved in KwaZulu–Natal where AL was deployed. These have been strong factors in the selection of AL as first-line therapy by many African countries. However, the sharp decline in malaria in KwaZulu–Natal commenced after DDT spraying of households was initiated and before AL was deployed; therefore, the relative contribution of AL remains unclear. Sustained success with ACTs in Thailand has been achieved with artemether-mefloquine. The long-term effectiveness of AL remains unproven. The extended half-life of lumefantrine and the short half-life of artesunate mean that many reinfections in Africa will be exposed to lumefantrine alone, increasing the odds that resistant parasites will be selected. Worrying reports from Zanzibar suggest that lumefantrine resistance may already be emerging there, not long after AL was introduced as second-line therapy [13]. Future studies should compare different artemisinin and nonartemisinin combination therapies for their long-term effectiveness. Finally, the huge new market for AL in Africa has outstripped the available supplies, delaying the launch of ACTs in some countries, and it remains uncertain when these supply problems will be fully resolved.
Learning from Success
Will ACTs “roll back malaria”? The Barnes et al. paper focuses on the efficacy and implementation of AL in KwaZulu–Natal, but active surveillance and treatment for asymptomatic carriers [1], as well as residual spraying, contributed to the success. Barnes et al. argue for an effective vector control program and ACT implementation in the context of a well-developed rural primary health-care infrastructure. While the cost of executing these programs throughout Africa may seem great, the cost of not doing so is likely to be greater, especially if resistance to ACTs emerges as a consequence. In any case, the widespread expectation that ACTs alone will turn the tide in the fight against malaria may be unrealistic.
Citation: Duffy PE, Mutabingwa TK (2005) Rolling back a malaria epidemic in South Africa. PLoS Med 2(11): e368.
Abbreviations
ACTartemisinin-based combination therapy
ALartemether-lumefantrine
AQamodiaquine
SPsulfadoxine-pyrimethamine
==== Refs
References
Craig MH Kleinschmidt I Le Sueur D Sharp BL Exploring 30 years of malaria case data in KwaZulu-Natal, South Africa: Part II. The impact of non-climatic factors Trop Med Int Health 2004 9 1258 1266 15598257
Barnes KI Durrheim DN Little F Jackson A Mehta U Effect of artemether-lumefantrine policy and improved vector control on malaria burden in KwaZulu-Natal, South Africa PLoS Med 2005 2 e330 10.1371/journal.pmed.0020330 16187798
Nabarro D Roll Back Malaria Parassitologia 1999 41 501 504 10697910
[Anonymous] Reversing the failures of Roll Back Malaria Lancet 2005 365 1439 15856540
White NJ Antimalarial drug resistance J Clin Invest 2004 113 1084 1092 15085184
Sutherland CJ Ord R Dunyo S Jawara M Drakeley C Reduction of malaria transmission to anopheles mosquitoes with a six-dose regimen of co-artemether PLoS Med 2005 2 e92 10.1371/journal.pmed.0020092 15839740
Nosten F van Vugt M Price R Luxemburger C Thway KL Effects of artesunate-mefloquine combination on incidence of Plasmodium falciparum malaria and mefloquine resistance in western Thailand: A prospective study Lancet 2000 356 297 302 11071185
Tanariya P Tippawangkoso P Karbwang J Na-Bangchang K Wernsdorfer WH In vitro sensitivity of Plasmodium falciparum and clinical response to lumefantrine (benflumetol) and artemether Br J Clin Pharmacol 2000 49 437 444 10792201
Ittarat W Pickard AL Rattanasinganchan P Wilairatana P Looareesuwan S Recrudescence in artesunate-treated patients with falciparum malaria is dependent on parasite burden not on parasite factors Am J Trop Med Hyg 2003 68 147 152 12641403
Snow RW Trape JF Marsh K The past, present and future of childhood malaria mortality in Africa Trends Parasitol 2001 17 593 597 11756044
Yeka A Banek K Bakyaita N Staedke SG Kamya MR Artemisinin versus nonartemisinin combination therapy for uncomplicated malaria: Randomized clinical trials from four sites in Uganda PLoS Med 2005 2 e190 10.1371/journal.pmed.0020190 16033307
Mutabingwa T Nzila A Mberu E Nduati E Winstanley P Chlorproguanildapsone for treatment of drug-resistant falciparum malaria in Tanzania Lancet 2001 358 1218 1223 11675058
Sisowath C Stromberg J Martensson A Msellum M Obondo C In vivo selection of Plasmodium falciparum pfmdr1 86N coding alleles by artemether-lumefantrine (Coartem) J Infect Dis 2005 191 1014 1017 15717281
|
16187736
|
PMC1240091
|
CC BY
|
2021-01-05 10:39:19
|
no
|
PLoS Med. 2005 Nov 4; 2(11):e368
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020368
|
oa_comm
|
Estrogenicity of Styrene
Oligomers and Assessment
of Estrogen Receptor Binding
Assays
Polystyrene is frequently used in resins, and
the styrene dimers and trimers eluted from
polystyrene have been reported to have
estrogenic activity (1). We have performed
a number of in vitro and in vivo tests [i.e.,
estrogen receptor (ER) and androgen
receptor binding assays, thyroid hormone
receptor binding assays, human breast cancer cell line MCF-7 proliferation assays
(E-SCREEN), uterotrophic assays in immature and ovariectomized rats, Hershberger
assays, and prolactin release assays and
steroidogenesis] and found no effects of
styrene dimers or trimers on sex hormones
in any of these assays (2-7). These results
are supported by Fail et al. (8), who reported that mixtures of styrene oligomers did
not show any estrogenic activity in the
immature rat uterotrophic assay and the
reporter gene assay. In addition, the Japan
Environment Agency referred to their studies (9) and removed the styrene dimers and
trimers from their list of endocrine disruptors (9). However, Ohyama et al. (10)
reported that high concentrations of certain
styrene dimers and trimers showed estrogenic effects in an ER binding assay and in
the E-SCREEN assay. Recently, several
assay systems have been used to assess
endocrine-disrupting effects, but a few of
these assay systems can cause false-positive
reactions when test compounds are at high
concentrations (11).
To assess the accuracy of the ER binding
assay system and the results of Ohyama et al.
(10), and to ascertain the safety of styrene
dimers and trimers, we used a solubility test
and three ER binding assays (12) (Table 1).
The ER binding assay, which detects the
direct reactivity of ligand to a receptor, is the
most standardized and simple test system for
the detection of specific mechanisms of
estrogenic activity.
Using the radiosotope method (Method
RI) as described previously (13,14), we
observed that styrene dimers and trimers
did not show statistically significant
inhibitory action against the binding of
[3H]-17-estradiol (E2) to ER.
We used Method A to detect the binding affinities of test samples to human ER
(hER). Using a fluorescence polarization
Screen-for-Competitor Kit ER (Takara,
Kyoto, Japan) as described by Bolger et al.
(15), we measured the difference of polarization between fluorescence-labeled E2
(ES1) bound to ER and ES1 only. Styrene
dimers and trimers did not show statistically
significant inhibitory action against the
binding of ESI to ER in this assay.
We also used Method B, the method
used by Ohyama et al. (10), to detect the
binding affinities of test samples to the
human recombinant ER coated on the
microplate by competition with fluorescence-labeled E2; this was performed using
the Estrogen Receptor () Competitor
Screening Kit (Wako PC, Osaka, Japan).
Styrene dimers and trimers showed weak
inhibitory effect on the binding of fluorescein E2 to hER at 5 mol/L, and their
binding abilities were < 30% in this assay.
To evaluate the ER binding assays themselves, we included vitamin D3, naphthalene,
5-dihydrotestosterone, and testosterone in
each of the three ER binding assays; none of
these compounds bound ER in vitro
(13,16,17). A cross-reaction between estrogen and androgens cannot occur in vivo
unless the androgens are metabolized. In
Method RI and Method A, these nonestrogenic compounds did not show any ability to
bind to the ER. However, in Method B,
these compounds showed binding affinity
for the recombinant hER coated on the
microplate at such high concentrations that
they did not dissolve, although the binding
affinity of E2 was similar in each assay.
These results suggest that Method B tends
to detect false-positive effects and that it is
less accurate at high concentrations because
of a decline of specificity to estrogen at high
concentrations at which compounds do not
dissolve. The manufacturer's instructions for
the Estrogen Receptor () Competitor
Screening Kit used for Method B say to
"make sure there is no precipitation in the
solution." Styrene dimers and trimers are so
hydrophobic that their solubility is very low
in the buffer solutions used in each assay.
On the basis of these results, styrene dimers
and trimers have no affinity for ER in
Methods RI and A. Nevertheless, styrene
dimers and trimers exhibited some affinity
for the recombinant hER in the Method B
study, similar to that described by Ohyama
et al. (10), but at high concentrations such
that the compounds were not completely
dissolved. This result is not because of the
difference of sensitivity between rat ER and
human ER, as shown in Method A with
the use of hER, but is caused by a
decrease in specificity to estrogen because
of the precipitation of test compounds.
Ohyama et al. (10) reported that high
concentrations of styrene dimers and trimers
showed proliferative activity in the
E-SCREEN assay. Cell proliferation can be
induced by other growth factors, although
proliferation of MCF-7 cell is basically E2
dependent (18-20), and the response to E2
in MCF-7 cells varies because of the various
mutation of ER (21). Therefore, a false-positive response might only be shown in tests
using proliferation as a target. The luciferase
reporter gene assay, which indicates direct
gene expression reactivity through the receptor, has been considered to be a more suitable assay for evaluating estrogenicity at the
cellular level because of specificity to E2
response (22,23). Styrene dimers and
trimers did not show any estrogenic effect in
the E-SCREEN assay and the reporter gene
assay in our previous study (6). In addition,
A 384 VOLUME 110 | NUMBER 7 | July 2002 * Environmental Health Perspectives
Correspondence
Table 1. Solubility and binding affinity for ER of tested compounds.
Solubilitya Binding affinity for ER (ED30) (mol/L)
Compounds (mol/L) Method RI Method A Method B
Estrogenic compounds
17-Estradiol > 10 0.0012*** 0.005*** 0.001***
Bisphenol A > 10 5.0*** 1.7*** 2.0**
Styrene dimers
2,4-Diphenyl-1-butene 1.3 NC NC > 10.0
cis-1,2-Diphenylcyclobutane 9.4 NC NC 10.0**
trans-1,2-Diphenylcyclobutane 4.0 NC NC > 10.0
Styrene trimers
2,4,6-Triphenyl-1-hexene < 0.16 NC NC > 10.0
1e-Phenyl-4e-(1-phenylethyl) tetralin < 0.16 NC NC > 10.0
1a-Phenyl-4e-(1-phenylethyl) tetralin < 0.16 NC NC > 10.0
1a-Phenyl-4a-(1-phenylethyl) tetralin 0.17 NC NC > 10.0
1e-Phenyl-4a-(1-phenylethyl) tetralin 0.16 NC NC 5.2**
1e-Phenyl-4a-(2-phenylethyl) tetralin < 0.16 NC NC > 10.0
1a-Phenyl-4a-(2-phenylethyl) tetralin < 0.16 NC NC > 10.0
Androgens
Testosterone < 10 NC NC 105.0***
5-Dihydrotestosterone < 10 NC NC 45.0***
Nonestrogenic compounds
Vitamin D3 0.19 NC NC 100.0***
Naphthalene 100 NC NC 1010.0***
Abbreviations: ED30, concentration equivalent to 30% activity of 100 nmol/L E2; NC, no competition for binding of labeled
E2. Each value represents the mean of triplicate assays. .
aConcentration at which test compounds are saturated. **p < 0.01, ***p < 0.001 (vs. control, Dunnett test).
at high concentrations at which test compounds were precipitated, cells indicated an
abnormal response in the luciferase activity
of control plasmids and in morphology (data
not shown). To construct a stable assay system, we used HeLa cells transfected with an
hER expression plasmid derived from normal human liver ER. In this assay system,
styrene dimers and trimers did not show any
increase in E2-dependent luciferase transcription activity. These results agreed with
the result of the ER binding assay. We presume that styrene dimers and trimers had no
binding affinity to ER and they did not
affect E2-dependent transcription.
As a result, in our comparison of three
ER binding assays using estrogenic and
nonestrogenic compounds, it appeared that
Method RI and Method A were useful for
evaluating binding affinity for the ER, but
Method B, similar to the method of
Ohyama et al. (10), tended to indicate
false-positives in high concentrations in
which test chemicals were insoluble; this
reduced the specificity of ER to E2. Based
on our present results and previous reports
(2-7), we found no endocrine-disrupting
activities in styrene dimers and trimers eluted from polystyrene-containing instant
noodle containers.
Katsutoshi Ohno
Yukimasa Azuma
Katsuhiro Date
Shigeru Nakano
Toru Kobayashi
Yasuhiro Nagao
Toshihiro Yamada
Central Research Institute
Nissin Food Products Co., Ltd.
Shiga, Japan
E-mail: [email protected]
REFERENCES AND NOTES
1. Colborn T, Dumanoski D, Myers JP. Our Stolen Future.
New York:Dutton, 1996.
2. Yamada T. Synthesis, analysis and biological evaluation
of styrene oligomers. Yuki Goseikagaku Kyokaishi
57:58-64 (1999).
3. Nobuhara Y, Hirano S, Azuma Y, Date K, Ohno K, Tanaka
K, Matsushiro S, Sakurai T, Shiozawa S, Chiba M, et al.
Biological evaluation of styrene oligomers for
endocrine-disrupting effects. J Food Hyg Soc Japan
40:36-45 (1999).
4. Yamada T, Hirano S, Kobayashi K, Sakurai T, Takagi K,
Tanaka M, Nagao Y, Azuma Y, Date K, Ohno K, et al.
Identification, determination and biological evaluation of
novel styrene trimer in polystyrene container. Bunseki
Kagaku 49:493-501 (2000).
5. Azuma Y, Nobuhara Y, Date K, Ohno K, Tanaka K, Hirano
S, Kobayashi K, Sakurai T, Chiba M, Yamada T. Biological
evaluation of styrene oligomers for endocrine-disrupting
effects (II). J Food Hyg Soc Japan 41:109-115 (2000).
6. Ohno K, Azuma Y, Nakano T, Kobayashi S, Hirano T,
Nobuhara Y, Yamada T. Assessment of styrene
oligomers eluted from polystyrene-made food container
for estrogenic effects in vitro assays. Food Chem Toxicol
39:1233-1241 (2001).
7. Date K, Ohno K, Azuma Y, Hirano S, Kobayashi K,
Sakurai T, Nobuhara Y, Yamada T. Endocrine-disrupting
effects of styrene oligomers that migrated from polystyrene containers into food. Food Chem Toxicol
40:129-139 (2001).
8. Fail PA, Hines JW, Zacharewski T, Wu ZF, Borodinsky L.
Assessment of polystyrene extract for estrogenic activity in the rat uterotrophic model and an in vitro recombinant receptor reporter gene assay. Drug Chem Toxicol
21(suppl 1):101-121 (1998).
9. JEA. Strategic Programs on Environmental Endocrine
Disruptors '98. Available: http://www.env.go.jp/en/pol/
speed98/sp98.pdf [cited 30 April 2002].
10. Ohyama K, Nagai F, Tsuchiya Y. Certain styrene oligomers
have proliferative activity on MCF-7 human breast tumor
cells and binding affinity for human estrogen receptor .
Environ Health Perspect 109:699-703 (2001).
11. Nakano S, Nagao Y, Kobayashi T, Tanaka M, Hirano S,
Nobuhara Y, Yamada T. Problems with methods used to
screen estrogenic chemicals by yeast two-hybrid
assays. J Environ Health 48:83-88 (2002).
12. The Japanese Pharmacopoeia. 13th ed. Tokyo:Society of
Japanese Pharmacopoeia, 1997.
13. Blair RM, Fang B, Braham WS, Hass BS, Dial SL, Moland
C-L, Tong W, Shi L, Perkins R, Sheehan DM. The estrogen receptor binding affinities of 188 natural and xenochemicals: Structural diversity of ligands. Toxicol Sci
54:138-153 (2000).
14. Laws SC, Carey SA, Kelce WR, Cooper RL, Gray LE.
Vinclozolin does not alter progesterone receptor (PR)
function in vivo despite inhibition of PR binding by its
metabolites in vitro. Toxicology 110:1-11 (1996).
15. Bolger R, Weise TE, Evin K, Nestich S, Checovich W.
Rapid screening of environmental chemicals for estrogen receptor binding capacity. Environ Health Perspect
106:551-557 (1998).
16. Swami S, Krishnan AV, Feldman D. 1, 25-dihydroxyvitamin D3 down-regulates estrogen receptor abundance
and suppresses estrogen actions in MCF-7 human
breast cancer cells. Clin Cancer Res 6:3371-3379 (2000).
17. Nishihara T, Nishikawa J, Kanayama T, Dakeyama F,
Saito K, Imagawa M, Takatori S, Kitagawa Y, Hori S,
Utsumi H. Estrogenic activity of 517 chemicals by yeast
two-hybrid assay. J Health Sci 46:282-298 (2000).
18. Karey KP, Sirbasku DA. Differential responsiveness of
human breast cancer cell lines MCF-7 and T47D to growth
factors and 17-estradiol. Cancer Res 48:4083-4092 (1988).
19. Soto AM, Sonnenschein C. The role of estrogens on the
proliferation of human breast tumor cells (MCF-7). J
Steroid Biochem 23:87-94 (1985).
20. Soto AM, Sonnenschein C, Chung KL, Fernandez MF, Olea
N, Serrano FO. The E-SCREEN assay as a tool to identify
estrogens: an update on estrogenic environmental pollutants. Environ Health Perspect 103(suppl 7):113-122 (1995).
21. Pink JJ, Fritsch M, Bilimoria MM, Assikis VJ, Jordan VC.
Cloning and characterization of a 77-kDa oestrogen
receptor isolated from a human breast cancer cell line.
Br J Cancer 75:17-27 (1997).
22. Pons M, Gagne D, Nicolas JC, Mehtali M. A new cellular
model of response to estrogens: a bioluminescent test to
characterize (anti)estrogen molecules. Biotechniques
9:450-459 (1990).
23. Saito K, Tomigahara Y, Ohe N, Isobe N, Nakatsuka I,
Kaneko H. Lack of significant estrogenic or antiestrogenic activity of pyrethroid insecticides in three in vitro
assays based on classic estrogen receptor -mediated
mechanisms. Toxicol Sci 57:54-60 (2000).
Estrogenicity of Styrene
Oligomers: Response to
Ohno et al.
The main point of the letter by Ohno et al.
is that styrene oligomers have no estrogenic
activity, that our statement about "some
styrene oligomers having binding affinity
for hER" was inaccurate, and that the
MCF-7 cell proliferation assay is useless in
detecting estrogenicity.
It seems that Ohno et al. have misunderstood our article. We are confident that the
results of the MCF-7 cell proliferative assay
and the binding assay of styrene oligomers to
hER in our paper are accurate.
The inhibition of fluorescence-labeled
E2 binding to hER by styrene oligomers
tested is shown in Figure 3 of our paper
(1). The inhibition by styrene trimers 1a-phenyl-4e-(1-phenylethyl)tetralin (ST-3)
and 1e-phenyl-4a-(1-phenylethyl)tetralin
(ST-4) was detected at 5 x 10-7 M, and
the inhibition by styrene trimers 2,4,6-triphenyl-1-hexene (ST-1), 1a-phenyl-4a-(1-phenylethyl)tetralin (ST-2), and
1e-phenyl-4e-(1-phenylethyl)tetralin
(ST-5) was detected at 5 x 10-6 M,
both sufficiently soluble concentrations.
This means that ST-1, ST-2, ST-3, ST-4,
and ST-5 bound to hER at "not high"
concentrations. The maximum inhibition
by styrene trimers (ST-1, ST-2, ST-3, ST-4, and ST-5) was detected at 5 x 10-5 M;
this concentration is relatively low.
Although the maximum inhibition by
styrene dimers 1,3-diphenyl propane,
(SD-1), 2,4-diphenyl-1-butene (SD-2), cis-1,2-diphenyl cyclobutane (SD-3), and
trans-1,2-diphenyl cyclobutane (SD-4) was
detected at 5 x 10-4 M, these styrene
trimers and dimers were almost soluble at 5
x 10-5 M and 5 x 10-4 M, respectively. It
is important that the inhibition hardly
increased at each 10-times-higher concentration at which chemicals tested were partially insoluble. This result indicates that
the soluble chemicals reacted with hER in
saturated solution, and insoluble compounds did not influence the binding. In
Table 1 of their letter, Ohno et al. did not
clarify the solubility of the compounds. It
appears that the compounds were dissolved
in water because of the extremely low solubility. In our study we dissolved the compounds
in DMSO--the styrene dimers at 100,000
mol/L and the styrene trimers at 10,0000
mol/L, except for ST-2 (1,000 mol/L).
Ohno et al. should have included the concentrations of the saturating chemicals in the
reaction solutions of each method in their
Table 1, because when various concentrations of the chemical solvents (DMSO) are
added to the reaction solutions, the solubility
will become much higher.
If our binding assay indicated false positives in the range of concentrations in
which test chemicals were insoluble, the
inhibition by 1e,3e,5a-triphenylcyclohexa-
ne (ST-6) and 1e,3e,5e-triphenylcyclohexa-
ne (ST-7) would also increase, but no binding activity was observed for ST-6 and ST-7 at any concentration tested. This method
(Ohno et al.'s Method B) showed an
Environmental Health Perspectives * VOLUME 110 | NUMBER 7 | July 2002 A 385
Correspondence
increase in the inhibition of binding by
some soluble styrene oligomers but no
effect by the same chemicals at insoluble
concentrations. Ohno et al.'s Method B
showed no ED30 values of the styrene
oligomers at > 10 mol/L. Therefore,
Ohno et al.'s Method B also indicated no
effect by the styrene oligomers at insoluble
concentrations except testosterone, 5-dihydrotestosterone, vitamin D3, and
naphthalene. It seems that testosterone, 5-dihydrotestosterone, vitamin D3, and
naphthalene used by Ohno et al. had special characteristics for the competitive binding assay kit (Wako, Osaka, Japan).
The MCF-7 cell proliferation assay is a
recognized method for estrogenic screening.
Ohno et al. overemphasize other growth
factors. All of the styrene oligomers we tested did not have proliferative activity (1).
ST-6 and ST-7 had no proliferative activity
at all, but the proliferative potency of ST-3
and ST-4 was comparable with that of
bisphenol A. Moreover we confirmed that
OH-tamoxifen, an antagonist, inhibited
cell proliferation by ST-1, ST-3, ST-4, ST-5, SD-3, and SD-4 (2).
Recently, we reported that ST-1 and
ST-4 were estrogenic in the reporter gene
assay using MVLN cells established by stable transfection with the luciferase gene (3).
Moreover we found that some other styrene
oligomers were also estrogenic in this
reporter gene assay (2).
We are confident that our paper (1)
does not include any inaccurate results.
Ken-ichi Ohyama
Fumiko Nagai
Tokyo Metropolitan Research Laboratory
of Public Health
Tokyo, Japan
E-mail: [email protected]
REFERENCES AND NOTES
1. Ohyama K, Nagai F, Tsuchiya Y. Certain styrene oligomers
have proliferative activity on MCF-7 human breast tumor
cells and binding affinity for human estrogen receptor .
Environ Health Perspect 109:699-703 (2001).
2. Ohyama K, Nagai F, Satoh K, Aoki N. Unpublished data.
3. Ohyama K, Nagai F, Satoh K, Uehara A, Ohba M, Uehara
S, Aoki N. Hormonal activity of styrene oligomers determined by reporter gene assay using MVLN cells for
estrogen and competitive receptor binding assay for
androgen. Environ Sci 9:200 (2002).
A 386 VOLUME 110 | NUMBER 7 | July 2002 * Environmental Health Perspectives
Correspondence
Corrections and Clarifications
In "3-Chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone (MX) and
Mutagenic Activity in Massachusetts
Drinking Water" by Wright et al. [Environ
Health Perspect 110:157-164 (2002)],
there are two errors in "Materials and
Methods." In lines 12-16 of the second
paragraph describing analytical protocol,
"700C" should be "70C" and "600
mg/L aqueous NaHCO3" should be "600
g/L 2% aqueous NaHCO3." The correct
sentences are as follows:
The solution was heated to 70C to accelerate
the reaction. The mixture was neutralized by
addition of 600 g/L 2% aqueous NaHCO3
and extracted twice with 600 L n-hexane.
EHP regrets the errors.
In "Certain Styrene Oligomers Have
Proliferative Activity on MCF-7 Human
Breast Tumor Cells and Binding Affinity
for Human Estrogen Receptor " by
Ohyama et al. [Environ Health Perspect
109:699-703 (2001)], the grids in Figure
3 are incorrect. The corrected figure
appears at left. EHP regrets any confusion
caused by the incorrect grids.
Figure 3. The inhibition of fluorescence-labeled E2 binding to hER by various concentrations of styrene
oligomers. Percent of inhibition was calculated as [1 - (optical density in the presence of competitor) /
(optical density in the absence of competitor)] x 100. Each point is the mean SD of two independent
assays performed in duplicate.
*Significantly different from hormone-free control (p < 0.01).
|
28319026
|
PMC1240932
|
CC0
|
2021-01-05 11:06:51
|
no
|
Environ Health Perspect. 2002 Jul 7; 110(7):a385-a386
|
utf-8
|
Environ Health Perspect
| 2,002 |
10.1289/ehp.110-a385
|
oa_comm
|
126 VOLUME 112 | NUMBER 2 | February 2004 * Environmental Health Perspectives
Research | Commentary
Applying processed sewage sludges (biosolids)
to crop land, strip mines, public parks, and
other areas has become common in the
United States and elsewhere. This practice, in
which several tons or more of processed
municipal wastes are applied per acre annually, is regulated by the U.S. Environmental
Protection Agency (EPA) under the 503
sludge rule (U.S. EPA 1993). The rule provides guidance for the beneficial reuse of
municipal wastes and sets standards intended
to protect public health and the environment
from exposure to heavy metals, toxic chemicals, and pathogens. In recent years, land
application has been increasingly scrutinized
because of nuisance complaints and growing
numbers of anecdotal reports of illnesses and
deaths attributed to exposure to commercially
processed sewage sludges.
Our laboratories investigated public complaints and concluded that irritant chemicals
associated with volatile chemicals and dusts
blowing from treated land (e.g., bacterial toxins, lime, organic amines) may cause nearby
residents to be more susceptible to infections
(Lewis et al. 2002, Lewis and Gattie 2002).
We documented an outbreak of Staphylococcus
aureus among individuals exposed to a
Pennsylvania land-application site and attributed the infections to secondary exposure
routes (animal-to-human or person-to-person).
Overall, we questioned the efficacy of methods
used to treat sewage sludges and determine
pathogen levels; and, we recommended that
new research focus on chemical-pathogen
interactions, airborne contaminants (especially
organic dusts), and risks posed to immunocompromised individuals and other sensitive
populations (Lewis 1998; Lewis et al. 1999,
2000, 2001, 2002; Lewis and Gattie 2002).
The National Research Council (NRC
2002) echoed these same concerns, and the
U.S. EPA intends to address some of the issues
through additional research (U.S. EPA 2003a).
In this paper we provide a more detailed
overview of the risks that land application of
sewage sludge poses to human health and how
those risks can be better managed.
High-Level Disinfection
Current federal standards for pathogen reduction in sewage sludge are based on levels of
indicator organisms, such as Escherichia coli
and Salmonella. Class A sludges have no
detectible pathogens, whereas low levels of
indicator pathogens are permitted in class B
sludges. Sludges contain a wide variety of bacteria, viruses, protozoa, fungi, and parasitic
worms, including some species that are more
difficult to kill than the indicator organisms.
Table 1 shows levels of disinfection
required to destroy different groups of
pathogens found in sewage sludges. Low-level
disinfection reduces numbers of vegetative bacteria (e.g., E. coli, Salmonella) and enveloped
viruses [e.g., hepatitis B, human immunodeficiency virus (HIV), influenza viruses]. More
resistant organisms require intermediate-level
disinfection. These include mycobacteria (e.g.,
Mycobacterium tuberculosis), protozoa (e.g.,
Cryptosporidium, Giardia), parasitic worms
(e.g., Ascaris, Toxocara) and fungi (e.g.,
Candida). Intermediate to high-level disinfection is required to kill some of the most important pathogens found in sludges, including
small, nonenveloped viruses (e.g., Norovirus)
and bacterial endospores (e.g., Clostridium
perfringens).
Nonenveloped viruses comprise an important group of pathogens that require a higher
level of disinfection than the indicator organisms recommended in the 503 rule (U.S. EPA
1993). Rotaviruses, for example, cause
30-40% of acute diarrhea that requires infants
to be hospitalized, and Norovirus (Norwalk-like viruses) is responsible for 40% of the cases
of nonbacterial diarrhea in children and adults
(Berkow and Fletcher 1992). Other important
infectious agents in this group include
hepatitis A, hepatitis E, encephalomyocarditis
virus, polioviruses, coxsacki viruses, reoviruses,
rhinoviruses, astroviruses, caliciviruses,
echoviruses, parvoviruses, and aphthovirus.
Many of these viruses pose a particular threat
to infants, the elderly, and individuals with
chronic diseases.
The National Institute for Occupational
Safety and Health (NIOSH 2002) recently
concluded that Class B biosolids likely contain
infectious levels of bacteria, viruses, protozoa,
and helminths and recommended that workers use protective gear and take basic infection
control precautions when handling the material. In issuing these guidelines, NIOSH
acknowledged that current methods for processing Class B sewage sludges may fail to
achieve even low-level disinfection.
Also recognizing that freshly processed
Class B sludges may pose a significant risk of
infection under certain conditions, the
U.S. EPA included protective measures in the
503 rule (U.S. EPA 1993), such as temporarily restricting public access to Class B land-application sites with warning signs and
fences. The U.S. EPA, however, failed to consider some potentially important exposure factors; for example, dusts from treated fields
could expose surrounding communities, and
certain chemicals in sludge may increase risks
of infections. Moreover, stockpiling sludge and
spreading it without incorporating it into soil
are commonplace. In practice, the 503 rule is
ineffective in preventing public exposure.
Based on the types of pathogens present
in municipal wastes, sewage sludges should be
treated with high-level disinfection. To meet
this standard, treatment methods should
demonstrate the ability to kill even the most
Address correspondence to D.K. Gattie, Department
of Biological and Agricultural Engineering, Driftmier
Engineering Center, University of Georgia, Athens,
GA 30602-4435 USA. Telephone: (706) 542-0880.
Fax: (706) 542-8806. E-mail: [email protected]
We thank C. Snyder, Sierra Club Sludge Task Force,
for her assistance in taking a national survey of public
concerns. We also thank M. Novak for assisting with
the survey and providing other technical support.
The authors declare they have no competing financial
interests.
Received 13 January 2003; accepted 17 November
2003.
A High-Level Disinfection Standard for Land-Applied
Sewage Sludges (Biosolids)
David K. Gattie1 and David L. Lewis 2
1Department of Biological and Agricultural Engineering, and 2Department of Marine Sciences, University of Georgia, Athens, Georgia, USA
Complaints associated with land-applied sewage sludges primarily involve irritation of the skin,
mucous membranes, and the respiratory tract accompanied by opportunistic infections. Volatile
emissions and organic dusts appear to be the main source of irritation. Occasionally, chronic gastrointestinal problems are reported by affected residents who have private wells. To prevent acute
health effects, we recommend that the current system of classifying sludges based on indicator
pathogen levels (Class A and Class B) be replaced with a single high-level disinfection standard and
that methods used to treat sludges be improved to reduce levels of irritant chemicals, especially
endotoxins. A national opinion survey of individuals impacted by or concerned about the safety of
land-application practices indicated that most did not consider the practice inherently unsafe but
that they lacked confidence in research supported by federal and state agencies. Key words:
biosolids, sewage sludge. Environ Health Perspect 112:126-131 (2004). doi:10.1289/ehp.6207
available via http://dx.doi.org/ [Online 17 November 2003]
resistant organisms, including nonenveloped
viruses and bacterial spores. Because all federal and state requirements are based on less-resistant indicator organisms, it is not known
whether current methods, including aerobic
and anaerobic digestion, heat treatment, lime
stabilization, and composting, could achieve
high-level disinfection.
Pathogen Regrowth
Although high-level disinfection would afford
greater protection for both workers and the
public from pathogens in freshly processed
sewage sludge, the public can also be exposed
to pathogens that proliferate after the sludge
is applied (Gibbs et al. 1997). Viruses do not
replicate outside their hosts; therefore,
pathogen regrowth is mainly of concern with
bacteria and fungi. Consequently, while
viruses and other pathogens die off in the
field, some pathogens may rebound. Also,
new pathogens are introduced when sludge is
mixed with soil and comes in contact with
insects, birds, mammals, and other environmental sources of pathogens.
The potential for pathogen regrowth is
the downside to sewage sludge being rich in
nutrients that promote the growth of bacteria
and fungi. The problem is similar to food
poisoning with perishable foods, such as egg
products. Eggs, like raw sewage, are often contaminated with Salmonella. With a little cooking, however, egg-containing products are safe
for human consumption. Nevertheless, unless
these foods are desiccated or refrigerated, other
pathogens, such as S. aureus, multiply in them.
The source of S. aureus in spoiled food is not
the eggs, however, but normal skin microflora
from the hands of people who prepare or
handle the food.
Although sewage sludge is not a food
product, the principle is the same. Sludge is
rich in proteins and other nitrogen-rich
organic compounds that promote the growth
of S. aureus and other bacteria. These organisms multiply as sludges decompose in soil,
and can present a risk of infection when traces
of sludge enter skin abrasions or when the
dusts contact mucous membranes or are
inhaled. The risk is particularly high when
sewage sludge contacts tissues injured by
chemical irritants, burns, cuts, or abrasions.
People with chronic diseases and compromised immune systems are especially at risk.
Also, as is the case with food products,
sewage sludge that is heated or otherwise
treated to kill pathogens is still subject to
pathogen regrowth. In fact, because most of
the competing microorganisms are eliminated,
it is even more conducive to pathogen
regrowth. Leaving pathogens in sewage sludge,
however, is not the solution.
Unfortunately, pathogen regrowth is an
inherent problem with all sludges rich in proteins, amino acids, and other forms of organic
nitrogen and sulfur--regardless of how they
are processed. Once the materials are applied
and become wet, they are colonized by bacteria and fungi; the materials then decompose
and emit noxious odors in the form of
organic amines, organic sulfides, and other
small-molecular-weight compounds.
Offensive odors that form as sludge biologically decomposes in the field indicate
pathogen regrowth because they are produced
as bacteria break down proteins and other
organic compounds containing nitrogen and
sulfur. Most treatment methods produce
sludges that are only temporarily stable; that
is, the sludges produce noxious odors from
biological decomposition after they are
applied in the field.
One commercial process achieves longterm stability by chemically reacting sludge
under heat and pressure at high pH to drive
off organic nitrogen as ammonia (Reimers
et al. 2003). With this process, the combination of gaseous ammonia, high temperature,
and pressure effectively eliminates a wide
range of pathogens. The final wet product,
which is odorless and has a high pH, is used to
amend acidic soils. Because the nitrogen content is driven off, however, the product lacks
nutrient value.
Bacterial Toxins
Most bacteria found in sewage sludge produce
either endotoxins or exotoxins, both of which
can cause severe illness or death. As sludges
decompose, toxins can leach into groundwater,
enter surface water runoff, and be carried away
in airborne dusts. Considering that tons of
decomposing sewage sludge per acre are often
applied to hundreds or thousands of acres
many times a year, land-application sites have a
potential for producing and exporting large
quantities of toxins.
Exotoxins--proteins and peptides secreted
into the surrounding environment by growing
cells--are produced by both gram-negative
and gram-positive bacteria. They are usually
the most toxic of the two general types of bacterial toxins. Because they can retain their toxicity at extremely high dilutions, some
exotoxins, including staphylococcal enterotoxins and shigatoxin, are used as biological
warfare agents.
Although exotoxins are generally heat labile
and could therefore be destroyed by heat-treatment processes for sewage sludges, treated
sludges are still likely to become contaminated
with E. coli, Pseudomonas auruginosa, and other
exotoxin-producing bacteria in the field. Severe
gastrointestinal illnesses reported by individuals
using private wells near land-application sites
may have been caused by exotoxins leaching
into groundwater.
The same property that makes S. aureus a
common cause of food poisoning--its ubiquitous presence--may also make it one of the
more common pathogens to proliferate in
sewage sludges after they are applied to land.
The organism produces an exotoxin that is not
destroyed by cooking. Symptoms caused by
S. aureus food poisoning (e.g., nausea, cramps,
vomiting) are due to the presence of this toxin.
Land-application sites with high levels of S.
aureus could contaminate air and water with
potentially harmful levels of both the organism
and its toxin.
Endotoxins, on the other hand, are
lipopolysaccharide complexes in the cell walls
of gram-negative bacteria only. They are associated with proteins and other components of
the cell walls and are released when the bacteria
die and cell walls break apart (Rylander 1995).
Endotoxins are produced in large quantities
when wastes colonized with gram-negative
bacteria are treated (Sigsgaard et al. 1994).
They would also be produced as gram-negative bacteria growing in nutrient-rich sludges
die off in the field.
Unlike most exotoxins, endotoxins are
heat stable even upon autoclaving (Baines
2000). They can, however, be inactivated with
dry heat at > 200oC for 1 hr (Williams 2001).
Traces of endotoxins in food and water can
cause headaches, fever, fatigue, and severe gastrointestinal symptoms; however, their primary
target is the lungs. In addition to the former
symptoms, inhaling endotoxin-contaminated
dusts can cause acute airflow obstruction,
shock, and even death. Chronic respiratory
effects can also develop [American Conference
of Government Industrial Hygienists
(ACGIH) 1999].
Commentary | High-level disinfection of sewage sludge
Environmental Health Perspectives * VOLUME 112 | NUMBER 2 | February 2004 127
Table 1. Disinfection levels required to kill pathogens in sewage sludges.a
Group Disinfection level required
Bacterial endospores (e.g., Bacillus anthracis) High
Nonenveloped viruses (e.g., Norovirus, Coxsackie, Rotavirus) Intermediate/high
Helminths (e.g., Ascaris, Toxocara) Intermediate
Protozoa (e.g., Cryptosporidium, Giardia) Intermediate
Mycobacteria (e.g., M. tuberculosis) Intermediate
Fungi (e.g., Candida) Low/intermediate
Vegetative bacteria (e.g., Staphylococcus, Salmonella) Low
Enveloped viruses (e.g., hepatitis B, HIV, influenza) Low
Data from the Association for the Advancement of Medical Instrumentation (AAMI 1994).
aDisinfection levels are based on susceptibilities to liquid chemical germicides; groups increase similarly in resistance to heat,
with enveloped viruses being the most sensitive and bacterial endospores the most resistant.
Allergic and nonallergic reactions caused
by airborne endotoxins have been documented with exposures of 45-150 endotoxin
units (EU)/m3 and 300-400 EU/m3 (Milton
et al. 1996; Smid et al. 1994). Nearby residents exposed to dusts from land-application
sites report many of the same symptoms of
endotoxin poisoning that have been documented among sewage treatment plant workers. These include flu-like symptoms, nausea,
vomiting, diarrhea, headaches, and difficulty
breathing (Lewis et al. 2002). Rylander
(1987) proposed occupational exposure limits
to endotoxin-contaminated cotton dusts.
Based on average air concentrations over an
8- to 10-hr workday, he suggested limits
ranging from 200 EU/m3 to prevent airway
inflammation to 20,000 EU/m3 to avoid
toxic pneumonitis. The exposure levels of
endotoxin-contaminated aerosols with sewage
treatment plant workers have ranged from
80 to 4,100 EU/m3 (Liesvuori et al. 1994).
The toxins, however, have a greater effect on
people with immune systems compromised
by injury or illness (Baines 2000).
Chemical-Pathogen Interactions
Although many chemical contaminants found
in processed sewage sludges may potentially
interact with pathogens to cause, facilitate, or
exacerbate the disease process through allegeric
and nonallergic mechanisms, microbial byproducts formed during the processing and
decomposition of sewage sludge probably
account for most of the acute health effects.
Complaints among residents living near land-application sites are primarily respiratory
related and are consistent with hypersensitivity
reactions, including fever, cough, difficulty in
breathing, nausea, and vomiting.
Numerous diseases involving immunologically mediated hypersensitivity reactions have
been documented among workers exposed to
organic dusts containing microbial products.
Yi (2002) listed 27 diseases, each categorized
according to the source of the dusts and the
specific microorganisms identified as the primary cause of hypersensitivity. Sources include,
for example, dusts from molded hay, mushroom compost contaminated with fungi and
actinomycetes, Streptomyces-contaminated fertilizers, Caphaloporium-contaminated sewage,
and wood contaminated with Bacillus subtilis.
Byssinosis, perhaps the most studied of
these diseases, is attributed to traces of endotoxins from the breakdown of E. coli and
other gram-negative bacteria on raw cotton
fibers. Similarly, illnesses have been documented among wastewater treatment plant
workers exposed to endotoxins in aerosols
(Rylander 1987). Usually, the disease affects
only a small percentage of sensitive workers.
Compared with waste treatment plant
aerosols, however, endotoxin levels are probably
much higher in sewage sludge dusts, which
contain large numbers of predominantly
gram-negative bacteria killed during treatment processes and after land application.
Consequently, the frequency and severity of
hypersensitivity among groups exposed to
sewage sludge dusts may be much greater compared with exposure to other organic dusts.
Respiratory-related hypersensitivity is generally reversible when affected individuals are
removed from the source of exposure and
treated with high doses of corticosteroids.
Corticosteroids used to treat the underlying
inflammation, however, seriously impair the
immune system. In the case of sewage sludge,
this would render hypersensitive individuals
highly susceptible to infection from the low
levels of viruses, bacteria, fungi, and other
pathogens in processed sludges.
Treating residents near land-application
sites who experience hypersensitivity to
processed sewage sludges, therefore, is both
costly and risky. It would involve relocating
affected individuals to another area, making
certain that any potentially serious infections
have been eliminated or controlled with
antibiotics, then administering high doses of
corticosteroids and closely monitoring for any
new infections.
Exposure Studies
A conference on health effects of odors sponsored by Duke University and the U.S. EPA
concluded that gases and volatile emissions
from waste products including processed
sewage sludge may cause adverse health effects
(Schiffman et al. 2000). While acknowledging the complexity of the problem, the participants recommended undertaking controlled
studies of the odorous emissions. In responding to the NRC recommendations (NRC
2002), the U.S. EPA committed to measuring
field concentrations of selected volatile and
gaseous compounds at selected sites (U.S.
EPA 2003a).
Land-applied sewage sludge can emit
numerous volatile chemicals and gases that
may act alone or in combination with one
another to produce the kinds of symptoms
reported by people living near biosolids-
recycling operations. The composition of air
contaminants emitted by any land-application
site undoubtedly varies widely over space and
time, as do the susceptibilities of individuals to
the effects of these emissions. Consequently, it
is unlikely that such research will adequately
establish which components and combinations
of components can potentially cause adverse
health effects and under what conditions.
We propose an alternative approach with a
more modest goal aimed at determining the
extent to which emissions must be diluted to
eliminate malodor complaints and irritant
effects (e.g., burning eyes, coughing, breathing
difficulties). Based on meteorologic data from
local weather stations and publicly available
topographic data, the dilution of air contaminants over areas surrounding land-application
sites can be readily determined with an air-
dispersion model (Lewis et al. 2002).
Meteorologic data should be collected over
an extended period of time (e.g., 6-8 weeks)
during the maximum potential exposure, for
example, when land application is in progress,
temperatures are high, and sufficient rainfall
has occurred to support high levels of microbial activity. This approach does not require
measuring specific pollutants, and a number
of land-application sites could be studied with
a reasonable level of resources.
In addition to collecting meteorologic data,
local census data would be used in this type of
study to randomly select two groups of residents of similar demographic compositions:
one close to the land-application site (< 1 km)
and one farther away (3-5 km). Individuals in
each group would provide information on
medical histories and keep daily records of the
selected symptoms during the period when
meteorologic data are collected. Using a similar
approach, we found that residents living closer
to land-application sites were more severely
affected than those living farther away (Lewis
et al. 2002).
Quantitatively, what is needed is a simple
numerical index that captures the most important variables determining whether symptoms
develop. The amount by which volatile emissions are diluted when they reach a residence
and the number of times the dilution drops
below a certain level may be sufficient for predicting whether odor and health-related complaints are likely to develop.
For example, an exposure index could be
calculated based on a) the number of exposures
in which levels of volatile chemicals at a residence are 10% of the levels over the sludged
field and b) the average percent dilution for
these exposures. This approach is illustrated by
Equation 1, where Iv is the exposure index for
volatile emissions, n is the number of exposures
in which levels of gases and volatile chemicals
at a residence were 10% of the levels over the
sludged field, and d
-
is the average dilution
(percent) for all exposures 10% of the levels
over the field.
Iv = n * d
-
[1]
Once exposure indices and frequencies of
symptoms are collected for a number of land-application sites, a representative dilution
level required to eliminate odor complaints
and acute adverse health effects can be determined. This index provides a quantitative
measure of whether a land-application site is
likely to cause odor complaints and acute
adverse health effects at a particular location
Commentary | Gattie and Lewis
128 VOLUME 112 | NUMBER 2 | February 2004 * Environmental Health Perspectives
or distance from the site. Such information
could be used to evaluate the effectiveness of
treatment methods and land management
practices.
Public Concerns
To assess public concerns over the safety of
current land-application practices, we distributed questionnaires to 150 individuals concerned about land-application of sewage
sludges (Table 2). The group included farmers,
residents complaining of adverse health effects,
community leaders, and environmentalists.
Based on the responses of 87 respondents
from 15 states, a majority of respondents
(51.7%) desired a total ban on land application of sewage sludges, while 35.6% believed
that land application should just be suspended
until proven safe. Most respondents (74.7%)
lived near land-applications sites and most
(67.5%) reported that they had been personally affected by the practice. Overwhelming
malodor, vector attraction (flies, mosquitoes),
and adverse health effects (e.g., difficulty
breathing, chronic sinusitis) were the primary
adverse effects reported by individuals living
near the sites.
The need for additional research was
strongly supported. Respondents, however,
expressed little trust in federal and state environmental agencies to provide a reliable scientific evaluation of potential public health and
environmental effects. More confidence was
expressed if assessments were done by public
health agencies, such as the Centers for
Disease Control and Prevention (CDC) and
Commentary | High-level disinfection of sewage sludge
Environmental Health Perspectives * VOLUME 112 | NUMBER 2 | February 2004 129
Table 2. Summary of survey results from 87 respondents indicating their level of public concern about land application practices.
Topic Question Choices Percent or mean SD (n)
Background Why are you interested Live or lived near land application site 74.7% (87)
informationa in the issue of land-applied Work as a farmer/grower 4.6% (87)
sewage sludges? Engaged in environmental activism 16.1% (87)
Other 14.9% (87)
Have you ever been personally Yes 67.5% (77)
affected by land application of
sewage sludges?
How do you think land application Current practices are safe;no new restrictions are needed 0% (87)
of sewage sludges should be All land application should be completely banned 51.7% (87)
handled? Only certain land application practices should be banned 8.0% (87)
All land application should be suspended until proven safe 35.6% (87)
Land application should be continued with certain new restrictions 2.3% (87)
Other 5.7% (87)
Level of On a scale of 0 to 10 (0 = no concern; Microorganisms that may cause infection 9.7 0.9 (83)
concernb 10 = highest level of concern), indicate Chemicals, metals and microorganism that may cause cancer 9.6 1.0 (84)
your level of concern regarding the Odor-causing emissions 8.9 1.9 (84)
following issues Bacterial toxins 9.7 0.8 (82)
Property value 8.6 2.3 (82)
Other 9.7 0.7 (25)
Kinds of contamination from sludges Contamination of food supply 9.3 1.4 (83)
that cause the most concern (0 = no Contamination of water 9.9 0.7 (84)
concern; 10 = highest level Contamination of soil 9.8 0.5 (84)
of concern) Contamination of air 9.6 1.0 (84)
Other 9.8 0.5 (26)
Level of Using a scale of 0 to 10 (0 = no trust; Congress 2.0 2.3 (81)
trustb 10 = highest level of trust) U.S. EPA 1.3 2.4 (82)
indicate your level of trust in organizations U.S. Department of Agriculture 1.7 2.5 (80)
dealing with land application of State agencies 1.2 2.1 (81)
sewage sludges Local governments (city/county) 2.4 3.1 (81)
Environmental organizations 7.0 2.8 (81)
Trade groups (e.g., WEF, NEBRA) 0.8 2.1 (75)
National Biosolids Partnership 0.5 1.7 (72)
Industry 0.5 1.5 (80)
Otherc (e.g., departments of health, independent scientists) 5.3 4.4 (19)
Need for On a scale of 0 to 10 (0 = don't feel that 9.6 1.7 (84)
additional more research is needed; 10 = feel very
researchb strongly that more research is needed),
indicate how strongly you feel that more
scientific research is needed before we
will know whether land applying sewage
sludges is safe for public health and
the environment
On a scale of 0 to 10 (0 = no trust; National Science Foundation/National Institutes of Health 5.1 3.3 (72)
10 = highest level of trust), Trade groups (WEF/WERF) 1.4 2.2 (72)
indicate your level of trust in the work U.S. EPA Office of Water 1.4 2.4 (77)
of scientists supported by various U.S. EPA Office of Research and Development 2.8 3.1 (75)
organizations dealing with land U.S. Department of Agriculture 1.8 2.6 (76)
application of sewage sludges Centers for Disease Control and Prevention 4.9 3.4 (75)
Industry 0.5 1.4 (76)
State agencies 1.2 2.1 (78)
Otherc (e.g., universities, independent scientists, environmental groups) 5.9 4.7 (19)
Abbreviations: NEBRA, New England Biosolids and Residuals Association; WEF, Water Environment Federation; WERF, Water Environment Research Federation. Responses not following
survey instructions were omitted.
aBased on yes/no responses from all 87 respondents; values shown are percent SD (n). bBased on a 0-10 scale, with averages determined from the actual number of responses to
each category; values shown are mean SD (n). cCategories given by respondents.
the National Institutes of Health (NIH).
Although few respondents (16.1%) reported
engaging in environmental activism, the group
rated environmentalist organizations as being
the most trustworthy to assess the safety of
land-application practices.
Overall, the survey results indicate that
most people concerned about sewage sludge do
not believe land application is inherently
unsafe but object to the practice because they
lack confidence in scientific studies funded by
government and industry groups defending the
status quo. By contrast, survey respondents
indicated a greater level of confidence in studies of land-application practices if done by the
CDC or the NIH. It appears, therefore, that
overcoming opposition through additional scientific research will require strong involvement
with respected public health organizations. It
will also require supportive findings from
researchers independent of the federal agencies
and trade organizations that have historically
overseen the development and marketing of
land-application practices.
Discussion
Politics of land application. Since the 503
sludge rule (U.S. EPA 1993) was promulgated
in 1993, the U.S. EPA, the U.S. Department of
Agriculture, and the industry and its trade associations have vigorously defended the rule as
fully protective of public health and the environment. The primary basis has been a lack of
documented cases of illnesses and the results of
research supported with congressional funds
earmarked for promoting land application as
safe and beneficial (U.S. EPA 2002).
In 2000, the Committee on Science in the
U.S. House of Representatives held hearings
into allegations that the U.S. EPA retaliates
against scientists and private citizens who
report adverse environmental and health
effects associated with sewage sludge (U.S.
House of Representatives 2000a, 2000b).
During the hearings, the U.S. EPA Office
of the Inspector General released a report confirming that concerns were widespread among
U.S. EPA scientists who reviewed the 503 rule
(U.S. EPA 2002). The assistant administrator
refused to approve the rule without a major
commitment from the Office of Water to support additional in-house research within the
Office of Research and Development. The
Inspector General noted that this research was
never carried out.
The U.S. EPA responded to the congressional hearings by calling for a study by the
National Research Council; Congress debated,
and overwhelmingly passed, the No Fear
Act (Notification and Federal Employee
Antidiscrimination and Retaliation Act of
2002). The act, which was aimed at better protecting employees against retaliation, was
signed by President Bush last year. The NRC
published its findings and recommendations in
July 2002 (NRC 2002), and the U.S. EPA
addressed them in a research strategy in the
Federal Register earlier this year (U.S. EPA
2003a). The U.S. EPA's final response to the
NRC report is due to be released in January
2004. In the meantime, the U.S. EPA Office
of Water has provided a docket for public
comments (U.S. EPA 2003b).
Public comments to the Office of Water
docket have largely mirrored those in the survey we report in this article. There is an overall lack of confidence in the the U.S. EPA's
willingness to conduct or support objective
research in this area. As evidence, the Sierra
Club (San Francisco, CA) and others pointed
to the fact that the Office of Water intends to
address the NRC recommendations extramurally by funding the same researchers it has
historically supported with congressional
appropriations for promoting the safety of
biosolids. The NRC study (NRC 2002),
therefore, had its beginning and ending
rooted in controversy over the U.S. EPA
using congressional appropriations to support
federal policies on the beneficial reuse of
sewage sludge and to oppose scientists and
private citizens who question them.
Trends in land application. With increasing numbers of residents who live near
sludged fields reporting respiratory and gastrointestinal illnesses (Shields 2003), many local
governments have banned land application of
Class B sewage sludges. However, we found
that some Class A sludges generate the same
complaints and concluded that going to Class
A products will not resolve the pathogens issue
(Lewis and Gattie 2002). The reason for this is
that infections appear to be primarily opportunistic, following irritation of the skin,
mucous membranes, and respiratory tract by
chemical components.
Consequently, an important aspect of preventing infections lies in reducing levels of
microbial toxins and other chemicals that
cause inflammation as well as other responses
that predispose individuals to infection. As
such, the infections arise from many sources,
both community and environmental. The
problem with Class A sludges is probably primarily endotoxin related. This is because
gram-negative bacteria comprise much of the
biomass and because most conditions used to
kill bacteria in treatment processes are insufficient to break down endotoxins. The Class A
standard, therefore, while reducing the risks of
acquiring infections directly from processed
sludge, could increase risks of infections from
other environmental and community sources.
One outcome of local bans is that land
application of sewage sludge is being forced out
of areas where residents have the political and
economic resources to oppose the practice and
into economically depressed areas. Whether
this is intentional or not, sewage sludge is
being dumped more and more into those communities least able to have their complaints
heard, and where residents are least capable of
relocating or obtaining medical treatment.
The changing demographics of land
application of sewage sludge, therefore, need
to be studied. First, census data should be
used to assess the socioeconomic makeup of
communities living near land-application sites
(< 1 km away). Steps then need to be taken to
ensure that land-application practices do not
disproportionately impact low-income and
minority communities.
Recommendations
We recommend that the U.S. EPA undertake
and complete five top-priority measures by
January 2006 to address the immediate adverse
health and environmental effects associated
with land-applied sewage sludges:
* Develop a universal high-level disinfection
standard to replace the current Class A/
Class B standards, and require industry to
provide efficacy data showing treatment
methods meet this standard for sporocidal,
fungicidal, bactericidal, tuberculocidal,
virucidal, anti-protozoal, and anti-parasitic
activity
* Develop treatment methods and land management practices for reducing airborne
endotoxin levels associated with processed
sewage sludges and set limits for public
exposure. (We recommend that maximum
levels be set at 0.1 times the limit recommended for 10-hr occupational exposure
for preventing airway inflammation, which
would be 20 EU/m3.)
* Conduct a national assessment of groundwater contamination from pathogens,
microbial (bacterial, fungal) toxins, organic
chemicals, and metals at land-application
sites
* Require that industry ensure that land-application practices do not disproportionately target low-income and minority
subpopulations in rural communities
* Work with the National Biosolids
Partnership (Association of Metropolitan
Sewerage Agencies, U.S. EPA, and the
Water Environment Federation) to develop
and enforce a policy supporting open competition for research funding and prohibiting discrimination and retaliation against
individuals raising concerns over adverse
environmental and health effects from
land-applied sewage sludges.
REFERENCES
ACGIH. 1999. Bioaerosols Assessment and Controls. Cincinnati,
OH:American Conference of Government Industrial
Hygienists.
AAMI. 1994. Designing, Testing & Labeling Reusable Medical
Devices for Reprocessing in Health Care Facilities: A Guide
Commentary | Gattie and Lewis
130 VOLUME 112 | NUMBER 2 | February 2004 * Environmental Health Perspectives
Commentary | High-level disinfection of sewage sludge
Environmental Health Perspectives * VOLUME 112 | NUMBER 2 | February 2004 131
for Device Manufacturers. Technical Information Report
No. 12. Arlington, VA:Association for the Advancement of
Medical Instrumentation.
Baines A. 2000. Endotoxin testing. In: Handbook of Microbiological
Quality Control (Baird R, Hodges N, Denyer S, eds). New
York:Taylor & Francis, 144-167.
Berkow R, Fletcher A. 1992. Gastroenteritis: infective and toxic. In:
The Merck Manual. 16th ed. Rahway, NJ:Merck Research
Laboratories, 812-821.
Gibbs RA, Hu CJ, Ho GE, Unkovich I. 1997. Regrowth of faecal coliforms and salmonellae in stored biosolids and soil amended
with biosolids. Water Sci Technol 35:269-275.
Lewis DL. 1998. Microbes in the environment: challenges to exposure assessment. Presented at the Annual Meeting and
Science Innovation Exposition of the American Association
for the Advancement of Science, 12-17 February 1998.
Philadelphia, PA.
Lewis DL, Garrison AW, Wommack KE, Whittemore A, Steudler P,
Melillo J. 1999. Influence of environmental changes on
degradation of chiral pollutants in soils. Nature. 401:898-901.
Lewis DL, Gattie DK. 2002. Pathogen risks from applying
sewage sludge to land. Environ Sci Technol 36:286A-293A.
Lewis DL, Gattie DK, Novak M, Sanchez S, Pumphrey C. 2001.
Interactions of pathogens and irritant chemicals in
land-applied sewage sludge (biosolids). In: New Solutions,
Vol 12, No 4 (Clapp R, Orlando L, eds). Amityville,
NY:Baywood Publishing Co.
------. 2002. Interactions of pathogens and irritant chemicals in
land-applied sewage sludges (biosolids). BMC Public
Health 2:11.
Lewis DL, Shephard S, Gattie DK, Sanchez S, Novak M. 2000.
Enhanced susceptibility to infection from exposure to gases
emitted by sewage sludge: a case study. In: Proceedings of
Biosolids Management in the 21st Century, 10-11 April 2000,
College Park, MD. College Park, MD:Department of Civil &
Environmental Engineering, University of Maryland,
168-174.
Liesivuori J, Kotimaa M, Laitinen S, Louhelainen K, Ponni J,
Sarantila R, et al. 1994. Airborne endotoxin concentrations
in different work conditions. Am J Ind Med 25(1):123-124.
Milton DK, Wypij D, Kriebel D, Walters MD, Hammond SK,
Evans JS. 1996. Endotoxin exposure-response in a fiberglass manufacturing facility. Am J Ind Med 29(1):3-13.
National Research Council. 2002. Biosolids Applied to Land:
Advancing Standards and Practice. Washington,
DC:National Academy Press.
NIOSH. 2002. Guidance for Controlling Potential Risks to Workers
Exposed to Class B Biosolids. NIOSH Publication No.
2002-149. Cincinnati, OH:National Institute for Occupational
Safety and Health. Available: http://www.cdc.gov/niosh/
docs/2002-149/pdfs/2002-149.pdf [accessed 4 December
2003].
Notification and Federal Employee Antidiscrimination and
Retaliation Act of 2002. 2002. Public Law 107-174.
Reimers RS, Oleszkiewicz JA, Shepherd SL, Bakeer RM,
Fitzmorris KB. 2003. Advances in Alkaline Stabilization/
Disinfection of Agricultural and Municipal Biosolids.
Baltimore, MD:Water Environment Federation.
Rylander R. 1987. The role of endotoxin for reactions after
exposure to cotton dust. Am J Ind Med 12(6):687-697.
Rylander R. 1995. Endotoxins in the environment. In:
Lipopolysaccharides from Genes to Therapy (Levin J, Alving
C, Munford R, Redl H, eds). New York:Wiley-Liss, 79-90.
Schiffman SS, Walker JM, Dalton P, Lorig TS, Raymer JH,
Shustermann D, et al. 2000. Potential health effects of odor
from animal operations, wastewater treatment, and recycling of byproducts. J Agromed 7(1):1-81.
Shields H. 2003. Sludge Victims. Fall 2002 Update. Alton,
NH:Citizens for a Sludge-Free Land and New Hampshire
Sierra Club.
Sigsgaard T, Malmros P, Nersting L, Petersen C. 1994.
Respiratory disorders and atopy in Danish refuse workers.
Am J Respir Crit Care Med 149(6):1407-1412.
Smid T, Heederick D, Houba R, Quanjer PH. 1994. Dust- and endotoxin-related acute lung function changes and work-related
symptoms in workers in animal feed industry. Am J Ind Med
25(6):877-888.
U.S. EPA. 1993. 40 CFR Part 503. Fed Reg 58(32):9248-9415.
------. 2002. Land Application of Biosolids Status Report. 2002-S-000004. Washington, DC:U.S. Environmental Protection
Agency, Office of Inspector General.
------. 2003a. Standards for the Use or Disposal of Sewage
Sludge; Agency Response to the National Research Council
Report on Biosolids Applied to Land and the Results of
EPA's Review of Existing Sewage Sludge Regulations. Fed
Reg 68:17379-17395.
------. 2003b. Standards for the Use or Disposal of Sewage
Sludge; Agency Response to the National Research Council
Report on Biosolids Applied to Land and the Results of EPA's
Review of Existing Sewage Sludge Regulations. Docket No.
OW-2003-0006. Washington, DC:U.S. Environmental
Protection Agency. Available http://cascade.epa.gov/
RightSite/dk_public_home.htm [accessed 5 December 2003].
U.S. House of Representatives, Committee on Science. 2000a.
EPA's Sludge Rule: Closed Minds or Open Debate? No.
106-95. Washington, DC:U.S. Government Printing Office.
------. 2000b. Intolerance at EPA--Harming People, Harming
Science? No. 106-103. Washington, DC:U.S. Government
Printing Office.
Williams KL, ed. 2001. Pyrogen, endotoxin, and fever. In:
Endotoxins: Pyrogens, LAL Testing, and Depyrogenation.
2nd ed. New York:Marcel Dekker, Inc., 12-24.
Yi ES. 2002. Hypersensitivity pneumonitis. Crit Rev Clin Lab Sci
39(6):581-629.
|
14754565
|
PMC1241820
|
CC0
|
2021-01-05 07:07:44
|
no
|
Environ Health Perspect. 2004 Feb; 112(2):126-131
|
utf-8
|
Environ Health Perspect
| 2,004 |
10.1289/ehp.6207
|
oa_comm
|
126 VOLUME 112 | NUMBER 2 | February 2004 * Environmental Health Perspectives
Research | Commentary
Applying processed sewage sludges (biosolids)
to crop land, strip mines, public parks, and
other areas has become common in the
United States and elsewhere. This practice, in
which several tons or more of processed
municipal wastes are applied per acre annually, is regulated by the U.S. Environmental
Protection Agency (EPA) under the 503
sludge rule (U.S. EPA 1993). The rule provides guidance for the beneficial reuse of
municipal wastes and sets standards intended
to protect public health and the environment
from exposure to heavy metals, toxic chemicals, and pathogens. In recent years, land
application has been increasingly scrutinized
because of nuisance complaints and growing
numbers of anecdotal reports of illnesses and
deaths attributed to exposure to commercially
processed sewage sludges.
Our laboratories investigated public complaints and concluded that irritant chemicals
associated with volatile chemicals and dusts
blowing from treated land (e.g., bacterial toxins, lime, organic amines) may cause nearby
residents to be more susceptible to infections
(Lewis et al. 2002, Lewis and Gattie 2002).
We documented an outbreak of Staphylococcus
aureus among individuals exposed to a
Pennsylvania land-application site and attributed the infections to secondary exposure
routes (animal-to-human or person-to-person).
Overall, we questioned the efficacy of methods
used to treat sewage sludges and determine
pathogen levels; and, we recommended that
new research focus on chemical-pathogen
interactions, airborne contaminants (especially
organic dusts), and risks posed to immunocompromised individuals and other sensitive
populations (Lewis 1998; Lewis et al. 1999,
2000, 2001, 2002; Lewis and Gattie 2002).
The National Research Council (NRC
2002) echoed these same concerns, and the
U.S. EPA intends to address some of the issues
through additional research (U.S. EPA 2003a).
In this paper we provide a more detailed
overview of the risks that land application of
sewage sludge poses to human health and how
those risks can be better managed.
High-Level Disinfection
Current federal standards for pathogen reduction in sewage sludge are based on levels of
indicator organisms, such as Escherichia coli
and Salmonella. Class A sludges have no
detectible pathogens, whereas low levels of
indicator pathogens are permitted in class B
sludges. Sludges contain a wide variety of bacteria, viruses, protozoa, fungi, and parasitic
worms, including some species that are more
difficult to kill than the indicator organisms.
Table 1 shows levels of disinfection
required to destroy different groups of
pathogens found in sewage sludges. Low-level
disinfection reduces numbers of vegetative bacteria (e.g., E. coli, Salmonella) and enveloped
viruses [e.g., hepatitis B, human immunodeficiency virus (HIV), influenza viruses]. More
resistant organisms require intermediate-level
disinfection. These include mycobacteria (e.g.,
Mycobacterium tuberculosis), protozoa (e.g.,
Cryptosporidium, Giardia), parasitic worms
(e.g., Ascaris, Toxocara) and fungi (e.g.,
Candida). Intermediate to high-level disinfection is required to kill some of the most important pathogens found in sludges, including
small, nonenveloped viruses (e.g., Norovirus)
and bacterial endospores (e.g., Clostridium
perfringens).
Nonenveloped viruses comprise an important group of pathogens that require a higher
level of disinfection than the indicator organisms recommended in the 503 rule (U.S. EPA
1993). Rotaviruses, for example, cause
30-40% of acute diarrhea that requires infants
to be hospitalized, and Norovirus (Norwalk-like viruses) is responsible for 40% of the cases
of nonbacterial diarrhea in children and adults
(Berkow and Fletcher 1992). Other important
infectious agents in this group include
hepatitis A, hepatitis E, encephalomyocarditis
virus, polioviruses, coxsacki viruses, reoviruses,
rhinoviruses, astroviruses, caliciviruses,
echoviruses, parvoviruses, and aphthovirus.
Many of these viruses pose a particular threat
to infants, the elderly, and individuals with
chronic diseases.
The National Institute for Occupational
Safety and Health (NIOSH 2002) recently
concluded that Class B biosolids likely contain
infectious levels of bacteria, viruses, protozoa,
and helminths and recommended that workers use protective gear and take basic infection
control precautions when handling the material. In issuing these guidelines, NIOSH
acknowledged that current methods for processing Class B sewage sludges may fail to
achieve even low-level disinfection.
Also recognizing that freshly processed
Class B sludges may pose a significant risk of
infection under certain conditions, the
U.S. EPA included protective measures in the
503 rule (U.S. EPA 1993), such as temporarily restricting public access to Class B land-application sites with warning signs and
fences. The U.S. EPA, however, failed to consider some potentially important exposure factors; for example, dusts from treated fields
could expose surrounding communities, and
certain chemicals in sludge may increase risks
of infections. Moreover, stockpiling sludge and
spreading it without incorporating it into soil
are commonplace. In practice, the 503 rule is
ineffective in preventing public exposure.
Based on the types of pathogens present
in municipal wastes, sewage sludges should be
treated with high-level disinfection. To meet
this standard, treatment methods should
demonstrate the ability to kill even the most
Address correspondence to D.K. Gattie, Department
of Biological and Agricultural Engineering, Driftmier
Engineering Center, University of Georgia, Athens,
GA 30602-4435 USA. Telephone: (706) 542-0880.
Fax: (706) 542-8806. E-mail: [email protected]
We thank C. Snyder, Sierra Club Sludge Task Force,
for her assistance in taking a national survey of public
concerns. We also thank M. Novak for assisting with
the survey and providing other technical support.
The authors declare they have no competing financial
interests.
Received 13 January 2003; accepted 17 November
2003.
A High-Level Disinfection Standard for Land-Applied
Sewage Sludges (Biosolids)
David K. Gattie1 and David L. Lewis 2
1Department of Biological and Agricultural Engineering, and 2Department of Marine Sciences, University of Georgia, Athens, Georgia, USA
Complaints associated with land-applied sewage sludges primarily involve irritation of the skin,
mucous membranes, and the respiratory tract accompanied by opportunistic infections. Volatile
emissions and organic dusts appear to be the main source of irritation. Occasionally, chronic gastrointestinal problems are reported by affected residents who have private wells. To prevent acute
health effects, we recommend that the current system of classifying sludges based on indicator
pathogen levels (Class A and Class B) be replaced with a single high-level disinfection standard and
that methods used to treat sludges be improved to reduce levels of irritant chemicals, especially
endotoxins. A national opinion survey of individuals impacted by or concerned about the safety of
land-application practices indicated that most did not consider the practice inherently unsafe but
that they lacked confidence in research supported by federal and state agencies. Key words:
biosolids, sewage sludge. Environ Health Perspect 112:126-131 (2004). doi:10.1289/ehp.6207
available via http://dx.doi.org/ [Online 17 November 2003]
resistant organisms, including nonenveloped
viruses and bacterial spores. Because all federal and state requirements are based on less-resistant indicator organisms, it is not known
whether current methods, including aerobic
and anaerobic digestion, heat treatment, lime
stabilization, and composting, could achieve
high-level disinfection.
Pathogen Regrowth
Although high-level disinfection would afford
greater protection for both workers and the
public from pathogens in freshly processed
sewage sludge, the public can also be exposed
to pathogens that proliferate after the sludge
is applied (Gibbs et al. 1997). Viruses do not
replicate outside their hosts; therefore,
pathogen regrowth is mainly of concern with
bacteria and fungi. Consequently, while
viruses and other pathogens die off in the
field, some pathogens may rebound. Also,
new pathogens are introduced when sludge is
mixed with soil and comes in contact with
insects, birds, mammals, and other environmental sources of pathogens.
The potential for pathogen regrowth is
the downside to sewage sludge being rich in
nutrients that promote the growth of bacteria
and fungi. The problem is similar to food
poisoning with perishable foods, such as egg
products. Eggs, like raw sewage, are often contaminated with Salmonella. With a little cooking, however, egg-containing products are safe
for human consumption. Nevertheless, unless
these foods are desiccated or refrigerated, other
pathogens, such as S. aureus, multiply in them.
The source of S. aureus in spoiled food is not
the eggs, however, but normal skin microflora
from the hands of people who prepare or
handle the food.
Although sewage sludge is not a food
product, the principle is the same. Sludge is
rich in proteins and other nitrogen-rich
organic compounds that promote the growth
of S. aureus and other bacteria. These organisms multiply as sludges decompose in soil,
and can present a risk of infection when traces
of sludge enter skin abrasions or when the
dusts contact mucous membranes or are
inhaled. The risk is particularly high when
sewage sludge contacts tissues injured by
chemical irritants, burns, cuts, or abrasions.
People with chronic diseases and compromised immune systems are especially at risk.
Also, as is the case with food products,
sewage sludge that is heated or otherwise
treated to kill pathogens is still subject to
pathogen regrowth. In fact, because most of
the competing microorganisms are eliminated,
it is even more conducive to pathogen
regrowth. Leaving pathogens in sewage sludge,
however, is not the solution.
Unfortunately, pathogen regrowth is an
inherent problem with all sludges rich in proteins, amino acids, and other forms of organic
nitrogen and sulfur--regardless of how they
are processed. Once the materials are applied
and become wet, they are colonized by bacteria and fungi; the materials then decompose
and emit noxious odors in the form of
organic amines, organic sulfides, and other
small-molecular-weight compounds.
Offensive odors that form as sludge biologically decomposes in the field indicate
pathogen regrowth because they are produced
as bacteria break down proteins and other
organic compounds containing nitrogen and
sulfur. Most treatment methods produce
sludges that are only temporarily stable; that
is, the sludges produce noxious odors from
biological decomposition after they are
applied in the field.
One commercial process achieves longterm stability by chemically reacting sludge
under heat and pressure at high pH to drive
off organic nitrogen as ammonia (Reimers
et al. 2003). With this process, the combination of gaseous ammonia, high temperature,
and pressure effectively eliminates a wide
range of pathogens. The final wet product,
which is odorless and has a high pH, is used to
amend acidic soils. Because the nitrogen content is driven off, however, the product lacks
nutrient value.
Bacterial Toxins
Most bacteria found in sewage sludge produce
either endotoxins or exotoxins, both of which
can cause severe illness or death. As sludges
decompose, toxins can leach into groundwater,
enter surface water runoff, and be carried away
in airborne dusts. Considering that tons of
decomposing sewage sludge per acre are often
applied to hundreds or thousands of acres
many times a year, land-application sites have a
potential for producing and exporting large
quantities of toxins.
Exotoxins--proteins and peptides secreted
into the surrounding environment by growing
cells--are produced by both gram-negative
and gram-positive bacteria. They are usually
the most toxic of the two general types of bacterial toxins. Because they can retain their toxicity at extremely high dilutions, some
exotoxins, including staphylococcal enterotoxins and shigatoxin, are used as biological
warfare agents.
Although exotoxins are generally heat labile
and could therefore be destroyed by heat-treatment processes for sewage sludges, treated
sludges are still likely to become contaminated
with E. coli, Pseudomonas auruginosa, and other
exotoxin-producing bacteria in the field. Severe
gastrointestinal illnesses reported by individuals
using private wells near land-application sites
may have been caused by exotoxins leaching
into groundwater.
The same property that makes S. aureus a
common cause of food poisoning--its ubiquitous presence--may also make it one of the
more common pathogens to proliferate in
sewage sludges after they are applied to land.
The organism produces an exotoxin that is not
destroyed by cooking. Symptoms caused by
S. aureus food poisoning (e.g., nausea, cramps,
vomiting) are due to the presence of this toxin.
Land-application sites with high levels of S.
aureus could contaminate air and water with
potentially harmful levels of both the organism
and its toxin.
Endotoxins, on the other hand, are
lipopolysaccharide complexes in the cell walls
of gram-negative bacteria only. They are associated with proteins and other components of
the cell walls and are released when the bacteria
die and cell walls break apart (Rylander 1995).
Endotoxins are produced in large quantities
when wastes colonized with gram-negative
bacteria are treated (Sigsgaard et al. 1994).
They would also be produced as gram-negative bacteria growing in nutrient-rich sludges
die off in the field.
Unlike most exotoxins, endotoxins are
heat stable even upon autoclaving (Baines
2000). They can, however, be inactivated with
dry heat at > 200oC for 1 hr (Williams 2001).
Traces of endotoxins in food and water can
cause headaches, fever, fatigue, and severe gastrointestinal symptoms; however, their primary
target is the lungs. In addition to the former
symptoms, inhaling endotoxin-contaminated
dusts can cause acute airflow obstruction,
shock, and even death. Chronic respiratory
effects can also develop [American Conference
of Government Industrial Hygienists
(ACGIH) 1999].
Commentary | High-level disinfection of sewage sludge
Environmental Health Perspectives * VOLUME 112 | NUMBER 2 | February 2004 127
Table 1. Disinfection levels required to kill pathogens in sewage sludges.a
Group Disinfection level required
Bacterial endospores (e.g., Bacillus anthracis) High
Nonenveloped viruses (e.g., Norovirus, Coxsackie, Rotavirus) Intermediate/high
Helminths (e.g., Ascaris, Toxocara) Intermediate
Protozoa (e.g., Cryptosporidium, Giardia) Intermediate
Mycobacteria (e.g., M. tuberculosis) Intermediate
Fungi (e.g., Candida) Low/intermediate
Vegetative bacteria (e.g., Staphylococcus, Salmonella) Low
Enveloped viruses (e.g., hepatitis B, HIV, influenza) Low
Data from the Association for the Advancement of Medical Instrumentation (AAMI 1994).
aDisinfection levels are based on susceptibilities to liquid chemical germicides; groups increase similarly in resistance to heat,
with enveloped viruses being the most sensitive and bacterial endospores the most resistant.
Allergic and nonallergic reactions caused
by airborne endotoxins have been documented with exposures of 45-150 endotoxin
units (EU)/m3 and 300-400 EU/m3 (Milton
et al. 1996; Smid et al. 1994). Nearby residents exposed to dusts from land-application
sites report many of the same symptoms of
endotoxin poisoning that have been documented among sewage treatment plant workers. These include flu-like symptoms, nausea,
vomiting, diarrhea, headaches, and difficulty
breathing (Lewis et al. 2002). Rylander
(1987) proposed occupational exposure limits
to endotoxin-contaminated cotton dusts.
Based on average air concentrations over an
8- to 10-hr workday, he suggested limits
ranging from 200 EU/m3 to prevent airway
inflammation to 20,000 EU/m3 to avoid
toxic pneumonitis. The exposure levels of
endotoxin-contaminated aerosols with sewage
treatment plant workers have ranged from
80 to 4,100 EU/m3 (Liesvuori et al. 1994).
The toxins, however, have a greater effect on
people with immune systems compromised
by injury or illness (Baines 2000).
Chemical-Pathogen Interactions
Although many chemical contaminants found
in processed sewage sludges may potentially
interact with pathogens to cause, facilitate, or
exacerbate the disease process through allegeric
and nonallergic mechanisms, microbial byproducts formed during the processing and
decomposition of sewage sludge probably
account for most of the acute health effects.
Complaints among residents living near land-application sites are primarily respiratory
related and are consistent with hypersensitivity
reactions, including fever, cough, difficulty in
breathing, nausea, and vomiting.
Numerous diseases involving immunologically mediated hypersensitivity reactions have
been documented among workers exposed to
organic dusts containing microbial products.
Yi (2002) listed 27 diseases, each categorized
according to the source of the dusts and the
specific microorganisms identified as the primary cause of hypersensitivity. Sources include,
for example, dusts from molded hay, mushroom compost contaminated with fungi and
actinomycetes, Streptomyces-contaminated fertilizers, Caphaloporium-contaminated sewage,
and wood contaminated with Bacillus subtilis.
Byssinosis, perhaps the most studied of
these diseases, is attributed to traces of endotoxins from the breakdown of E. coli and
other gram-negative bacteria on raw cotton
fibers. Similarly, illnesses have been documented among wastewater treatment plant
workers exposed to endotoxins in aerosols
(Rylander 1987). Usually, the disease affects
only a small percentage of sensitive workers.
Compared with waste treatment plant
aerosols, however, endotoxin levels are probably
much higher in sewage sludge dusts, which
contain large numbers of predominantly
gram-negative bacteria killed during treatment processes and after land application.
Consequently, the frequency and severity of
hypersensitivity among groups exposed to
sewage sludge dusts may be much greater compared with exposure to other organic dusts.
Respiratory-related hypersensitivity is generally reversible when affected individuals are
removed from the source of exposure and
treated with high doses of corticosteroids.
Corticosteroids used to treat the underlying
inflammation, however, seriously impair the
immune system. In the case of sewage sludge,
this would render hypersensitive individuals
highly susceptible to infection from the low
levels of viruses, bacteria, fungi, and other
pathogens in processed sludges.
Treating residents near land-application
sites who experience hypersensitivity to
processed sewage sludges, therefore, is both
costly and risky. It would involve relocating
affected individuals to another area, making
certain that any potentially serious infections
have been eliminated or controlled with
antibiotics, then administering high doses of
corticosteroids and closely monitoring for any
new infections.
Exposure Studies
A conference on health effects of odors sponsored by Duke University and the U.S. EPA
concluded that gases and volatile emissions
from waste products including processed
sewage sludge may cause adverse health effects
(Schiffman et al. 2000). While acknowledging the complexity of the problem, the participants recommended undertaking controlled
studies of the odorous emissions. In responding to the NRC recommendations (NRC
2002), the U.S. EPA committed to measuring
field concentrations of selected volatile and
gaseous compounds at selected sites (U.S.
EPA 2003a).
Land-applied sewage sludge can emit
numerous volatile chemicals and gases that
may act alone or in combination with one
another to produce the kinds of symptoms
reported by people living near biosolids-
recycling operations. The composition of air
contaminants emitted by any land-application
site undoubtedly varies widely over space and
time, as do the susceptibilities of individuals to
the effects of these emissions. Consequently, it
is unlikely that such research will adequately
establish which components and combinations
of components can potentially cause adverse
health effects and under what conditions.
We propose an alternative approach with a
more modest goal aimed at determining the
extent to which emissions must be diluted to
eliminate malodor complaints and irritant
effects (e.g., burning eyes, coughing, breathing
difficulties). Based on meteorologic data from
local weather stations and publicly available
topographic data, the dilution of air contaminants over areas surrounding land-application
sites can be readily determined with an air-
dispersion model (Lewis et al. 2002).
Meteorologic data should be collected over
an extended period of time (e.g., 6-8 weeks)
during the maximum potential exposure, for
example, when land application is in progress,
temperatures are high, and sufficient rainfall
has occurred to support high levels of microbial activity. This approach does not require
measuring specific pollutants, and a number
of land-application sites could be studied with
a reasonable level of resources.
In addition to collecting meteorologic data,
local census data would be used in this type of
study to randomly select two groups of residents of similar demographic compositions:
one close to the land-application site (< 1 km)
and one farther away (3-5 km). Individuals in
each group would provide information on
medical histories and keep daily records of the
selected symptoms during the period when
meteorologic data are collected. Using a similar
approach, we found that residents living closer
to land-application sites were more severely
affected than those living farther away (Lewis
et al. 2002).
Quantitatively, what is needed is a simple
numerical index that captures the most important variables determining whether symptoms
develop. The amount by which volatile emissions are diluted when they reach a residence
and the number of times the dilution drops
below a certain level may be sufficient for predicting whether odor and health-related complaints are likely to develop.
For example, an exposure index could be
calculated based on a) the number of exposures
in which levels of volatile chemicals at a residence are 10% of the levels over the sludged
field and b) the average percent dilution for
these exposures. This approach is illustrated by
Equation 1, where Iv is the exposure index for
volatile emissions, n is the number of exposures
in which levels of gases and volatile chemicals
at a residence were 10% of the levels over the
sludged field, and d
-
is the average dilution
(percent) for all exposures 10% of the levels
over the field.
Iv = n * d
-
[1]
Once exposure indices and frequencies of
symptoms are collected for a number of land-application sites, a representative dilution
level required to eliminate odor complaints
and acute adverse health effects can be determined. This index provides a quantitative
measure of whether a land-application site is
likely to cause odor complaints and acute
adverse health effects at a particular location
Commentary | Gattie and Lewis
128 VOLUME 112 | NUMBER 2 | February 2004 * Environmental Health Perspectives
or distance from the site. Such information
could be used to evaluate the effectiveness of
treatment methods and land management
practices.
Public Concerns
To assess public concerns over the safety of
current land-application practices, we distributed questionnaires to 150 individuals concerned about land-application of sewage
sludges (Table 2). The group included farmers,
residents complaining of adverse health effects,
community leaders, and environmentalists.
Based on the responses of 87 respondents
from 15 states, a majority of respondents
(51.7%) desired a total ban on land application of sewage sludges, while 35.6% believed
that land application should just be suspended
until proven safe. Most respondents (74.7%)
lived near land-applications sites and most
(67.5%) reported that they had been personally affected by the practice. Overwhelming
malodor, vector attraction (flies, mosquitoes),
and adverse health effects (e.g., difficulty
breathing, chronic sinusitis) were the primary
adverse effects reported by individuals living
near the sites.
The need for additional research was
strongly supported. Respondents, however,
expressed little trust in federal and state environmental agencies to provide a reliable scientific evaluation of potential public health and
environmental effects. More confidence was
expressed if assessments were done by public
health agencies, such as the Centers for
Disease Control and Prevention (CDC) and
Commentary | High-level disinfection of sewage sludge
Environmental Health Perspectives * VOLUME 112 | NUMBER 2 | February 2004 129
Table 2. Summary of survey results from 87 respondents indicating their level of public concern about land application practices.
Topic Question Choices Percent or mean SD (n)
Background Why are you interested Live or lived near land application site 74.7% (87)
informationa in the issue of land-applied Work as a farmer/grower 4.6% (87)
sewage sludges? Engaged in environmental activism 16.1% (87)
Other 14.9% (87)
Have you ever been personally Yes 67.5% (77)
affected by land application of
sewage sludges?
How do you think land application Current practices are safe;no new restrictions are needed 0% (87)
of sewage sludges should be All land application should be completely banned 51.7% (87)
handled? Only certain land application practices should be banned 8.0% (87)
All land application should be suspended until proven safe 35.6% (87)
Land application should be continued with certain new restrictions 2.3% (87)
Other 5.7% (87)
Level of On a scale of 0 to 10 (0 = no concern; Microorganisms that may cause infection 9.7 0.9 (83)
concernb 10 = highest level of concern), indicate Chemicals, metals and microorganism that may cause cancer 9.6 1.0 (84)
your level of concern regarding the Odor-causing emissions 8.9 1.9 (84)
following issues Bacterial toxins 9.7 0.8 (82)
Property value 8.6 2.3 (82)
Other 9.7 0.7 (25)
Kinds of contamination from sludges Contamination of food supply 9.3 1.4 (83)
that cause the most concern (0 = no Contamination of water 9.9 0.7 (84)
concern; 10 = highest level Contamination of soil 9.8 0.5 (84)
of concern) Contamination of air 9.6 1.0 (84)
Other 9.8 0.5 (26)
Level of Using a scale of 0 to 10 (0 = no trust; Congress 2.0 2.3 (81)
trustb 10 = highest level of trust) U.S. EPA 1.3 2.4 (82)
indicate your level of trust in organizations U.S. Department of Agriculture 1.7 2.5 (80)
dealing with land application of State agencies 1.2 2.1 (81)
sewage sludges Local governments (city/county) 2.4 3.1 (81)
Environmental organizations 7.0 2.8 (81)
Trade groups (e.g., WEF, NEBRA) 0.8 2.1 (75)
National Biosolids Partnership 0.5 1.7 (72)
Industry 0.5 1.5 (80)
Otherc (e.g., departments of health, independent scientists) 5.3 4.4 (19)
Need for On a scale of 0 to 10 (0 = don't feel that 9.6 1.7 (84)
additional more research is needed; 10 = feel very
researchb strongly that more research is needed),
indicate how strongly you feel that more
scientific research is needed before we
will know whether land applying sewage
sludges is safe for public health and
the environment
On a scale of 0 to 10 (0 = no trust; National Science Foundation/National Institutes of Health 5.1 3.3 (72)
10 = highest level of trust), Trade groups (WEF/WERF) 1.4 2.2 (72)
indicate your level of trust in the work U.S. EPA Office of Water 1.4 2.4 (77)
of scientists supported by various U.S. EPA Office of Research and Development 2.8 3.1 (75)
organizations dealing with land U.S. Department of Agriculture 1.8 2.6 (76)
application of sewage sludges Centers for Disease Control and Prevention 4.9 3.4 (75)
Industry 0.5 1.4 (76)
State agencies 1.2 2.1 (78)
Otherc (e.g., universities, independent scientists, environmental groups) 5.9 4.7 (19)
Abbreviations: NEBRA, New England Biosolids and Residuals Association; WEF, Water Environment Federation; WERF, Water Environment Research Federation. Responses not following
survey instructions were omitted.
aBased on yes/no responses from all 87 respondents; values shown are percent SD (n). bBased on a 0-10 scale, with averages determined from the actual number of responses to
each category; values shown are mean SD (n). cCategories given by respondents.
the National Institutes of Health (NIH).
Although few respondents (16.1%) reported
engaging in environmental activism, the group
rated environmentalist organizations as being
the most trustworthy to assess the safety of
land-application practices.
Overall, the survey results indicate that
most people concerned about sewage sludge do
not believe land application is inherently
unsafe but object to the practice because they
lack confidence in scientific studies funded by
government and industry groups defending the
status quo. By contrast, survey respondents
indicated a greater level of confidence in studies of land-application practices if done by the
CDC or the NIH. It appears, therefore, that
overcoming opposition through additional scientific research will require strong involvement
with respected public health organizations. It
will also require supportive findings from
researchers independent of the federal agencies
and trade organizations that have historically
overseen the development and marketing of
land-application practices.
Discussion
Politics of land application. Since the 503
sludge rule (U.S. EPA 1993) was promulgated
in 1993, the U.S. EPA, the U.S. Department of
Agriculture, and the industry and its trade associations have vigorously defended the rule as
fully protective of public health and the environment. The primary basis has been a lack of
documented cases of illnesses and the results of
research supported with congressional funds
earmarked for promoting land application as
safe and beneficial (U.S. EPA 2002).
In 2000, the Committee on Science in the
U.S. House of Representatives held hearings
into allegations that the U.S. EPA retaliates
against scientists and private citizens who
report adverse environmental and health
effects associated with sewage sludge (U.S.
House of Representatives 2000a, 2000b).
During the hearings, the U.S. EPA Office
of the Inspector General released a report confirming that concerns were widespread among
U.S. EPA scientists who reviewed the 503 rule
(U.S. EPA 2002). The assistant administrator
refused to approve the rule without a major
commitment from the Office of Water to support additional in-house research within the
Office of Research and Development. The
Inspector General noted that this research was
never carried out.
The U.S. EPA responded to the congressional hearings by calling for a study by the
National Research Council; Congress debated,
and overwhelmingly passed, the No Fear
Act (Notification and Federal Employee
Antidiscrimination and Retaliation Act of
2002). The act, which was aimed at better protecting employees against retaliation, was
signed by President Bush last year. The NRC
published its findings and recommendations in
July 2002 (NRC 2002), and the U.S. EPA
addressed them in a research strategy in the
Federal Register earlier this year (U.S. EPA
2003a). The U.S. EPA's final response to the
NRC report is due to be released in January
2004. In the meantime, the U.S. EPA Office
of Water has provided a docket for public
comments (U.S. EPA 2003b).
Public comments to the Office of Water
docket have largely mirrored those in the survey we report in this article. There is an overall lack of confidence in the the U.S. EPA's
willingness to conduct or support objective
research in this area. As evidence, the Sierra
Club (San Francisco, CA) and others pointed
to the fact that the Office of Water intends to
address the NRC recommendations extramurally by funding the same researchers it has
historically supported with congressional
appropriations for promoting the safety of
biosolids. The NRC study (NRC 2002),
therefore, had its beginning and ending
rooted in controversy over the U.S. EPA
using congressional appropriations to support
federal policies on the beneficial reuse of
sewage sludge and to oppose scientists and
private citizens who question them.
Trends in land application. With increasing numbers of residents who live near
sludged fields reporting respiratory and gastrointestinal illnesses (Shields 2003), many local
governments have banned land application of
Class B sewage sludges. However, we found
that some Class A sludges generate the same
complaints and concluded that going to Class
A products will not resolve the pathogens issue
(Lewis and Gattie 2002). The reason for this is
that infections appear to be primarily opportunistic, following irritation of the skin,
mucous membranes, and respiratory tract by
chemical components.
Consequently, an important aspect of preventing infections lies in reducing levels of
microbial toxins and other chemicals that
cause inflammation as well as other responses
that predispose individuals to infection. As
such, the infections arise from many sources,
both community and environmental. The
problem with Class A sludges is probably primarily endotoxin related. This is because
gram-negative bacteria comprise much of the
biomass and because most conditions used to
kill bacteria in treatment processes are insufficient to break down endotoxins. The Class A
standard, therefore, while reducing the risks of
acquiring infections directly from processed
sludge, could increase risks of infections from
other environmental and community sources.
One outcome of local bans is that land
application of sewage sludge is being forced out
of areas where residents have the political and
economic resources to oppose the practice and
into economically depressed areas. Whether
this is intentional or not, sewage sludge is
being dumped more and more into those communities least able to have their complaints
heard, and where residents are least capable of
relocating or obtaining medical treatment.
The changing demographics of land
application of sewage sludge, therefore, need
to be studied. First, census data should be
used to assess the socioeconomic makeup of
communities living near land-application sites
(< 1 km away). Steps then need to be taken to
ensure that land-application practices do not
disproportionately impact low-income and
minority communities.
Recommendations
We recommend that the U.S. EPA undertake
and complete five top-priority measures by
January 2006 to address the immediate adverse
health and environmental effects associated
with land-applied sewage sludges:
* Develop a universal high-level disinfection
standard to replace the current Class A/
Class B standards, and require industry to
provide efficacy data showing treatment
methods meet this standard for sporocidal,
fungicidal, bactericidal, tuberculocidal,
virucidal, anti-protozoal, and anti-parasitic
activity
* Develop treatment methods and land management practices for reducing airborne
endotoxin levels associated with processed
sewage sludges and set limits for public
exposure. (We recommend that maximum
levels be set at 0.1 times the limit recommended for 10-hr occupational exposure
for preventing airway inflammation, which
would be 20 EU/m3.)
* Conduct a national assessment of groundwater contamination from pathogens,
microbial (bacterial, fungal) toxins, organic
chemicals, and metals at land-application
sites
* Require that industry ensure that land-application practices do not disproportionately target low-income and minority
subpopulations in rural communities
* Work with the National Biosolids
Partnership (Association of Metropolitan
Sewerage Agencies, U.S. EPA, and the
Water Environment Federation) to develop
and enforce a policy supporting open competition for research funding and prohibiting discrimination and retaliation against
individuals raising concerns over adverse
environmental and health effects from
land-applied sewage sludges.
REFERENCES
ACGIH. 1999. Bioaerosols Assessment and Controls. Cincinnati,
OH:American Conference of Government Industrial
Hygienists.
AAMI. 1994. Designing, Testing & Labeling Reusable Medical
Devices for Reprocessing in Health Care Facilities: A Guide
Commentary | Gattie and Lewis
130 VOLUME 112 | NUMBER 2 | February 2004 * Environmental Health Perspectives
Commentary | High-level disinfection of sewage sludge
Environmental Health Perspectives * VOLUME 112 | NUMBER 2 | February 2004 131
for Device Manufacturers. Technical Information Report
No. 12. Arlington, VA:Association for the Advancement of
Medical Instrumentation.
Baines A. 2000. Endotoxin testing. In: Handbook of Microbiological
Quality Control (Baird R, Hodges N, Denyer S, eds). New
York:Taylor & Francis, 144-167.
Berkow R, Fletcher A. 1992. Gastroenteritis: infective and toxic. In:
The Merck Manual. 16th ed. Rahway, NJ:Merck Research
Laboratories, 812-821.
Gibbs RA, Hu CJ, Ho GE, Unkovich I. 1997. Regrowth of faecal coliforms and salmonellae in stored biosolids and soil amended
with biosolids. Water Sci Technol 35:269-275.
Lewis DL. 1998. Microbes in the environment: challenges to exposure assessment. Presented at the Annual Meeting and
Science Innovation Exposition of the American Association
for the Advancement of Science, 12-17 February 1998.
Philadelphia, PA.
Lewis DL, Garrison AW, Wommack KE, Whittemore A, Steudler P,
Melillo J. 1999. Influence of environmental changes on
degradation of chiral pollutants in soils. Nature. 401:898-901.
Lewis DL, Gattie DK. 2002. Pathogen risks from applying
sewage sludge to land. Environ Sci Technol 36:286A-293A.
Lewis DL, Gattie DK, Novak M, Sanchez S, Pumphrey C. 2001.
Interactions of pathogens and irritant chemicals in
land-applied sewage sludge (biosolids). In: New Solutions,
Vol 12, No 4 (Clapp R, Orlando L, eds). Amityville,
NY:Baywood Publishing Co.
------. 2002. Interactions of pathogens and irritant chemicals in
land-applied sewage sludges (biosolids). BMC Public
Health 2:11.
Lewis DL, Shephard S, Gattie DK, Sanchez S, Novak M. 2000.
Enhanced susceptibility to infection from exposure to gases
emitted by sewage sludge: a case study. In: Proceedings of
Biosolids Management in the 21st Century, 10-11 April 2000,
College Park, MD. College Park, MD:Department of Civil &
Environmental Engineering, University of Maryland,
168-174.
Liesivuori J, Kotimaa M, Laitinen S, Louhelainen K, Ponni J,
Sarantila R, et al. 1994. Airborne endotoxin concentrations
in different work conditions. Am J Ind Med 25(1):123-124.
Milton DK, Wypij D, Kriebel D, Walters MD, Hammond SK,
Evans JS. 1996. Endotoxin exposure-response in a fiberglass manufacturing facility. Am J Ind Med 29(1):3-13.
National Research Council. 2002. Biosolids Applied to Land:
Advancing Standards and Practice. Washington,
DC:National Academy Press.
NIOSH. 2002. Guidance for Controlling Potential Risks to Workers
Exposed to Class B Biosolids. NIOSH Publication No.
2002-149. Cincinnati, OH:National Institute for Occupational
Safety and Health. Available: http://www.cdc.gov/niosh/
docs/2002-149/pdfs/2002-149.pdf [accessed 4 December
2003].
Notification and Federal Employee Antidiscrimination and
Retaliation Act of 2002. 2002. Public Law 107-174.
Reimers RS, Oleszkiewicz JA, Shepherd SL, Bakeer RM,
Fitzmorris KB. 2003. Advances in Alkaline Stabilization/
Disinfection of Agricultural and Municipal Biosolids.
Baltimore, MD:Water Environment Federation.
Rylander R. 1987. The role of endotoxin for reactions after
exposure to cotton dust. Am J Ind Med 12(6):687-697.
Rylander R. 1995. Endotoxins in the environment. In:
Lipopolysaccharides from Genes to Therapy (Levin J, Alving
C, Munford R, Redl H, eds). New York:Wiley-Liss, 79-90.
Schiffman SS, Walker JM, Dalton P, Lorig TS, Raymer JH,
Shustermann D, et al. 2000. Potential health effects of odor
from animal operations, wastewater treatment, and recycling of byproducts. J Agromed 7(1):1-81.
Shields H. 2003. Sludge Victims. Fall 2002 Update. Alton,
NH:Citizens for a Sludge-Free Land and New Hampshire
Sierra Club.
Sigsgaard T, Malmros P, Nersting L, Petersen C. 1994.
Respiratory disorders and atopy in Danish refuse workers.
Am J Respir Crit Care Med 149(6):1407-1412.
Smid T, Heederick D, Houba R, Quanjer PH. 1994. Dust- and endotoxin-related acute lung function changes and work-related
symptoms in workers in animal feed industry. Am J Ind Med
25(6):877-888.
U.S. EPA. 1993. 40 CFR Part 503. Fed Reg 58(32):9248-9415.
------. 2002. Land Application of Biosolids Status Report. 2002-S-000004. Washington, DC:U.S. Environmental Protection
Agency, Office of Inspector General.
------. 2003a. Standards for the Use or Disposal of Sewage
Sludge; Agency Response to the National Research Council
Report on Biosolids Applied to Land and the Results of
EPA's Review of Existing Sewage Sludge Regulations. Fed
Reg 68:17379-17395.
------. 2003b. Standards for the Use or Disposal of Sewage
Sludge; Agency Response to the National Research Council
Report on Biosolids Applied to Land and the Results of EPA's
Review of Existing Sewage Sludge Regulations. Docket No.
OW-2003-0006. Washington, DC:U.S. Environmental
Protection Agency. Available http://cascade.epa.gov/
RightSite/dk_public_home.htm [accessed 5 December 2003].
U.S. House of Representatives, Committee on Science. 2000a.
EPA's Sludge Rule: Closed Minds or Open Debate? No.
106-95. Washington, DC:U.S. Government Printing Office.
------. 2000b. Intolerance at EPA--Harming People, Harming
Science? No. 106-103. Washington, DC:U.S. Government
Printing Office.
Williams KL, ed. 2001. Pyrogen, endotoxin, and fever. In:
Endotoxins: Pyrogens, LAL Testing, and Depyrogenation.
2nd ed. New York:Marcel Dekker, Inc., 12-24.
Yi ES. 2002. Hypersensitivity pneumonitis. Crit Rev Clin Lab Sci
39(6):581-629.
|
16135254
|
PMC1242107
|
CC BY
|
2021-01-04 16:39:13
|
no
|
BMC Ear Nose Throat Disord. 2005 Aug 31; 5:6
|
latin-1
|
BMC Ear Nose Throat Disord
| 2,005 |
10.1186/1472-6815-5-6
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr10361616810910.1186/bcr1036Research ArticleAnticancer properties of propofol-docosahexaenoate and propofol-eicosapentaenoate on breast cancer cells Siddiqui Rafat A [email protected] Mustapha [email protected] Min [email protected] Alicia [email protected] Kevin [email protected] Gary P [email protected] William [email protected] Methodist Research Institute, Clarian Health Partners, Indianapolis, IN, USA2 Department of Medicine, Indiana University School of Medicine, Indianapolis, IN, USA3 Department of Biology, Indiana University-Purdue University, Indianapolis, IN, USA2005 7 6 2005 7 5 R645 R654 23 8 2004 24 11 2004 21 1 2005 8 4 2005 Copyright © 2005 Siddiqui et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Epidemiological evidence strongly links fish oil, which is rich in docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA), with low incidences of several types of cancer. The inhibitory effects of omega-3 polyunsaturated fatty acids on cancer development and progression are supported by studies with cultured cells and animal models. Propofol (2,6-diisopropylphenol) is the most extensively used general anesthetic–sedative agent employed today and is nontoxic to humans at high levels (50 μg/ml). Clinically relevant concentrations of propofol (3 to 8 μg/ml; 20 to 50 μM) have also been reported to have anticancer activities. The present study describes the synthesis, purification, characterization and evaluation of two novel anticancer conjugates, propofol-docosahexaenoate (propofol-DHA) and propofol-eicosapentaenoate (propofol-EPA).
Methods
The conjugates linking an omega-3 fatty acid, either DHA or EPA, with propofol were synthesized and tested for their effects on migration, adhesion and apoptosis on MDA-MB-231 breast cancer cells.
Results
At low concentrations (25 μM), DHA, EPA or propofol alone or in combination had minimal effect on cell adhesion to vitronectin, cell migration against serum and the induction of apoptosis (only 5 to 15% of the cells became apoptotic). In contrast, the propofol-DHA or propofol-EPA conjugates significantly inhibited cell adhesion (15 to 30%) and migration (about 50%) and induced apoptosis (about 40%) in breast cancer cells.
Conclusion
These results suggest that the novel propofol-DHA and propofol-EPA conjugates reported here may be useful for the treatment of breast cancer.
==== Body
Introduction
Omega-3 polyunsaturated long-chain fatty acids (ω-3 PUFAs) have been documented to inhibit or even prevent cancer. Epidemiological evidence strongly links fish oil (rich in docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA)) with low incidences of several types of cancer [1-6]. The inhibitory effects of ω-3 PUFAs on cancer development and progression are supported by studies using cultured cells and animal models [7-14]. However, the mechanisms by which ω-3 PUFAs inhibit cancer remain unclear. Of particular interest are the many reports demonstrating anticancer properties of ω-3 PUFAs on the growth and survival of various cancer cell lines cultured in vitro. Included in the growing list of affected cell lines are breast cancer cells [7,8,15-17]. Several of these reports indicate that at low concentrations ω-3 PUFAs (10 to 100 μM) produce anticancer effects through the induction of apoptosis rather than via cytotoxicity. Propofol (2,6 diisopropylphenol) is the most extensively used general anesthetic-sedative agent employed today [18,19] and is nontoxic to humans at high levels (3 to 8 μg/ml; 20 to 50 μM) [20]. Propofol is a potent antioxidant [21-24] and has been shown to stimulate protein kinase C [25,26], inhibit calcium entry in muscle cells [27] and increase the calcium sensitivity of myofilaments in ventricular myocytes [28]. Propofol is also a potent direct vasodilator and bronchodilator and has recently been shown to possess anti-inflammatory and antiseizure properties. Although the exact signaling systems responsible for these effects are unclear, it does indicate that propofol alters signaling pathways within cells.
Although most studies concerning the mode of action of propofol have concentrated on its action as an anesthetic, there are a few reports indicating that this compound may also affect cellular processes related to cancer. Clinically relevant concentrations of propofol (3 to 8 μg/ml) were reported to decrease the metastatic potential of human cancer cells, including HeLa, HT1080, HOS and RPMI-7951 cells [29]. In addition, continuous infusion of propofol inhibited pulmonary metastasis of murine osteosarcoma (LM8) cells in mice through the modulation of Rho A [29]. In HL-60 human promyelocytic leukemia cells, propofol was shown to inhibit growth and induce the formation of apoptotic bodies, increase DNA fragmentation and laddering, activate caspase-3, caspase-6, caspase-8 and caspase-9, and induce the cytosolic release of cytochrome c [30]. The conclusion from these studies was that propofol induces apoptosis through both a cell-surface death receptor (extrinsic) and the mitochondrial (intrinsic) pathway. These studies suggest that propofol possesses anticancer properties in addition to its sedative effects.
The omega-3 PUFAs DHA and EPA are natural nontoxic food substances that have interesting anticancer properties. Propofol is a widely employed nontoxic anesthetic that also has anticancer properties. Although several preparations of propofol with lipid mixtures are available for anesthetic usage (AstraZeneca, Wilmington, DE), to our knowledge propofol has not previously been covalently conjugated with fatty acids. The purpose of this study was to synthesize and investigate novel compounds composed of DHA or EPA conjugated to propofol. The conjugates were tested for their ability to inhibit breast cancer cell migration, alter adhesion to the matrix protein vitronectin and induce apoptosis. The results indicate that these novel conjugates might represent a new class of anticancer agents.
Materials and methods
Materials
MDA-MB-231 breast cancer cells were purchased from the American Type Culture Collection (ATCC; Manassas, VA). The Vybrant Apoptosis assay kit was from Molecular Probes (Eugene, OR), and DMEM, penicillin, streptomycin and glutamine were from Invitrogen Corporation (Grand Island, NY). Fetal bovine serum was from BioWhittaker (Walkersville, MD). Transwell chemotaxis chamber plates were from Corning Incorporated (Corning, NY). Cytomatrix Human Vitronectin-Coated Strips were from Chemicon International, Inc. (Temecula, CA). DHA and EPA fatty acid standards for thin-layer chromatography (TLC) and gas chromatography (GC) were from Nu-Chek Prep, Inc. (Elysian, MN). Methanol, chloroform, petroleum ether, diethyl ether, acetic acid, hexane and ethanol were from Fisher Scientific (Fair Lane, NY). Propofol, N,N-dicyclohexylcarbodiimide, 2,6-di-tert-butyl-4-methylphenol (BHT), 4-(dimethyl amino)pyridine, hematoxylin, crystal violet and all other reagents were from Sigma Chemical Co. (St Louis, MO).
Synthesis and purification of propofol-DHA and propofol-EPA
Although only propofol-DHA synthesis is described here, the procedure is analogous for the propofol-EPA conjugate. Synthesis was performed in two steps. First, docosahexaenoic acid anhydride (DHA-anhydride) was synthesized, followed by its esterification to propofol. Synthesis of the conjugate was performed under reduced light and under nitrogen to minimize auto-oxidation. In brief, DHA or EPA (0.300 mmol), a coupling reagent, N,N-dicyclohexylcarbodiimide (0.45 mmol), and an antioxidant, BHT (1.5 μM), were dissolved in 5 ml of chloroform and the reaction was stirred for 60 min at room temperature (23–25°C). To this mixture, propofol (0.29 mmol) and 4-(dimethyl amino)pyridine (0.152 mmol) were added. The mixture was stirred for a period of 12 hours; the suspension was then filtered, washed with light petroleum (38.3–53.2°C) and subjected to purification on analytical thin-layer plates (silica gel, 60 A, 0.2 mm thickness; Alltech Associates, Inc., Deerfield, IL). The plates were developed in a solvent mixture of petroleum ether and ethyl acetate (92:8, v/v) and the products were revealed with iodine vapor. The spots were compared with authentic DHA, propofol, BHT, pyridine and N,N-dicyclohexylcarbodiimide and the spot corresponding to the new compound was scraped, suspended in chloroform/methanol (20:80, v/v), passed through a glass filter and stored at -70°C. The propofol-DHA and propofol-EPA conjugates were characterized as described below.
Characterization of propofol-DHA
Characterization of the propofol-DHA conjugate was performed by a combination of techniques. First, the presence of propofol in the conjugate was assessed by UV spectroscopy with an SLM Aminco 3000 spectrophotometer. The conjugate was dissolved in ethanol and the absorption spectra was measured from 200 to 600 nm. Next, the new compound was hydrolyzed and methylated in 3 M methanolic HCl [31]. The product of the reaction was extracted with hexane/water (2:1, v/v) and the organic phase was analyzed with a Shimatzu 17A gas chromatograph with a 0.25 mm × 30 m Stabilwax capillary column (Resteck, Belfont, PA). The temperature ramp was 180 to 240°C at 3°C/min (hold 3 min), followed by 240 to 245°C at 1°C/min. To determine the presence of a possible ester group formed in the new product, an infrared spectrum was taken on a Perkin Elmer/2000FT-IR spectrometer. A thin film of the product was made by evaporation from chloroform. Finally, the molecular mass of the product was determined by mass spectrometry on a MAT95XP mass spectrometer (Thermo Electron Corp., San Jose, CA) by Dr Jonathan Karty at the mass spectrometry facility of Indiana University (Bloomington, IN) using an electrospray method.
Cell culture
MDA-MB-231 breast cancer cells were grown in DMEM containing 10% fetal bovine serum, 100 units/ml penicillin and 100 μg/ml streptomycin at a density of 106 cells/ml for routine culture. For experimental purposes, cells were cultured at the cell density indicated and treated with DHA, EPA, propofol or the conjugates. The test compounds were stored in hexane at -80°C. An aliquot of the conjugate was dried under nitrogen and the compounds were diluted in ethanol just before use. The final concentration of ethanol (less than 0.1%) in the treated cultures did not exhibit any cytotoxic effects as measured by lactate dehydrogenase release and a WST-cell proliferation assay (results not shown).
Cell growth assay
The effect of the fatty acids on cell growth was determined with a WST-1 assay in accordance with the manufacturer's instructions (Roche Biosciences, Indianapolis, IN).
Cell migration assay
Cell migration was performed with Transwell Chemotaxis Chamber Plates (Corning Inc.). The bottom chamber was supplemented with 300 μl of DMEM containing 10% fetal calf serum, and 100 μl of MDA-MB-231 cells (105/ml) in serum-free DMEM was added to the top chamber with or without (control) test compounds. The plates were incubated for 4 hours at 37°C in a humidified CO2 incubator. After incubation, the insides of the inserts were washed and cleaned with cotton swabs, and the filters were fixed with 5% formaldehyde. The cells were stained with hematoxylin as described [32] and cells that migrated through the filters were counted under a microscope on the bottom side of the filters.
Cell adhesion assay
The cell adhesion assay was performed with Cytomatrix human vitronectin-coated strips (Chemicon International, Inc., Temecula, CA). The strips were incubated with 100 μl of MDA-MB-231 cells (105/ml) at 37°C for 45 min in a CO2 incubator under serum-free conditions with control or test compounds. The wells were washed three times with PBS and the adhered cells were stained with 0.2% crystal violet in 10% ethanol for 5 min at room temperature. The excess stain was removed by washing six times with PBS. The stained cells were dissolved in 100 μl of solubilization buffer (1:1 mixture of 0.1 M NaH2PO4, pH 4.5, and 50% ethanol) and the absorbance was read at 540 nm. Absorbance of dye in the control (vehicle-treated) cells was regarded as 100% adherence and the percentage adherence of treated cells was calculated in comparison with that of the control cells.
Cell viability and apoptosis assay
A Vybrant apoptosis assay kit (Molecular Probes) was used in accordance with the manufacturer's protocol. In brief, 100 μl of MDA-MB-231 cells (105/ml) suspended in serum-free DMEM were incubated in 96-well plates at 37°C in a humidified CO2 incubator with control or test compounds. At the end of the incubation period (24 hours), reagents A (YO-PRO-1, 100 μM) and B (propidium iodide, 1 mg/ml) from the kit were added to each well (1 μl/ml) and the plates were left to incubate for 30 min on ice. The cells were revealed by using a fluorescence microscope with appropriate filters. Live cells do not exhibit any fluorescence because the dyes are impermeable to living cells, dead and necrotic cells exhibit red fluorescence, and apoptotic cells fluoresce green. Total and apoptotic cells were counted and the percentage of cells exhibiting apoptosis was calculated.
Caspase-3 activation
Activation of caspase-3 in MDA-MB-231 cells was determined with a caspase-3 activity assay kit (Oncogenes Research Products, San Diego, CA). Cells (5 × 105 per well) were grown in 48-well plates in serum-free DMEM for 24 hours at 37°C in a humidified CO2 incubator with control or test compounds. After incubation, a caspase-3 fluorescent substrate (Asp-Glu-Val-Asp-fluoromethylketone-aminotrifluoromethylcoumarin conjugate (DEVD-AFC)) was added to each well (10 μl per well) and the plates were incubated for a further 1 hour. Cells were revealed under a fluorescence microscope and pictures were taken with a MagnaFire charge-coupled device camera (Optronics, Goleta, CA).
Cytochrome c release assay
Induction of apoptosis was also assayed by detecting cytochrome c release from the apoptotic cells by western blot analysis as described [33]. Cells (5 × 105 per well) were grown in 48-well plates in serum-free DMEM for 24 hours at 37°C in a humidified CO2 incubator with control or test compounds. Cells were then homogenized in a buffer containing 10 mM HEPES, pH 7.4, 0.25 mM sucrose and 1 mM EDTA; a post-nuclear fraction was prepared by centrifugation at 2,000 g for 5 min at 4°C. The supernatant was further centrifuged at 100,000 g for 20 min at 4°C and the resultant cytosolic fraction was used for cytochrome c detection by immunoblotting. Proteins were separated by 8 to 15% SDS-PAGE and the blot was incubated with monoclonal anti-cytochrome c (BD Bioscience Pharmingen, San Diego, CA) or monoclonal antibodies against glyceraldehyde-3-phosphate dehydrogenase (1:1,000 dilution; Santa Cruz Biotech, Santa Cruz, CA) overnight at 4°C and detected with secondary anti-mouse peroxidase-conjugated antibodies (Amersham Pharmacia Biotech, Little Chalfont, Buckinghamshire, UK). The bands were detected with a chemiluminescence detection kit (Pierce, Rockford, IL). The relative distributions of cytochrome c and glyceraldehyde-3-phosphate dehydrogenase (loading control) were determined by densitometric analysis with the Kodak imaging system (Eastman Kodak Company, Rochester, NY).
Statistical analysis of data
For each experiment, means and standard errors were found for each treatment group and were plotted accordingly. Analysis of variance was performed to test for an overall effect across treatments. Individual treatments were tested against the control by using Dunnett's multiple comparison test to control the Type I experimental wise error. Analyses were conducted with SAS version 8.2 (SAS Institute, Cary, NC).
Results
Characterization of propofol-docosahexaenoate
The chemical synthesis of the propofol-DHA conjugate is shown in Fig. 1. The initial isolation and characterization of the synthetic propofol-DHA conjugate were performed with TLC. The results shown in Fig. 2 demonstrate that reaction between DHA and propofol resulted in the formation of a new product with an RF of 0.90. The RF of this product is very different from that of either DHA or propofol. The band corresponding to the new compound was scraped and subjected to spectrometric characterization. The propofol spectra showed two absorption peaks, at 216 and 274 nm; these peaks were shifted to 205 and 265 nm, respectively, for the propofol-DHA conjugate (data not shown). Next, the conjugate was hydrolyzed and then methylated for analysis by gas chromatography. The results demonstrate that the commercially available propofol (Aldrich-Sigma Chemical Co.) was 97% pure, and analysis of the hydrolyzed product of the propofol-DHA conjugate showed a DHA : propofol ratio of 1: 0.8. The ratio varied from the expected 1:1 because other isoforms of propofol were also conjugated to DHA (data not shown).
Further characterization of the conjugate was performed by infrared spectroscopy. The infrared absorption spectra of the propofol-DHA conjugate (Fig. 3) showed two broad, strong absorption bands at 1,750 and 1,250 cm-1, which are attributable to C = O and C–O bonds, respectively, indicating the presence of an ester. The band at 3,030 cm-1 is characteristic of an aromatic C–H bond (propofol), and the band at 2,800 to 2,960 cm-1 is characteristic of aliphatic C–H bonds. No O–H absorption band was seen, indicating the absence of non-esterified propofol. The presence of an ester bond, aromatic C–H absorbance, and absence of free O–H group absorbance further confirms the formation of a propofol-DHA conjugate.
The propofol-EPA conjugate was also synthesized; its characterization by TLC, UV spectroscopy, GC analysis of hydrolyzed product and infrared spectroscopy was performed in a similar fashion (data not shown).
Finally, a mass spectroscopic analysis was performed to determine the molecular mass of the product on a MAT 95XP high-resolution, high-mass-accuracy mass spectrometer. Its high mass accuracy (less than 5 p.p.m. in magnetic scan and less than 2 p.p.m. in electric scan) enables its data to be used to provide a formula match. That is, the mass spectra are precise enough to verify a chemical formula based on the sum of the mass defects from the constituent atoms. Results demonstrate that the propofol-DHA mass spectra identified a product with a parent molecular mass of 488.36 Da, which is very close to the calculated molecular mass, 488.74 Da. The observed monoisotopic molecular mass was within 0.8 part per million (0.0004 atomic mass units) of that predicted for propofol-DHA. The propofol-EPA spectra identified a product of molecular mass 462.34 Da, which is very close to the calculated molecular mass of 462.70 Da. The observed monoisotopic molecular mass was within 1.1 parts per million (0.0006 atomic mass units) of that predicted for propofol-EPA (see mass spectrometry data as additional file 1).
Characterization of the propofol-DHA conjugate and propofol-EPA conjugates suggests that the coupling reaction between DHA or EPA and propofol resulted in the formation of new products. The conjugates are a one-to-one ester of DHA or EPA and propofol.
Effect of DHA, EPA and propofol on cell growth
Results shown in Fig. 4 demonstrate that at 25 μM DHA or EPA significantly inhibited MDA-MB-231 breast cancer cell growth by 20 to 30%, whereas only at high concentration (100 μM) did propofol significantly inhibit breast cancer cell growth. Subsequent experiments were therefore performed with a concentration of 25 μM for DHA, EPA, propofol, propofol-DHA and propofol-EPA.
Effect of the conjugates on breast cancer cell migration
Because MDA-MB-231 breast cancer cells are highly invasive, we monitored the effect of the conjugates on cell migration. The results shown in Fig. 5 demonstrate that DHA, EPA and propofol alone or in combination do not have a significant effect on breast cancer cell migration; however, propofol-DHA and propofol-EPA are equally effective and inhibited cell migration by about 50% (P < 0.05).
Effect of the conjugates on breast cancer cell adhesion
We also tested the effect of the propofol-DHA and propofol-EPA conjugates on adhesion of the breast cancer cells to a vitronectin substrate. DHA and EPA alone did not significantly affect breast cancer cell adhesion, whereas propofol itself or in combination with DHA or EPA slightly increased cell adhesion (Fig. 6). In contrast, propofol-DHA and propofol-EPA significantly inhibited cell adhesion by 15% (P < 0.05) and 30% (P < 0.05), respectively.
Effect of the conjugates on breast cancer cell apoptosis
Propofol-DHA and propofol-EPA were tested for their ability to initiate apoptosis within MDA-MB-231 breast cancer cells. Induction of apoptosis was assayed by incubating the cells with DHA, EPA or propofol alone or mixtures of DHA and EPA with propofol as well as with the propofol-DHA and propofol-EPA conjugates. Results shown in Fig. 7 indicate that, at 25 μM, DHA, EPA or propofol alone induced apoptosis in only 5 to 15% of the cells. Incubating DHA or EPA with propofol did not further enhance apoptosis, whereas propofol-DHA or propofol-EPA conjugates were strongly apoptotic, inducing apoptosis in about 40% (P < 0.05) of the breast cancer cells. To confirm the apoptotic response in these cells, we further assayed for caspase-3 activity, a protease involved in the execution phase of apoptosis. The assay was performed by using a specific fluorescent caspase-3 substrate (DEVD-AFC) that on hydrolysis produces a green fluorescent product. The results shown in Fig. 8 demonstrate that cells incubated with DHA, EPA or propofol have only a few caspase-3 positive cells (2 to 5%), whereas many more cells treated with propofol-DHA or propofol-EPA conjugates exhibited caspase-3 activation (15 to 20%).
The effect of propofol-DHA and propofol-EPA conjugates on apoptosis was further analyzed by assaying for cytochrome c release (Fig. 9). DHA, EPA and propofol alone had no significant effect on cytochrome c release, whereas DHA or EPA with propofol caused an approximately 150 to 200% (P < 0.05) increase in cytochrome c release. However, propofol-DHA or propofol-EPA conjugates caused significant increases in cytochrome c release by about 300 to 400% (P < 0.05).
These results confirm that the propofol-DHA and propofol-EPA conjugates are far more effective at inducing apoptosis in breast cancer cells than are the unconjugated parent compounds DHA, EPA or propofol.
Discussion
Often a major obstacle to the successful use of a drug is its ability to be taken up and retained by cells. Either the drug must have its target on the outer membrane surface or it must cross the plasma membrane through either an existing transport system or by simple diffusion to affect intracellular targets. One approach to overcoming the problem of cell entry and retention has been to link water-soluble drugs to lipophilic carriers. Several attempts have been made in the past to synthesize novel compounds by conjugating fatty acids with drugs. For example, chlorambucil-fatty acid conjugates (Chl-fatty acid) were synthesized and tested on human lymphoma cell lines [34]. These studies found that the conjugates (including those with DHA) selectively affected neoplastic lymphocytes, with minimal effect on quiescent lymphocytes [34]. The cell toxicity observed with Chl-arachidonic acid and Chl-docosahexaenoic acid against lymphoma cells was equal to or higher than the individual toxic potential of either chlorambucil or the fatty acids, whereas the Chl-oleic acid conjugate was much less toxic than Chl alone. The authors concluded that the coupling of chlorambucil with polyunsaturated fatty acids was selective against neoplastic versus quiescent lymphocytes [34]. During the present investigation, propofol was conjugated with the omega-3 fatty acids DHA and EPA. It is possible that propofol conjugated with arachidonic acid, an omega-6 polyunsaturated fatty acid, or with a saturated fatty acid might be as effective. However, there is now a large body of evidence that DHA and EPA are far more effective than shorter and less unsaturated fatty acids [35]. Clearly additional studies are needed to establish the specificity of propofol conjugates with different fatty acids.
Similarly, DHA-paclitaxel was synthesized by Bradley and colleagues, who demonstrated that the conjugate possessed antitumor activity in mice with lung tumors [36]. In the M109 mouse tumor model, DHA-paclitaxel was less toxic to animals than paclitaxel alone and cured all tumor-bearing animals; in contrast, unconjugated paclitaxel cured none. This study indicated a limited plasma area under the drug concentration–time curve (AUC) for paclitaxel, and an increase in tumor AUC of DHA-paclitaxel administration was consistent with the increase in therapeutic index of DHA-paclitaxel relative to paclitaxel in the M109 mouse tumor model. During the present investigation we did not measure the tissue concentrations of propofol, DHA or EPA after treatment with propofol-DHA or propofol-EPA, but we plan to do so in a further study.
We have previously employed the DHA-conjugate approach to enhance the availability and hence the activity of the anticancer drug methotrexate [37]. Two phosphatidylcholines were synthesized to contain methotrexate in the sn-2 position and either stearic acid or DHA in the sn-1 position. The DHA-containing and methotrexate-containing phosphatidylcholines were more effective than conjugates containing stearic acid. Synthesis of phosphatidylcholines containing DHA and propofol is also a possibility, but such a synthesis is more cumbersome. In the study described here, we therefore used another approach to synthesize a class of novel compounds by directly conjugating DHA or EPA with propofol. Results presented in Figs 1 to 4 show that our synthetic process produced the propofol-DHA and propofol-EPA conjugates. Separations by TLC (Fig. 2), UV absorption, hydrolysis and GC analysis (not shown), infrared spectroscopy (Fig. 3) and mass spectroscopy (not shown) confirm the identity of the conjugates.
After obtaining the conjugates, we investigated their anticancer effects on breast cancer cells. MDA-MB-231 is a highly invasive breast cancer cell line that has the potential for uncontrolled proliferation and metastasis in animal models [38]. We investigated the effects of these novel conjugates on cell migration, adhesion and cancer cell death through apoptosis. Results presented in Fig. 5 show that the conjugates do indeed inhibit cell migration, whereas DHA, EPA or propofol at the same concentration has no effect. We then assayed these compounds for their effect on cell adhesion to a vitronectin substrate. Results in Fig. 6 show that DHA or EPA alone or in combination with propofol did not inhibit cancer cell adhesion, whereas the same concentrations of propofol-DHA and propofol-EPA caused significant inhibition. Cell migration and adhesion are essential processes in tumor metastasis [39]. The results of this study suggest that these novel conjugates are able to affect the metastatic potential of this breast cancer cell line by inhibiting both migration and adhesion at concentrations at which DHA, EPA or propofol alone are not effective.
We then further analyzed the effect on these conjugates on the induction of apoptosis. Results in Fig. 7 show that both conjugates are able to induce apoptosis (Vybrant apoptosis assay kit) in breast cancer cells. Induction of apoptosis was further confirmed by assaying for caspase-3 activation (Fig. 8) and cytochrome c release (Fig. 9). Cytochrome c has a function in the intrinsic pathway of apoptosis and leads to the activation of caspase-3, which is a downstream enzyme in the apoptosis process and is involved in the execution phase of the death pathway [40]. Release of cytochrome c and activation of caspase-3 by propofol-DHA and propofol-EPA conjugates confirm that these compounds induce a cell signaling pathway for apoptosis that eventually leads to the death of breast cancer cells.
During the present study, the detailed molecular mechanisms by which these conjugates inhibit migration and adhesion and induce apoptosis were not investigated. Our previous studies [41-43] and other reports [44,45] have shown that DHA is rapidly taken up by cells and incorporated into membrane phospholipids. Propofol, being a partly lipophilic agent, also interacts with cellular membranes [46]. However, because of its volatile nature, it is rapidly removed from membranes and has a very short half-life in the circulation [19]. It is possible that these conjugates provide a mechanism whereby propofol can be retained in cell membranes for a longer duration and therefore enhance its anticancer effects. Studies have shown that the incorporation of DHA into membranes results in reorganization and the formation of membrane microdomains [47]. These conjugates might therefore influence cell-signaling pathways involved in cell migration, adhesion and apoptosis.
We have previously shown that DHA induces apoptosis in Jurkat leukemic cells by activating protein phosphatases [48,49]. We have not yet tested this possibility with the conjugates; however, it is likely that these conjugates also activate protein phosphatases, inducing apoptosis. Further investigation is required to explore the molecular mechanisms by which propofol-DHA and propofol-EPA conjugates affect the growth, migration and adhesion of breast cancer cells. We also plan to investigate the effect of the conjugates on other cancer cell lines. Importantly, our preliminary observations indicate that similar concentrations of these conjugates do not induce any cytotoxic effects in normal skeletal muscle cells, cardiomyocytes or hepatocyte cell lines from rat (data not shown). Further studies are required to test these compounds on normal human cell lines.
Conclusion
Our synthesis has yielded two novel lipid compounds, namely propofol-DHA and propofol-EPA. These conjugates exhibit anticancer effects that include the inhibition of cell migration and adhesion and the induction of apoptosis within MDA-MB-231 breast cancer cells. The conjugates are more active than the parent compounds and possess unique anticancer activities not found with the latter (namely inhibition of adherence). These conjugates were not tested in other breast cancer cell lines or other cancers of different anatomical locations. Experiments are under way to test these conjugates on different cancer cells lines and also in model systems in vivo.
Abbreviations
BHT = 2,6-di-tert-butyl-4-methylphenol; Chl = chlorambucil; DEVD-AFC = Asp-Glu-Val-Asp-fluoromethylketone-aminotrifluoromethylcoumarin conjugate; DHA = docosahexaenoic acid; DMEM = Dulbecco's modified Eagle's medium; EPA = eicosapentaenoic acid; GC = gas chromatography; ω-3 PUFAs = omega-3 polyunsaturated fatty acids; TLC = thin-layer chromatography.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RAS, GZ, and SW designed conjugates, supervised experiments and contributed in manuscript writing. MZ led the propofol-DHA and propofol-EPA synthesis and characterization. MW, AC, and KH were responsible for laboratory work for adhesion, migration and apoptosis assays. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
A Word document containing mass spectrometry data for propofol-DHA and propofol-EPA conjugates are presented as additional file 1.
Click here for file
Acknowledgements
We thank Dr Bruce Young and Dr JJ Breen (Department of Chemistry, Indiana University-Purdue University, Indianapolis) and Dr Jonathan Karty (Department of Chemistry, Indiana University, Bloomington) for their help in infrared and mass spectroscopy studies, Ms Heather Richardson and Charlene Shaffer for secretarial assistance, Mr Colin Terry for statistical analysis of the data, and Dr Karen Spear for editorial assistance.
Figures and Tables
Figure 1 Chemical synthesis of the propofol-docosahexaenoic acid (propofol-DHA) conjugate. Synthesis was performed in two steps. First, docosahexaenoic acid anhydride (DHA-anhydride) was synthesized with a coupling reagent, N,N-dicyclohexylcarbodiimide (DCC); subsequently, conjugation of propofol with DHA was performed in the presence of 4-(dimethylamino)pyridine (DMAP).
Figure 2 Thin-layer chromatography (TLC) analysis of propofol-docosahexaenoic acid (propofol-DHA) conjugate synthesis. Synthesis of propofol-DHA was performed as described in the text. The product of the reaction was separated by TLC with a light petroleum (38.3–53.2°C)/ethyl ether (98:2, v/v) solvent system and the products were revealed with iodine vapors. Lane 1, product of the reaction mixture that contained all starting material except propofol; lane 2, product of the complete reaction mixture; lane 3, propofol and DHA standards.
Figure 3 Characterization of propofol-docosahexaenoic acid conjugate by infrared spectroscopy. The infrared spectrum of the conjugate was recorded on a Perkin Elmer/2000 FT-IR. The sample was run as a thin film after evaporation of solvent. The broad, strong absorption bands at 1,750 cm-1 and 1,250 cm-1 are attributable to C=O and C–O bonds, respectively, and indicate the presence of an ester. The band at 3,030 cm-1 is characteristic of an aromatic C–H (propofol) and the band at 2,800 to 2,960 cm-1 is characteristic of aliphatic C–H bonds. No O–H absorption band was seen, indicating the absence of nonesterified propofol.
Figure 4 Effect of docosahexaenoic acid (DHA), eicosapentaenoic acid (EPA) and propofol on cellular growth. Cells (104 per well) were seeded overnight in a 96-well plate and then treated for 24 hours with various concentrations of DHA, EPA or propofol in serum-free medium. Cell growth was assayed with a WST-1 assay as described in the Materials and methods section. Results are means ± SEM for three experiments. The results were analyzed by analysis of variance and Dunnett's multiple comparison test to control the Type I experimental wise error. Significant differences from the control (P < 0.05) are indicated with an asterisk.
Figure 5 Effect of propofol-docosahexaenoic acid (propofol-DHA) and propofol-eicosapentaenoic acid (propofol-EPA) conjugates on cell migration. MDA-MB-231 cells (104) were incubated for 4 hours with 25 μM DHA (D), 25 μM EPA (E), 25 μM propofol (P), 25 μM each of D+P, 25 μM each of E+P, and 25 μM propofol-DHA (D-P) or 25 μM propofol-EPA (E-P) using transwell plates. The control cells (C) were treated with equal amounts of ethanol. Cells that migrated through the filter were counted under a microscope as described in the text. Results are means ± SEM for four experiments. The results were analyzed by analysis of variance and Dunnett's multiple comparison test to control the Type I experimental wise error. Significant differences from the control (P < 0.05) are indicated with an asterisk.
Figure 6 Effect of propofol-docosahexaenoic acid (propofol-DHA) and propofol-eicosapentaenoic acid (propofol-EPA) conjugates on cell adhesion. MDA-MB-231 cells (104) were incubated in vitronectin-coated plates in the presence of 25 μM DHA (D), 25 μM EPA (E), 25 μM propofol (P), 25 μM each of D+P, 25 μM each of E+P, and 25 μM propofol-DHA (D-P) or 25 μM propofol-EPA (E-P) for 45 min. The control cells were treated with equal amounts of ethanol. Cells adhering to the plates were quantified by staining and measurement of the absorbance in a spectrophotometer as described in the text. Results are means ± SEM for four experiments and are presented as the percentages of adhered cells in comparison with the controls. The results were analyzed by analysis of variance and Dunnett's multiple comparison test to control the Type I experimental wise error. Significant differences from the control (P < 0.05) are indicated with an asterisk.
Figure 7 Effect of propofol-docosahexaenoic acid (propofol-DHA) and propofol-eicosapentaenoic acid (propofol-EPA) conjugates on breast cancer cell apoptosis. MDA-MB-231 cells (104) were incubated with 25 μM DHA (D), 25 μM EPA (E), 25 μM propofol (P), 25 μM each of D+P, 25 μM each of E+P, 25 μM propofol-DHA (D-P) or 25 μM propofol-EPA (E-P) for 24 hours. The control cells (C) were treated with equal amounts of ethanol. The apoptotic cells were stained by using the Vybrant Apoptosis assay kit as described in the text. Results are means ± SEM for four experiments. The results were analyzed by analysis of variance and Dunnett's multiple comparison test to control the Type I experimental wise error. Significant differences from the control (P < 0.05) are indicated with an asterisk.
Figure 8 Effect of propofol-docosahexaenoic acid (propofol-DHA) and propofol-eicosapentaenoic acid (propofol-EPA) conjugates on caspase-3 activation. MDA-MB-231 cells (5 × 105 per well) were grown in 96-well plates for 24 hours at 37°C in a humidified CO2 incubator with control or test compounds. Cells containing activated caspase-3 fluoresced green as the result of cleavage of a fluorogenic substrate (Asp-Glu-Val-Asp-fluoromethylketone-aminotrifluoromethylcoumarin conjugate).
Figure 9 Effect of propofol-docosahexaenoic acid (propofol-DHA) and propofol-eicosapentaenoic (propofol-EPA) conjugates on cytochrome c release. MDA-MB-231 cells (0.5 × 106/well) were grown in 96-well plates for 24 hours at 37°C in a humidified CO2 incubator with control or test compounds. (a) Postnuclear supernatant was used for the determination of cytochrome c release using immunobloting as described in the Materials and Methods section. The relative distributions of cytochrome c and glyceraldehyde-3-phosphate dehydrogenase (loading control) were determined. (b) Quantification was by densitometry analysis with the Kodak imaging system. Results are means ± SEM for three experiments. The results were analyzed by analysis of variance and Dunnett's multiple comparison test to control the Type I experimental wise error. Significant differences from the control (P < 0.05) are indicated with an asterisk. D, 25 μM DHA; E, 25 μM EPA; P, 25 μM propofol; D+P, 25 μM DHA plus 25 μM propofol; E+P, 25 μM EPA plus 25 μM propofol; D-P, 25 μM propofol-DHA; E-P, 25 μM propofol-EPA.
==== Refs
Bang HO Dyerberg J Plasma lipids and lipoproteins in Greenlandic west coast Eskimos Acta Med Scand 1972 192 85 94 5052396
Vogel VG McPherson RS Dietary epidemiology of colon cancer Hematol Oncol Clin North Am 1989 3 35 63 2537285
Kaizer L Boyd NF Kriukov V Tritchler D Fish consumption and breast cancer risk: an ecological study Nutr Cancer 1989 12 61 68 2710648
Schloss I Kidd MS Tichelaar HY Young GO O'Keefe SJ Dietary factors associated with a low risk of colon cancer in coloured west coast fishermen S Afr Med J 1997 87 152 158 9107220
Berg JP Glattre E Haldorsen T Hostmark AT Bay IG Johansen AF Jellum E Longchain serum fatty acids and risk of thyroid cancer: a population-based case-control study in Norway Cancer Causes Control 1994 5 433 439 7999965 10.1007/BF01694757
Terry P Lichtenstein P Feychting M Ahlbom A Wolk A Fatty fish consumption and risk of prostate cancer Lancet 2001 357 1764 1766 11403817 10.1016/S0140-6736(00)04889-3
Noguchi M Minami M Yagasaki R Kinoshita K Earashi M Kitagawa H Taniya T Miyazaki I Chemoprevention of DMBA-induced mammary carcinogenesis in rats by low-dose EPA and DHA Br J Cancer 1997 75 348 353 9020478
Rose DP Connolly JM Coleman M Effect of omega-3 fatty acids on the progression of metastases after the surgical excision of human breast cancer cell solid tumors growing in nude mice Clin Cancer Res 1996 2 1751 1756 9816126
Jenski LJ Zerouga M Stillwell W Omega-3 fatty acid-containing liposomes in cancer therapy Proc Soc Exp Biol Med 1995 210 227 233 8539260
Norrish AE Skeaff CM Arribas GL Sharpe SJ Jackson RT Prostate cancer risk and consumption of fish oils: a dietary biomarker-based case-control study Br J Cancer 1999 81 1238 1242 10584888 10.1038/sj.bjc.6690835
Calviello G Palozza P Maggiano N Franceschelli P Di Nicuolo F Marcocci ME Bartoli GM Effects of eicosapentaenoic and docosahexaenoic acids dietary supplementation on cell proliferation and apoptosis in rat colonic mucosa Lipids 1999 S111 10419112
Calviello G Palozza P Maggiano N Piccioni E Franceschelli P Frattucci A Di Nicuolo F Bartoli GM Cell proliferation, differentiation, and apoptosis are modified by n-3 polyunsaturated fatty acids in normal colonic mucosa Lipids 1999 34 599 604 10405974
Calviello G Palozza P Piccioni E Maggiano N Frattucci A Franceschelli P Bartoli GM Dietary supplementation with eicosapentaenoic and docosahexaenoic acid inhibits growth of Morris hepatocarcinoma 3924A in rats: effects on proliferation and apoptosis Int J Cancer 1998 75 699 705 9495237 10.1002/(SICI)1097-0215(19980302)75:5<699::AID-IJC7>3.0.CO;2-U
Madhavi N Das UN Effect of n–6 and n–3 fatty acids on the survival of vincristine sensitive and resistant human cervical carcinoma cells in vitro Cancer Lett 1994 84 31 41 8076361 10.1016/0304-3835(94)90355-7
Rose DP Connolly JM Antiangiogenicity of docosahexaenoic acid and its role in the suppression of breast cancer cell growth in nude mice Int J Oncol 1999 15 1011 1015 10536186
Hatala MA Rayburn J Rose DP Comparison of linoleic acid and eicosapentaenoic acid incorporation into human breast cancer cells Lipids 1994 29 831 837 7854008
Rose DP Connolly JM Effects of fatty acids and inhibitors of eicosanoid synthesis on the growth of a human breast cancer cell line in culture Cancer Res 1990 50 7139 7144 2224849
Covington H Use of propofol for sedation in the ICU Crit Care Nurse 1998 18 34 39 9814186
Miller RD Miller RD Local anesthetics: anesthesia Local Anesthetics 2000 5 New York: Churchill Livingstone 491 521
Coetzee JF Glen JB Wium CA Boshoff L Pharmacokinetic model selection for target controlled infusions of propofol. Assessment of three parameter sets Anesthesiology 1995 82 1328 1345 7793646 10.1097/00000542-199506000-00003
Eriksson O Pollesello P Saris NE Inhibition of lipid peroxidation in isolated rat liver mitochondria by the general anaesthetic propofol Biochem Pharmacol 1992 44 391 393 1322668 10.1016/0006-2952(92)90026-F
Murphy PG Myers DS Davies MJ Webster NR Jones JG The antioxidant potential of propofol (2,6-diisopropylphenol) Br J Anaesth 1992 68 613 618 1319189
Tsuchiya M Asada A Maeda K Ueda Y Sato EF Shindo M Inoue M Propofol versus midazolam regarding their antioxidant activities Am J Respir Crit Care Med 2001 163 26 31 11208621
Aarts L van der Hee R Dekker I de Jong J Langemeijer H Bast A The widely used anesthetic agent propofol can replace alpha-tocopherol as an antioxidant FEBS Lett 1995 357 83 85 8001686 10.1016/0014-5793(94)01337-Z
Hemmings HC JrAdamo AI Effects of halothane and propofol on purified brain protein kinase C activation Anesthesiology 1994 81 147 155 8042784
Kanaya N Gable B Murray PA Damron DS Propofol increases phosphorylation of troponin I and myosin light chain 2 via protein kinase C activation in cardiomyocytes Anesthesiology 2003 98 1363 1371 12766644 10.1097/00000542-200306000-00010
Horibe M Kondo I Damron DS Murray PA Propofol attenuates capacitative calcium entry in pulmonary artery smooth muscle cells Anesthesiology 2001 95 681 688 11575542 10.1097/00000542-200109000-00022
Kanaya N Murray PA Damron DS Propofol increases myofilament Ca2+ sensitivity and intracellular pH via activation of Na+-H+ exchange in rat ventricular myocytes Anesthesiology 2001 94 1096 1104 11465603 10.1097/00000542-200106000-00026
Mammoto T Mukai M Mammoto A Yamanaka Y Hayashi Y Mashimo T Kishi Y Nakamura H Intravenous anesthetic, propofol inhibits invasion of cancer cells Cancer Lett 2002 184 165 170 12127688 10.1016/S0304-3835(02)00210-0
Tsuchiya M Asada A Arita K Utsumi T Yoshida T Sato EF Utsumi K Inoue M Induction and mechanism of apoptotic cell death by propofol in HL-60 cells Acta Anaesthesiol Scand 2002 46 1068 1074 12366500 10.1034/j.1399-6576.2002.460903.x
Eder K Reichlmayr-Lais AM Kirchgessner M Studies on the methanolysis of small amounts of purified phospholipids for gas chromatographic analysis of fatty acid methyl esters J Chromatogr 1992 607 55 67 1447360 10.1016/0021-9673(92)87054-C
Siddiqui RA English D Phosphatidylinositol 3'-kinase-mediated calcium mobilization regulates chemotaxis in phosphatidic acid-stimulated human neutrophils Biochim Biophys Acta 2000 1483 161 173 10601705
Ito A Uehara T Tokumitsu A Okuma Y Nomura Y Possible involvement of cytochrome c release and sequential activation of caspases in ceramide-induced apoptosis in SK-N-MC cells Biochim Biophys Acta 1999 1452 263 274 10590315 10.1016/S0167-4889(99)00131-7
Takahashi M Fukutake M Isoi T Fukuda K Sato H Yazawa K Sugimura T Wakabayashi K Suppression of azoxymethane-induced rat colon carcinoma development by a fish oil component, docosahexaenoic acid (DHA) Carcinogenesis 1997 18 1337 1342 9230276 10.1093/carcin/18.7.1337
Kafrawy O Zerouga M Stillwell W Jenski LJ Docosahexaenoic acid in phosphatidylcholine mediates cytotoxicity more effectively than other omega-3 and omega-6 fatty acids Cancer Lett 1998 132 23 29 10397449 10.1016/S0304-3835(98)00163-3
Bradley MO Swindell CS Anthony FH Witman PA Devanesan P Webb NL Baker SD Wolff AC Donehower RC Tumor targeting by conjugation of DHA to paclitaxel J Control Release 2001 74 233 236 11489499 10.1016/S0168-3659(01)00321-2
Zerouga M Stillwell W Jenski LJ Synthesis of a novel phosphatidylcholine conjugated to docosahexaenoic acid and methotrexate that inhibits cell proliferation Anticancer Drugs 2002 13 301 311 11984074 10.1097/00001813-200203000-00012
Connolly JM Gilhooly EM Rose DP Effects of reduced dietary linoleic acid intake, alone or combined with an algal source of docosahexaenoic acid, on MDA-MB-231 breast cancer cell growth and apoptosis in nude mice Nutr Cancer 1999 35 44 49 10624705 10.1207/S1532791444-49
Bogenrieder T Herlyn M Axis of evil: molecular mechanisms of cancer metastasis Oncogene 2003 22 6524 6536 14528277 10.1038/sj.onc.1206757
Stennicke HR Salvesen GS Properties of the caspases Biochim Biophys Acta 1998 1387 17 31 9748481
Williams EE Jenski LJ Stillwell W Docosahexaenoic acid (DHA) alters the structure and composition of membranous vesicles exfoliated from the surface of a murine leukemia cell line Biochim Biophys Acta 1998 1371 351 362 9630720
Scherer JM Stillwell W Jenski LJ Spleen cell survival and proliferation are differentially altered by docosahexaenoic acid Cell Immunol 1997 180 153 161 9341745 10.1006/cimm.1997.1187
Zerouga M Stillwell W Stone J Powner A Jenski LJ Phospholipid class as a determinant in docosahexaenoic acid's effect on tumor cell viability Anticancer Res 1996 16 2863 2868 8917399
Dratz EA Deese AJ Simopoulos AP, Kifer RR, Martin RE The role of docosahexaenoic acid (22:6ω3) in biological membranes: examples from photoreceptors and model membrane bilayers Health Effects of Polyunsaturated Fatty Acids in Seafoods 1986 Orlando, FL: Academic Press 319 351
Salem NJ Kim HY Yergey JA Simopoulos AP, Kifer RR, Martin RE Docosahexaenoic acid: membrane function and metabolism Health Effects of Polyunsaturated Fatty Acids in Seafoods 1986 Orlando, FL: Academic Press 263 317
Balasubramanian SV Campbell RB Straubinger RM Propofol, a general anesthetic, promotes the formation of fluid phase domains in model membranes Chem Phys Lipids 2002 114 35 44 11841824 10.1016/S0009-3084(01)00199-2
Stillwell W Wassall SR Docosahexaenoic acid: membrane properties of a unique fatty acid Chem Phys Lipids 2003 126 1 27 14580707 10.1016/S0009-3084(03)00101-4
Siddiqui RA Jenski LJ Neff K Harvey K Kovacs R Stillwell W Docosahexanoic acid induces apoptosis in Jurkat cells by a protein phosphatase-mediated process Biochim Biophys Acta 2001 1499 265 275 11341974 10.1016/S0167-4889(00)00128-2
Siddiqui RA Jenski LJ Wiesehan JD Hunter MV Kovacs RJ Stillwell W Prevention of docosahexaenoic acid-induced cytotoxicity by phosphatidic acid in Jurkat leukemic cells: the role of protein phosphatase-1 Biochim Biophys Acta 2001 1541 188 200 11755213 10.1016/S0167-4889(01)00143-4
|
16168109
|
PMC1242121
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jun 7; 7(5):R645-R654
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1036
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr11991616810110.1186/bcr1199Research ArticlePrime–boost vaccination with plasmid and adenovirus gene vaccines control HER2/neu+ metastatic breast cancer in mice Wang Xiaoyan [email protected] Jian-Ping [email protected] Xiao-Mei [email protected] Janet E [email protected] Heshan S [email protected] Lawrence B [email protected] Department of Experimental Therapeutics, The University of Texas MD Anderson Cancer Center, Houston, TX, USA2 Program in Immunology, The University of Texas Graduate School of Biomedical Sciences at Houston, Houston, TX, USA3 Department of Medicine, University of Louisville, Louisville, KY, USA4 Department of Cancer Biology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA2005 23 5 2005 7 5 R580 R588 11 1 2005 4 3 2005 5 4 2005 21 4 2005 Copyright © 2005 Wang et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Once metastasis has occurred, the possibility of completely curing breast cancer is unlikely, particularly for the 30 to 40% of cancers overexpressing the gene for HER2/neu. A vaccine targeting p185, the protein product of the HER2/neu gene, could have therapeutic application by controlling the growth and metastasis of highly aggressive HER2/neu+ cells. The purpose of this study was to determine the effectiveness of two gene vaccines targeting HER2/neu in preventive and therapeutic tumor models.
Methods
The mouse breast cancer cell line A2L2, which expresses the gene for rat HER2/neu and hence p185, was injected into the mammary fat pad of mice as a model of solid tumor growth or was injected intravenously as a model of lung metastasis. SINCP-neu, a plasmid containing Sindbis virus genes and the gene for rat HER2/neu, and Adeno-neu, an E1,E2a-deleted adenovirus also containing the gene for rat HER2/neu, were tested as preventive and therapeutic vaccines.
Results
Vaccination with SINCP-neu or Adeno-neu before tumor challenge with A2L2 cells significantly inhibited the growth of the cells injected into the mammary fat or intravenously. Vaccination 2 days after tumor challenge with either vaccine was ineffective in both tumor models. However, therapeutic vaccination in a prime–boost protocol with SINCP-neu followed by Adeno-neu significantly prolonged the overall survival rate of mice injected intravenously with the tumor cells. Naive mice vaccinated using the same prime–boost protocol demonstrated a strong serum immunoglobulin G response and p185-specific cellular immunity, as shown by the results of ELISPOT (enzyme-linked immunospot) analysis for IFNγ.
Conclusion
We report herein that vaccination of mice with a plasmid gene vaccine and an adenovirus gene vaccine, each containing the gene for HER2/neu, prevented growth of a HER2/neu-expressing breast cancer cell line injected into the mammary fat pad or intravenously. Sequential administration of the vaccines in a prime–boost protocol was therapeutically effective when tumor cells were injected intravenously before the vaccination. The vaccines induced high levels of both cellular and humoral immunity as determined by in vitro assessment. These findings indicate that clinical evaluation of these vaccines, particularly when used sequentially in a prime–boost protocol, is justified.
==== Body
Introduction
Once metastasis has occurred, the possibility of completely curing breast cancer is unlikely [1], particularly for the 30 to 40% of cancers overexpressing the gene for HER2/neu [2]. A vaccine targeting HER2/neu could have considerable preventive and therapeutic application by controlling the growth and metastasis of highly aggressive HER2/neu+ tumor cells [3,4]. Gene vaccines, which are bacterial expression plasmids encoding the DNA sequence for tumor antigens, have induced strong antitumor immunity in animals [5]. Although gene vaccines have shown effectiveness in clinical trials [6-13], only a few trials have been completed in oncology patients, and the results have been mixed [14-19]. However, it was recently demonstrated that a gene vaccine for prostate-specific antigen broke immunologic tolerance and induced cellular immunity [20]. Immunotherapy for cancer using gene vaccines is in its infancy, and the development of new approaches and techniques is anticipated [21-24].
The fields of gene therapy and gene vaccines have recently converged, as shown by the use of identical delivery systems for both purposes [25]. Alphaviruses such as the Sindbis, Semliki Forest, and Venezuelan equine encephalitis viruses may be used for both gene therapy and gene vaccines [26-28]. We [29] and others [30] have shown that plasmids containing Sindbis virus genes induce excellent antitumor immunity in murine tumor models. Adenoviruses, the workhorse of gene replacement therapy, are quickly moving to the forefront as gene vaccine vectors [31-35]. Unlike alphaviruses, which contain self-replicating RNA [36], adenoviruses contain DNA as their genetic material. For both families of viruses, essential genes for replication or packaging are deleted or mutated to ensure the safety of the gene therapy or gene vaccine vector.
Although animal models of tumor growth are limited in their ability to represent clinical cancer, models can provide valuable information about drug candidates. We have used both preventive and therapeutic murine tumor models to evaluate the effectiveness of two gene vaccines. Our results demonstrated that each gene vaccine was effective in prevention models, but neither was effective when used in a therapeutic model. However, prime–boost vaccination with SINCP-neu followed by Adeno-neu significantly prolonged the overall survival rate when used therapeutically in a murine model of breast cancer metastasis. This finding indicates that effective vaccine immunotherapy may require treatment with at least two gene vaccines delivered in a precise order.
Materials and methods
Tumor cell line and reagents
The A2L2 cell line, which expresses high levels of rat HER2/neu, was derived from the mouse tumor cell line 66.3 in our laboratory [29]. The A2L2 cell line has expressed high levels of HER2/neu for more than 5 years and consistently induces tumors in BALB/c mice when injected into a mammary fat pad or intravenously (i.v.). The line was maintained in Eagle's minimal essential medium containing 5% fetal bovine serum, sodium pyruvate, nonessential amino acids, L-glutamine, and vitamins (Gibco, Carlsbad, CA, USA). The monolayer cultures were subdivided at approximately 75% confluence by treatment for 1 to 3 min with 0.25% trypsin and 0.02% ethylenediaminetetraacetic acid (EDTA) at 37°C.
SINCP-neu and Adeno-neu gene vaccines
The SINCP plasmid, which contains Sindbis virus genes, was provided by Dr John Polo (Chiron, Emeryville, CA, USA). SINCP is nearly identical to the ELVIS plasmid we previously tested [29] except for an internal ribosome entry site on the 5' side of the insertion point for an antigen gene. The gene for rat neu was excised from pSV2-neu and inserted into SINCP by standard recombinant DNA techniques [29]. The correct insertion of the complete neu gene was confirmed by sequence analysis. The E1,E2a-deleted adenovirus (Adeno) has been described in detail [37,38]. The gene for rat neu was excised from pSV2-neu and incorporated into Adeno as previously described [38].
Flow cytometric analysis
To evaluate the level of immunoglobulin G (IgG) in the serum of vaccinated mice, we incubated A2L2 cells for 1 hour at 37°C with either immune serum or control serum diluted 1:100 in PBS, pH 7.2. Fluorescein-isothiocyanate-labeled (FITC-labeled) goat antimouse IgG (Pharmingen, San Diego, CA, USA) diluted 1:1000 in PBS was added to the cells, and incubation continued for 1 hour at 37°C. The cells were washed three times in PBS and analyzed by flow cytometry using an EPICS Profile Analyzer (Coulter, Hialeah, FL, USA). As a positive control, A2L2 cells were stained with a 1:1000 dilution of a polyclonal rabbit antibody against p185 (sc-284; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and an FITC-labeled goat anti-rabbit IgG.
Mice
Female BALB/c mice 6 to 8 weeks old were obtained from the National Cancer Institute-Frederick Cancer Research Facility (Frederick, MD, USA). Mice were acclimated for at least 1 week before use. Our small-animal facility is approved by the Association for the Assessment and Accreditation of Laboratory Animal Care (AAALAC).
Vaccination of mice with SINCP-neu and Adeno-neu
SINCP-neu was prepared in milligram quantities by Aldevron (Fargo, ND, USA). Intramuscular (i.m.) injections of 0.1 ml containing 100 μg of plasmid DNA formulated with 0.25% bupivacaine (Sigma Chemical, St Louis, MO, USA) were administered to the quadriceps using a 24-gauge needle. Adeno-neu and adenovirus lacking an inserted gene (Adeno-empty) were suspended in 100 μl of 0.85% saline and injected into the quadriceps using a 24-gauge needle. Blood was collected at intervals from the tail vein, and the serum was separated by centrifugation after incubation at 37°C for 1 hour and overnight refrigeration.
Mammary fat pad tumor prevention model
We used a mammary fat pad tumor prevention model to assess the effect of vaccination on solid tumor development. A2L2 cells from cultures that had reached 75% confluence were harvested by treatment with 0.25% trypsin and 0.02% EDTA for 1 to 3 min at 37°C. The cells were washed once in serum-containing culture medium and once in PBS. Mice were anesthetized by inhalation of isoflurane using an apparatus developed by veterinarians at the University of Texas MD Anderson Cancer Center [39]. The fur was shaved over the lateral thorax, and a 5-mm-long incision was made to reveal mammary fat pad number 2 [40]. A 0.1-ml sample containing 2.5 × 104 A2L2 cells in normal saline was injected into the fat pad. The incision was closed with a wound clip. After 7 days, wound clips that had not already fallen off were removed. The mice were then observed daily, and their tumors were measured in perpendicular directions using a caliper. Mice with tumors at least 10 mm in the greater dimension were killed according to our approved Institutional Animal Care and Use protocol. At the end of the experiment, all mice were killed by CO2 inhalation, and all tumors were excised and weighed. Examination of the surface of the lungs during necropsy did not reveal tumor nodules.
Experimental metastasis prevention model
Although a solid tumor develops after injection of A2L2 cells into the mammary fat pad, it is not highly metastatic. The tumor induces moribundity in mice before an appreciable number of lung metastases develop. For this reason, we used an experimental lung metastasis model rather than a spontaneous lung metastasis model to assess the effect of vaccination on lung metastasis. A 0.1-ml sample containing 2.5 × 104 A2L2 cells in normal saline was injected into the tail vein of each immunized mouse. Thirty days later, the mice were killed by CO2 inhalation, and surface lung metastases in each animal were counted.
ELISPOT analysis of IFNγ production by immune spleen cells
To determine the number of interferon-γ-producing (IFNγ-producing) cells in the spleens of vaccinated mice, we used an IFNγ enzyme-linked immunospot (ELISPOT) technique (kit no. 552569; Pharmingen) and used reagents from Pharmingen whenever possible. Immune spleens were dissected from vaccinated mice and prepared exactly as previously described for tetramer analysis [41]. Wells containing only immune spleen cells served as negative controls, and spleen cells from vaccinated mice cultured with 5 μg/ml concanavalin A (Con A) (Sigma-Aldrich, St Louis, MO, USA) overnight served as a positive control. The finished plates were air-dried overnight at room temperature in the dark and sent to ZellNet Consulting (New York, NY, USA [42]), where the spots were counted automatically using an ImmunoSpot Series I analyzer (BD Bioscience, San Diego, CA, USA). If confluence (overlapping spots) was present in wells, the number of spots in a non-confluent area of that well was determined, and the following equation was used to estimate the total number of spots in each well with confluence: total spot number = spot count+2(spot count × % confluence/ [100% – %confluence]).
Statistical analysis
Student's t test was performed using Prism 4.0 GraphPad software (San Diego, CA, USA).
Results
Inhibition of solid tumor growth by vaccination with SINCP-neu and Adeno-neu
To determine the effectiveness of vaccination on solid tumor growth, two groups of 13 mice each were vaccinated once i.m. in the quadriceps with 100 μg of SINCP-neu or SINCP-βgal. Two weeks later, the mice were challenged with A2L2 cells injected into the mammary fat pad. Thirty-five days after the challenge, the mice were killed and if a solid tumor was present, its mass was determined. All of the mice vaccinated with SINCP-βgal developed tumors. In the group vaccinated with SINCP-neu, only six mice developed tumors, and the mean mass of these six tumors was significantly less than the mean mass of the tumors in the SINCP-βgal group (0.63 vs 1.02 g; P = 0.0186). When we calculated the mean mass for all 13 mice in the SINCP-neu group and compared it with the value of the SINCP-βgal group, the difference was significant (0.25 vs 1.02 g; P = 0.0001) (Fig. 1a).
Two groups of 10 mice were each vaccinated i.m. in the quadriceps once with 1 × 1010 particles of Adeno-neu or Adeno-empty. Six weeks later, the mice were challenged with A2L2 cells injected into the mammary fat pad. Thirty-five days after the challenge, the mice were killed and if a solid tumor was present, its mass was determined. All of the mice vaccinated with Adeno-empty developed tumors, whereas in the group vaccinated with Adeno-neu only 1 of 10 mice developed a tumor (mean mass 1.75 vs 0.2 g; P = 0.0001) (Fig. 1b).
Our findings show that reduced solid tumor growth is attributable to the induction of cellular immunity. This immunity is antigen specific, because it was absent in mice vaccinated with SINCP-βgal or Adeno-empty, both of which lack the neu gene.
Inhibition of experimental metastasis by vaccination with SINCP-neu and Adeno-neu
To determine the effectiveness of vaccination on experimental metastasis of A2L2 cells, two groups of five mice each were vaccinated i.m. with 100 μg of SINCP-neu or SINCP-βgal three times at 2-week intervals. Two weeks after the third vaccination, the mice were challenged with A2L2 cells injected into the tail vein. Twenty-one days after the challenge, the mice were killed, the lungs removed, and the number of tumor nodules on the surface of the lungs counted with the naked eye. All of the mice in the SINCP-βgal group had more than 25 surface lung nodules, whereas in the SINCP-neu group all of the mice had between 1 and 10 nodules (mean number of nodules 105 vs 15; P = 0.0079) (Fig. 2a).
Two groups of nine mice each were vaccinated once with Adeno-neu or Adeno-empty and 6 weeks later challenged with A2L2 cells injected into the tail vein. Thirty-five days after the challenge, the mice were killed, the lungs removed, and the number of surface lung nodules counted with the naked eye. All but one mouse in the Adeno-empty group had lung nodules, whereas in the Adeno-neu group five of nine mice had lung nodules (mean number of nodules 15 vs 1; P = 0.0078) (Fig. 2b).
Our results show that inhibition of experimental metastasis is attributable to the induction of cellular immunity. This immunity is antigen specific because it was absent in mice vaccinated with SINCP-βgal or Adeno-empty, both of which lack the neu gene.
neu-Specific IgG in the serum of mice vaccinated with SINCP-neu or Adeno-neu
To determine if vaccination induced a humoral immune response to p185, immediately before tumor challenge 0.1 ml of blood was collected from the tail vein of each of the mice treated as described for the experimental metastasis model. The sera for each group were pooled and analyzed for the presence of p185-specific IgG by flow cytometry using the A2L2 cell line as described previously [29] and in Materials and methods. Serum from mice vaccinated with SINCP-neu had a stronger IgG response to A2L2 cells than that from mice vaccinated with SINCP-βgal (Fig. 3a). Likewise, serum from mice vaccinated with Adeno-neu exhibited a stronger IgG response to A2L2 cells than that from mice vaccinated with Adeno-empty (Fig. 3b). These results show that the gene vaccines induce humoral immunity as well as the cellular immunity described above.
Therapeutic vaccination with SINCP and Adeno
Because SINCP-neu and Adeno-neu were effective as preventive vaccines delivered before tumor cell challenge, we evaluated both vaccines as therapeutic vaccines delivered after tumor cell challenge. Vaccination conditions were the same as those described for the experimental metastasis model. In both the therapeutic mammary fat pad tumor model and the therapeutic experimental metastasis model, neither vaccine prolonged the overall survival rate of mice when delivered after tumor cell challenge (data not shown). Even when administered as early as 2 days after tumor challenge, neither vaccine was effective.
Therapeutic vaccination with SINCP and Adeno using a prime–boost protocol
To determine whether the therapeutic effectiveness of the vaccines could be increased by sequential administration, we evaluated a prime–boost protocol in which the mice were primed with a single injection of SINCP 2 days after tumor cell challenge with A2L2 cells and boosted with injections of Adeno 9 and 16 days after the challenge. SINCP was always the prime and Adeno was always the boost, because numerous studies have reported that the most effective protocol is a plasmid followed by a virus [43-47]. All combinations of SINCP-neu or SINCP-βgal used as the prime and Adeno-neu and Adeno-empty used as the boost were tested in both the therapeutic mammary fat pad tumor model and the therapeutic experimental metastasis model.
In the therapeutic mammary fat pad tumor model, none of the combinations significantly increased the overall survival rate (data not shown). In the therapeutic experimental metastasis model, only one prime–boost combination was effective: the overall survival rate was significantly higher when SINCP-neu was the prime and Adeno-neu was the boost (P = 0.0004) (Fig. 4a). A comparison of mice vaccinated with SINCP-βgal as the prime showed that the overall survival rate was slightly higher when Adeno-neu was the boost than when Adeno-empty was the boost, but that this increase was not as high when neu was in both the prime and the boost (Fig. 4b). Thus, priming with SINCP-neu is essential for the effect, depends on the presence of the neu gene, and cannot be replaced by non-specific stimulation with SINCP-βgal.
neu-Specific IgG in the serum of prime–boost-vaccinated mice
We evaluated the effectiveness of prime–boost vaccination in an in vitro assay. Naive mice were primed by vaccination with SINCP-neu and boosted by vaccination twice with Adeno-neu, or they were primed by vaccination with SINCP-βgal and boosted by vaccination twice with Adeno-empty [48]. The presence of the neu gene in both the prime and the boost resulted in a strong IgG response to the A2L2 cells (Fig. 5a). The IgG response from vaccination with SINCP-βgal followed by Adeno-empty (Fig. 5a) was nearly identical to the IgG level in serum collected from the mice before vaccination (Fig. 5b). These observations indicated that prime–boost vaccination produces strong humoral immunity.
Induction of IFNγ-producing cells in the spleens of prime–boost-vaccinated mice
To determine whether prime–boost vaccination produced neu-specific T lymphocytes, we used an IFNγ ELISPOT assay to compare spleen cells of mice vaccinated with SINCP-neu as the prime and Adeno-neu as the boost with spleen cells of mice vaccinated with SINCP-βgal as the prime and Adeno-empty as the boost. Spleen cells from vaccinated mice were co-cultured overnight with A2L2 cells, and the next day the number of cells producing IFNγ was determined. Vaccination with SINCP-neu followed by Adeno-neu resulted in a mean of more than 150 IFNγ-producing cells per million spleen cells at all ratios of effector (spleen cell) to stimulator (A2L2 cells) cells, whereas vaccination with SINCP-βgal followed by Adeno-empty resulted in a mean of fewer than 20 IFNγ-producing cells per million spleen cells (Fig. 6). As a positive control and possibly as a measure, of a maximum response, spleen cells were stimulated with 5 μg/ml Con A, which resulted in a mean that was comparable to the means obtained with neu in both the prime and the boost. That the presence of neu in both the prime and the boost resulted in IFNγ production after co-culture with A2L2 cells and that this effect was not seen with vaccination with SINCP-βgal followed by Adeno-empty clearly demonstrated that prime–boost vaccination resulted in antigen-specific induction of cellular immunity.
Discussion
In this report, we showed that vaccination with a plasmid or adenovirus vaccine containing the gene for neu protected mice from challenge with breast cancer cells. In the mammary fat pad model, a single vaccination with either SINCP-neu or Adeno-neu was effective. In the experimental lung metastasis model, a single vaccination with Adeno-neu was effective, and the SINCP-neu vaccine was effective when delivered three times at 2-week intervals.
In our study, serum from mice vaccinated with SINCP-neu or Adeno-neu contained high levels of IgG that reacted with A2L2 cells, whereas serum from mice vaccinated with SINCP-βgal or Adeno-empty was at background levels. We previously showed that immune serum from gene-vaccinated mice does not react with the parental cell line (66.3) that was transfected with the neu gene to create A2L2 [29]. The level of immunoflourescence resulting from a single vaccination with Adeno-neu was one log greater than that resulting from three vaccinations with SINCP-neu. Taken together with the results described in the previous paragraph for the experimental metastasis model, this finding indicates that at these doses and this schedule, Adeno-neu is the more immunogenic vaccine. Of course, there is no way to accurately compare a plasmid vaccine with a viral vaccine except in relative terms.
Although both SINCP-neu and Adeno-neu produced robust protection in two tumor-prevention models, both gene vaccines were ineffective when used after A2L2 cells were injected into the mammary fat pad or i.v., even when vaccination was begun only 2 days after tumor cell injection and vaccination was repeated three times for SINCP-neu. Adeno-neu was injected only once in these therapeutic models. Because many publications have described the effectiveness of prime–boost vaccination in which the prime is a plasmid gene vaccine and the boost is a viral gene vaccine, and have shown that priming with a plasmid and boosting with a virus is more effective than the opposite [43,44,47], we tested this strategy in our two therapeutic tumor models.
In our experimental metastasis model, priming with SINCP-neu followed by boosting twice with Adeno-neu prolonged the lives of mice compared with prime–boost vaccination with SINCP-βgal and Adeno-empty. This effect was seen only in the experimental metastasis model and not in the experimental mammary fat pad model, and no other prime–boost combination worked. Although this combination of SINCP-neu and Adeno-neu was able to increase the survival rate, all mice eventually succumbed to tumor growth and were killed because of moribundity. We also found that prime–boost vaccination results in antigen-specific induction of both cellular and humoral immunity. A question that cannot be answered by our findings is the relative role of cellular immunity versus humoral immunity in the in vitro antitumor effect.
With the mandate to move promising laboratory findings to the clinic, it is most important to evaluate the implications of these findings with regard to treatment of patients. A gene vaccine against HER2/neu could be effective for patients with a HER2/neu-expressing tumor but could not be rationally used to manage a HER2/neu-negative tumor. Many types of HER2/neu-expressing tumors are noteworthy for their aggressive growth and high metastatic potential [50]. Our data indicate that combined vaccination in a prime–boost schedule may be the most likely to produce a clinical effect. However, these vaccines require phase I toxicity testing individually before they could be evaluated in a prime–boost protocol. A phase I/II study may yield valuable information regarding increased survival rates and induction of cellular and humoral immunity as measured by in vitro assays. Testing of either SINCP-neu or Adeno-neu in patients will require the production of clinical-grade material, which could be less of a hurdle for a plasmid than for a virus. Nonetheless, our data identify two promising gene vaccines against HER2/neu, an antigen associated with aggressive tumor growth and metastasis.
Conclusion
We demonstrated that in mice, vaccination with a plasmid gene vaccine and an adenovirus gene vaccine, each containing the gene for HER2/neu, prevented growth of a HER2/neu-expressing breast cancer cell line injected into the mammary fat pad or i.v. The gene vaccines were not effective individually in therapeutic vaccine models in which vaccination took place after tumor cells were injected. However, sequential administration of the vaccines in a prime–boost protocol was therapeutically effective when the tumor cells were injected i.v. The vaccines induced high levels of both cellular and humoral immunity as determined by in vitro assessment. Clinical evaluation of these vaccines, particularly when used sequentially, is justified.
Abbreviations
Adeno = adenovirus; Con A = concanavalin A; EDTA = ethylenediaminetetraacetic acid; ELISPOT = enzyme-linked immunospot; FITC = fluorescein isothiocyanate; i.m. = intramuscular or intramuscularly; i.v. = intravenous or intravenously; IgG = immunoglobulin G; IFNγ = interferon γ ; PBS = phosphate-buffered saline.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
The authors' contributions to this research are reflected in the order shown. XW performed all molecular biology in Dr. Lachman's laboratory, all animal experiments and all laboratory analysis; J-PW was responsible for cell culture; X-MR prepared and purified the Adeno-neu; JEP supervised the initial animal experimentation and developed the tumor model; HSZ performed all molecular biology associated with the adenovirus work and LBL supervised all aspects of the research and wrote the original manuscript. All authors read and approved the final manuscript.
Acknowledgements
This research was supported by the US Army Medical Research and Materiel Command, Department of Defense Breast Cancer Research Program grant BC980071, the WM Keck Center for Cancer Gene Therapy, and National Cancer Institute grant CA16672. We thank Galina M Kiriakova in the Department of Cancer Biology for her cooperation and outstanding technical assistance. We also thank Elizabeth L Hess of the Department of Scientific Publications for expert editing of the manuscript.
Figures and Tables
Figure 1 Protection from tumor challenge in the mammary fat pad after vaccination with SINCP or Adeno. Groups of mice were vaccinated once with SINCP-neu (a plasmid containing Sindbis virus genes and the gene for rat HER2/neu) or SINCP-βgal. (a) or with Adeno-neu (an E1,E2a-deleted adenovirus also containing the gene for rat HER2/neu) or Adeno-empty (adenovirus lacking an inserted gene) (b). Two weeks after vaccination with SINCP and 6 weeks after vaccination with Adeno, the mice were challenged with A2L2 cells injected into a mammary fat pad. Thirty-five days after the tumor challenge, the mice were killed and if a solid tumor was present, the mass of the tumor was determined. Each symbol represents one mouse. Horizontal bars indicate means.
Figure 2 Protection from intravenous tumor challenge after vaccination with SINCP or Adeno. Groups of mice were vaccinated three times at 2-week intervals with SINCP-neu (a plasmid containing Sindbis virus genes and the gene for rat HER2/neu) or SINCP-βgal. (a) or once with Adeno-neu (an E1,E2a-deleted adenovirus also containing the gene for rat HER2/neu) or Adeno-empty (adenovirus lacking an inserted gene) (b). Two weeks after the final vaccination with SINCP or 6 weeks after vaccination with Adeno, the mice were challenged with A2L2 cells injected into the tail vein. Twenty-one days after the challenge in SINCP-vaccinated animals or 35 days after the challenge in Adeno-vaccinated animals, the mice were killed and the number of tumor nodules on the surfaces of the lungs was counted. Each symbol represents one mouse. Horizontal bars indicate means.
Figure 3 Flow cytometric analysis, using A2L2 cells, of serum from mice vaccinated with SINCP or Adeno. Serum was collected 2 weeks after the third vaccination of mice with SINCP-neu (a plasmid containing Sindbis virus genes and the gene for rat HER2/neu) or SINCP-βgal. (a) or 6 weeks after a single vaccination with Adeno-neu (an E1,E2a-deleted adenovirus also containing the gene for rat HER2/neu) or Adeno-empty (adenovirus lacking an inserted gene) (b). The immune serum was diluted with PBS (1:100), and fluorescein-isothiocyanate-labeled goat antimouse IgG diluted in PBS (1:1000) was used as the secondary antibody.
Figure 4 Survival in a therapeutic metastasis model after vaccination with SINCP (prime) and Adeno (boost). Mice were given injections into the tail vein of A2L2 cells on day 0. On day 2, the mice were vaccinated with either SINCP-neu (a plasmid containing Sindbis virus genes and the gene for rat HER2/neu) or SINCP-βgal and on days 9 and 16 the mice were vaccinated with either Adeno-neu (an E1,E2a-deleted adenovirus also containing the gene for rat HER2/neu) or Adeno-empty (adenovirus lacking an inserted gene). (a) Overall survival of mice vaccinated with SINCP-neu and Adeno-neu compared with mice vaccinated with SINCP-βgal and Adeno-empty. (b) Overall survival of mice vaccinated with SINCP-neu and Adeno-neu compared with mice vaccinated with SINCP-βgal and Adeno-neu. P values were determined by a log-rank test of Kaplan–Meier survival curves.
Figure 5 Flow cytometric analysis using A2L2 cells of serum from mice vaccinated with SINCP and Adeno. Naive mice were vaccinated with SINCP-neu (a plasmid containing Sindbis virus genes and the gene for rat HER2/neu) as the prime and twice, at 2-week intervals, with Adeno-neu (an E1,E2a-deleted adenovirus also containing the gene for rat HER2/neu) as the boost; or they were vaccinated with SINCP-βgal as the prime and then twice, at 2-week intervals, with Adeno-empty (adenovirus lacking an inserted gene) as the boost. (a) Immune serum collected 2 weeks after the second Adeno vaccination was diluted 1:100 and analyzed by flow cytometry. (b) A2L2 cells treated with fluorescein-isothiocyanate-labeled goat antimouse IgG (secondary antibody) and serum collected before vaccination (prebleed) served as the negative controls, and A2L2 cells treated with a commercial polyclonal antibody against p185 served as the positive control.
Figure 6 IFNγ ELISPOT analysis of spleen cells from mice vaccinated with SINCP (prime) and Adeno (boost). Naive mice were vaccinated with SINCP-neu (a plasmid containing Sindbis virus genes and the gene for rat HER2/neu) and then twice, at 2-week intervals, with Adeno-neu (an E1,E2a-deleted adenovirus also containing the gene for rat HER2/neu). Spleen cells were collected 2 weeks after the second Adeno vaccination and were cultured overnight with 5 μg/ml concanavalin A, without stimulation, or with A2L2 cells at three different effector:stimulator ratios. The number of IFNγ-secreting cells was determined using a commercial ELISPOT procedure and antibody pair. Horizontal bars indicate means. IFNγ, interferon γ.
==== Refs
Fornier M Munster P Seidman AD Update on the management of advanced breast cancer Oncology 1999 13 647 658 10356685
Slamon DJ Godolphin W Jones LA Holt JA Wong SG Keith DE Levin WJ Stuart SG Udove J Ullrich A Studies of the HER-2/neu proto-oncogene in human breast and ovarian cancer Science 1989 244 707 712 2470152
Disis ML Grabstein KH Sleath PR Cheever MA Generation of immunity to the HER-2/neu oncogenic protein in patients with breast and ovarian cancer using a peptide-based vaccine Clin Cancer Res 1999 5 1289 1297 10389911
Calogero RA Musiani P Forni G Cavallo F Towards a long-lasting immune prevention of HER2 mammary carcinomas: directions from transgenic mice Cell Cycle 2004 3 704 706 15153804
Stevenson FK Ottensmeier CH Johnson P Zhu D Buchan SL McCann KJ Roddick JS King AT McNicholl F Savelyeva N DNA vaccines to attack cancer Proc Natl Acad Sci USA 2004 101 14646 14652 15292504 10.1073/pnas.0404896101
MacGregor RR Boyer JD Ugen KE Lacy KE Gluckman SJ Bagarazzi ML Chattergoon MA Baine Y Higgins TJ Ciccarelli RB First human trial of a DNA-based vaccine for treatment of human immunodeficiency virus type 1 infection: safety and host response J Infect Dis 1998 178 92 100 9652427
Wang R Doolan DL Le TP Hedstrom RC Coonan KM Charoenvit Y Jones TR Hobart P Margalith M Ng J Induction of antigen-specific cytotoxic T lymphocytes in humans by a malaria DNA vaccine Science 1998 282 476 480 9774275 10.1126/science.282.5388.476
Roy MJ Wu MS Barr LJ Fuller JT Tussey LG Speller S Culp J Burkholder JK Swain WF Dixon RM Induction of antigen-specific CD8+ T cells, T helper cells, and protective levels of antibody in humans by particle-mediated administration of a hepatitis B virus DNA vaccine Vaccine 2000 19 764 778 11115698 10.1016/S0264-410X(00)00302-9
Boyer JD Cohen AD Vogt S Schumann K Nath B Ahn L Lacy K Bagarazzi ML Higgins TJ Baine Y Vaccination of seronegative volunteers with a human immunodeficiency virus type 1 env/rev DNA vaccine induces antigen-specific proliferation and lymphocyte production of β-chemokines J Infect Dis 2000 181 476 483 10669329 10.1086/315229
Le TP Coonan KM Hedstrom RC Charoenvit Y Sedegah M Epstein JE Kumar S Wang R Doolan DL Maguire JD Safety, tolerability and humoral immune responses after intramuscular administration of a malaria DNA vaccine to healthy adult volunteers Vaccine 2000 18 1893 1901 10699338 10.1016/S0264-410X(99)00407-7
Tacket CO Roy MJ Widera G Swain WF Broome S Edelman R Phase 1 safety and immune response studies of a DNA vaccine encoding hepatitis B surface antigen delivered by a gene delivery device Vaccine 1999 17 2826 2829 10438052 10.1016/S0264-410X(99)00094-8
Boyer JD Chattergoon MA Ugen KE Shah A Bennett M Cohen A Nyland S Lacy KE Bagarazzi ML Higgins TJ Enhancement of cellular immune response in HIV-1 seropositive individuals: a DNA-based trial Clin Immunol 1999 90 100 107 9884357 10.1006/clim.1998.4616
Ugen KE Nyland SB Boyer JD Vidal C Lera L Rasheid S Chattergoon M Bagarazzi ML Ciccarelli R Higgins T DNA vaccination with HIV-1 expressing constructs elicits immune responses in humans Vaccine 1998 16 1818 1821 9795386 10.1016/S0264-410X(98)00180-7
Rosenberg SA Yang JC Sherry RM Hwu P Topalian SL Schwartzentruber DJ Restifo NP Haworth LR Seipp CA Freezer LJ Inability to immunize patients with metastatic melanoma using plasmid DNA encoding the gp100 melanoma-melanocyte antigen Hum Gene Ther 2003 14 709 714 12804135 10.1089/104303403765255110
Conry RM Curiel DT Strong TV Moore SE Allen KO Barlow DL Shaw DR LoBuglio AF Safety and immunogenicity of a DNA vaccine encoding carcinoembryonic antigen and hepatitis B surface antigen in colorectal carcinoma patients Clin Cancer Res 2002 8 2782 2787 12231517
Tagawa ST Lee P Snively J Boswell W Ounpraseuth S Lee S Hickingbottom B Smith J Johnson D Weber JS Phase I study of intranodal delivery of a plasmid DNA vaccine for patients with Stage IV melanoma Cancer 2003 98 144 154 12833467 10.1002/cncr.11462
Timmerman JM Singh G Hermanson G Hobart P Czerwinski DK Taidi B Rajapaksa R Caspar CB Van Beckhoven A Levy R Immunogenicity of a plasmid DNA vaccine encoding chimeric idiotype in patients with B-cell lymphoma Cancer Res 2002 62 5845 5852 12384547
Mincheff M Tchakarov S Zoubak S Loukinov D Botev C Altankova I Georgiev G Petrov S Meryman HT Naked DNA and adenoviral immunizations for immunotherapy of prostate cancer: a phase I/II clinical trial Eur Urol 2000 38 208 217 10895014 10.1159/000020281
Rosenberg SA Yang JC Restifo NP Cancer immunotherapy: moving beyond current vaccines Nat Med 2004 10 909 915 15340416 10.1038/nm1100
Pavlenko M Roos AK Lundqvist A Palmborg A Miller AM Ozenci V Bergman B Egevad L Hellstrom M A phase I trial of DNA vaccination with a plasmid expressing prostate-specific antigen in patients with hormone-refractory prostate cancer Br J Cancer 2004 91 688 694 15280930
Mocellin S Mandruzzato S Bronte V Lise M Nitti D Part I: Vaccines for solid tumours Lancet Oncol 2004 5 681 689 15522656 10.1016/S1470-2045(04)01610-9
Mocellin S Semenzato G Mandruzzato S Riccardo RC Part II: Vaccines for haematological malignant disorders Lancet Oncol 2004 5 727 737 15581544 10.1016/S1470-2045(04)01649-3
Mocellin S Mandruzzato S Bronte V Marincola FM Correspondence 1: Cancer vaccines: pessimism in check Nat Med 2004 10 1278 1279 15580242 10.1038/nm1204-1278
Lisziewicz J Trocio J Xu J Whitman L Ryder A Bakare N Lewis MG Wagner W Pistorio A Arya S Control of viral rebound through therapeutic immunization with DermaVir AIDS 2005 19 35 43 15627031 10.1097/00002030-200501030-00004
Collins MK Cerundolo V Gene therapy meets vaccine development Trends Biotechnol 2004 22 623 626 15542151 10.1016/j.tibtech.2004.10.006
Lundstrom K Alphavirus vectors for vaccine production and gene therapy Expert Rev Vaccines 2003 2 447 459 12903809 10.1586/14760584.2.3.445
Yamanaka R Alphavirus vectors for cancer gene therapy Int J Oncol 2004 24 919 923 15010830
Atkins GJ Smyth JW Fleeton MN Galbraith SE Sheahan BJ Alphaviruses and their derived vectors as anti-tumor agents Curr Cancer Drug Targets 2004 4 597 607 15578917 10.2174/1568009043332754
Lachman LB Rao XM Kremer RH Ozpolat B Kiriakova G Price JE DNA vaccination against neu reduces breast cancer incidence and metastasis in mice Cancer Gene Ther 2001 8 259 268 11393278 10.1038/sj.cgt.7700300
Leitner WW Hwang LN De Veer MJ Zhou A Silverman RH Williams BR Dubensky TW Ying H Restifo NP Alphavirus-based DNA vaccine breaks immunological tolerance by activating innate antiviral pathways Nat Med 2003 9 33 39 12496961 10.1038/nm813
Zhang L Tang Y Akbulut H Zelterman D Linton PJ Deisseroth AB An adenoviral vector cancer vaccine that delivers a tumor-associated antigen/CD40-ligand fusion protein to dendritic cells Proc Natl Acad Sci USA 2003 100 15101 15106 14645711 10.1073/pnas.2135379100
Tang Y Zhang L Yuan J Akbulut H Maynard J Linton PJ Deisseroth A Multistep process through which adenoviral vector vaccine overcomes anergy to tumor-associated antigens Blood 2004 104 2704 2713 15238426 10.1182/blood-2003-12-4319
Tatsis N Ertl HC Adenoviruses as vaccine vectors Mol Ther 2004 10 616 629 15451446 10.1016/j.ymthe.2004.07.013
Chen Y Emtage P Zhu Q Foley R Muller W Hitt M Gauldie J Wan Y Induction of ErbB-2/neu-specific protective and therapeutic antitumor immunity using genetically modified dendritic cells: enhanced efficacy by cotransduction of gene encoding IL-12 Gene Ther 2001 8 316 323 11313806 10.1038/sj.gt.3301396
Gallo P Dharmapuri S Nuzzo M Maldini D Iezzi M Cavallo F Musiani P Forni G Monaci P Xenogeneic immunization in mice using HER2 DNA delivered by an adenoviral vector Int J Cancer 2005 113 67 77 15386429 10.1002/ijc.20536
Polo JM Dubensky TW Virus-based vectors for human vaccine applications Drug Discov Today 2002 7 719 727 12110228 10.1016/S1359-6446(02)02324-3
Zhou H O'Neal W Morral N Beaudet AL Development of a complementing cell line and a system for construction of adenovirus vectors with E1 and E2a deleted J Virol 1996 70 7030 7038 8794347
Zhou HS Zhao T Rao XM Beaudet AL Production of helper-dependent adenovirus vector relies on helper virus structure and complementing J Gene Med 2002 4 498 509 12221643 10.1002/jgm.301
Popesko P Rajtova V Horak J Colour Atlas of Anatomy of Small Laboratory Animals 2003 London: Baillière Tindall
Wang X Wang J-P Maughan MF Lachman LB Alphavirus replicon particles containing the gene for HER2/neu inhibit breast cancer growth and tumorigenesis Breast Cancer Res 2005 7 R145 R155 This is not a reference and this web address is appropriate for this company. I don't know why this has a reference number next to it 15642163 10.1186/bcr962
Robinson HL Prime boost vaccines power up in people Nat Med 2003 9 642 643 12778154 10.1038/nm0603-642
Ramshaw IA Ramsay AJ The prime–boost strategy: exciting prospects for improved vaccination Immunol Today 2000 21 163 165 10740236 10.1016/S0167-5699(00)01612-1
Vuola JM Keating S Webster DP Berthoud T Dunachie S Gilbert SC Hill AV Differential immunogenicity of various heterologous prime–boost vaccine regimens using DNA and viral vectors in healthy volunteers J Immunol 2005 174 449 455 15611270
Ivory C Chadee K DNA vaccines: designing strategies against parasitic infections Genet Vaccines Ther 2004 2 17 15579202 10.1186/1479-0556-2-17
Takakura M Okuda K Matsuda T Takeshita F Takakura H Ikezawa Z Xin KQ Combination of DNA vaccine and adenovirus vector by cutaneous administration induced strong HIV-specific cellular immune responses in mice Vaccine 2005 23 847 848 15603883 10.1016/j.vaccine.2004.07.051
Hobeika AC Morse MA Osada T Ghanayem M Niedzwiecki D Barrier R Lyerly HK Clay TM Enumerating antigen-specific T-cell responses in peripheral blood: a comparison of peptide MHC tetramer, ELISpot, and intracellular cytokine analysis J Immunother 2005 28 63 72 15614046 10.1097/00002371-200501000-00008
Barouch DH Pau MG Custers JH Koudstaal W Kostense S Havenga MJ Truitt DM Sumida SM Kishko MG Arthur JC Immunogenicity of recombinant adenovirus serotype 35 vaccine in the presence of pre-existing anti-Ad5 immunity J Immunol 2004 172 6290 6297 15128818
Perez-Soler R HER1/EGFR targeting: refining the strategy Oncologist 2004 9 58 67 14755015 10.1634/theoncologist.9-1-58
|
16168101
|
PMC1242122
|
CC BY
|
2021-01-04 16:04:33
|
no
|
Breast Cancer Res. 2005 May 23; 7(5):R580-R588
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1199
|
oa_comm
|
==== Front
Breast Cancer Res
Breast Cancer Res
Breast Cancer Research : BCR
1465-5411
1465-542X
BioMed Central
bcr1200
16168103
10.1186/bcr1200
Research Article
Immunohistochemical evaluation of human epidermal growth factor receptor 2 and estrogen and progesterone receptors in breast carcinoma in Jordan
Almasri Nidal M [email protected]
Hamad Mohammad Al [email protected]
1 Department of Pathology, Jordan University of Science and Technology, and King Abdullah University Hospital, Irbid, Jordan
2005
24 5 2005
7 5 R598R604
21 12 2004
24 1 2005
10 4 2005
26 4 2005
Copyright © 2005 Almasri and Hamad et al.; licensee BioMed Central Ltd.
2005
Almasri and Hamad et al.; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licences/by/2.0)
Introduction
Although breast carcinoma (BC) is the most common malignancy affecting Jordanian females and the affected population in Jordan is younger than that in the West, no information is available on its biological characteristics. Our aims in this study are to evaluate the expression of estrogen receptor (ER) and progesterone receptor (PR) and Her-2/neu overexpression in BC in Jordan, and to compare the expression of these with other prognostic parameters for BC such as histological type, histological grade, tumor size, patients' age, and number of lymph node metastases.
Method
This is a retrospective study conducted in the Department of Pathology at Jordan University of Science and Technology. A confirmed 91 cases of BC diagnosed in the period 1995 to 1998 were reviewed and graded. We used immunohistochemistry to evaluate the expression of ER, PR, and Her-2. Immunohistochemical findings were correlated with age, tumor size, grade and axillary lymph node status.
Results
Her-2 was overexpressed in 24% of the cases. The mean age of Her-2 positive cases was 42 years as opposed to 53 years among Her-2 negative cases (p = 0.0001). Her-2 expression was inversely related to ER and PR expression. Her-2 positive tumors tended to be larger than Her-2 negative tumors with 35% overexpression among T3 tumors as opposed to 22% among T2 tumors (p = 0.13). Her-2 positive cases tended to have higher rates of axillary metastases, but this did not reach statistical significance. ER and PR positive cases were seen in older patients with smaller tumor sizes.
Conclusion
Her-2 overexpression was seen in 24% of BC affecting Jordanian females. Her-2 overexpression was associated with young age at presentation, larger tumor size, and was inversely related to ER and PR expression. One-fifth of the carcinomas were Her-2 positive and ER negative. This group appears to represent an aggressive form of BC presenting at a young age with large primary tumors and a high rate of four or more axillary lymph node metastases.
==== Body
Introduction
According to data from the Jordan National Cancer Registry [1], breast carcinoma is the most common malignant neoplasm affecting Jordanian females. Others as well as our own group have shown that females with breast carcinoma in Jordan are significantly younger than those in the West, with an average age ranging from 44.5 to 47 years [2-5]. Such data may reflect the fact that the population of Jordan is, on average, younger than in the West, but it may also suggest that breast carcinoma in Jordan has some unique biological features that need to be explored. Estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (Her-2) expression are crucially important in the biology of breast carcinoma, and yet the expression of these have not been studied in breast carcinoma in Jordan. It is known that ER and PR expression are the only predictive factors with proven usefulness in selecting patients who are likely to respond to adjuvant endocrine therapy [6,7]. Patients lacking these receptors tend to have shorter disease free survival and earlier recurrences than those expressing these receptors [6]. Her-2, otherwise known as neu or c-erbB-2, is the product of an oncogene amplified and overexpressed in 20% to 30% of breast carcinomas [8-13]. In most studies, overexpression of Her-2 is associated with adverse prognosis independent of other prognostic factors, even when multivariate analysis was used for the outcome analysis [14]. The prognostic effects of Her-2 expression appear to be stronger in node positive carcinomas than in node negative carcinomas [15]. Although Her-2 expression in breast carcinomas is associated with resistance to regimens using cyclophosphamide and methotrexate [16], it is associated with better response to regimens using doxorubicin [17]. Her-2 has gained an even greater deal of attention lately after the introduction of a humanized monoclonal antibody known as trastuzumab that can be effective in the treatment of cases in which this oncogene product is overexpressed [18-23].
Our aims in this study are to evaluate the expression of ER and PR and the overexpression of Her-2/neu in patients with breast cancer in Jordan, and to compare the expression of these with other prognostic parameters for mammary carcinomas, such as histological type, histological grade, tumor size, patients' age and number of lymph node metastases.
Materials and methods
This is a retrospective study of breast carcinoma cases diagnosed in the period January 1995 to December 1998 at the Department of Pathology, Jordan University of Science and Technology (JUST) in Irbid, Jordan. The Department of Pathology at JUST is the sole provider of histopathology services in the north of Jordan. Only cases among Jordanian females with adequate tissue to perform immunhistochemical examination were included. Of 135 cases of breast carcinomas diagnosed in this period, 91 (67%) fulfilled these criteria. The cases reflect the overall distribution of all breast carcinomas diagnosed in the same period of time.
Sections from the cases were reviewed by one of us (NMA), and the tumors were typed according to the WHO classification system [24]. For invasive ductal carcinomas, the Nottingham combined histologic system was used for grading [25]. Grade 1 carcinoma includes tumors with combined scores of 3, 4 or 5; grade 2 includes scores of 6 and 7; and grade 3 includes tumors with the scores of 8 and 9.
Sections were cut at 4 μm thicknesses, mounted onto silanized slides, and left to dry overnight at 37°C. Sections were then deparaffinized and rehydrated. Antigen retrieval was achieved by heat retrieval using a bench autoclave. Briefly, slides were placed in Coplin jars containing enough 0.01 M sodium citrate solution (pH 6.0) to cover the sections, then autoclaved at 121°C for 10 minutes in the case of Her-2, and 15 minutes for both ER and PR. Slides were incubated with 100–200 μl of primary antibodies for 30 minutes at room temperature in a moisture chamber, then rinsed in PBS. The dilution of the primary antibodies against ER and PR (Biogenex, San Ramon, Ca, USA) was 1:130, and for Her-2/neu (Dako, Carpintera, Ca, USA) 1:50. After washing, binding of antibodies was detected by incubation for 10 minutes with biotinylated goat anti-mouse antibody ready to use (LSAB2) from Dako; the slides were then rinsed with PBS. Sections were then incubated with streptavidin-horse radish peroxidase for 10 minutes. Finally, the sections were washed in 4 times in 4 minute changes of PBS, followed by adding 3,3 diaminobenzidine tetra hydrochloride (Biogenex) as a chromogen to produce the characteristic brown stain.
For each run of staining, a positive and negative control slide were also prepared. The positive control slides were prepared from breast carcinoma known to be positive for the antigen under study. The negative control slides were prepared from the same tissue block, but incubated with PBS instead of the primary antibody.
A semi-quantitative histochemical score was used to record results of ER and PR staining according to the system established by Allred et al. [7]. This system considers both the proportion and intensity of stained cells. The intensity score (IS) ranges from 0 to 3, with 0 being no staining, 1 weak staining, 2 intermediate staining, and 3 intense staining. The proportion score (PS) estimates the proportion of positive tumor cells and ranges from 0 to 5, with 0 being non-reacting, 1 for 1% reacting tumor cells, 2 for 10% reacting tumor cells, 3 for one-third reacting tumor cells, 4 for two-thirds reacting tumor cells, and 5 if 100% of tumor cells show reactivity. The PS and IS are added to obtain a total score (TS) that ranges from 0 to 8. Tumor cells with a total score of 3 to 8 were considered positive, whereas those with a TS less than 3 were considered negative cases.
Her-2/neu was scored on a 0 to 3 scale according to the criteria set by Dako. The staining was scored as: negative (0) when no membrane staining was observed, or when membranous staining was observed in less than 10% of the tumor cells; weak positive (1+) if weak focal membrane staining was seen in more than 10% of the tumor cells; intermediate (2+) if weak to moderate, complete membrane staining was seen in more than 10% of the tumor cells; and strongly positive (3+) if intense membrane staining with weak to moderate cytoplasmic reactivity was seen in more than 10% of the tumor cells. Figure 1 illustrates scores 1+, 2+, and 3+ as uses in this study. In the final analysis, however, scores 0 and 1 were considered negative; score 2 was considered weakly positive; and score 3 was considered strongly positive. Only score 3 cases were considered as Her-2 overexpressing cases. Fluorescence in situ hybridization (FISH) was not performed on the weak positive cases (score 2) in this study.
Figure 1 Microscopy pictures illustrating the patterns of Her-2 immunostaining in breast carcinoma. (a) Weak positive (1+) pattern exemplified by weak focal membrane staining seen in more than 10% of the tumor cells. (b) Intermediate (2+) pattern, showing weak to moderate complete membrane staining in more than 10% of the tumor cells. (c) Strongly positive (3+) pattern shows intense membrane staining with weak to moderate cytoplasmic reactivity in more than 10% of the tumor cells. Her-2 overexpressing cases comprised only those having strongly positive (3+) patterns of immunostaining.
The Student's t-test was used for comparison of mean tumor size and mean patient age for each category of cases. The chi square test was used to compare the expression of ER, PR and Her-2 among different cases, including: those above or below 50 years of age; those with tumor size up to 2 cm with those between 2 and 5 cm and those larger than 5 cm in size; and those with up to 3 nodal metastases versus those with more than 3 nodal metastases. The results were considered statistically significant if the P value was < 0.05.
Results
Of the 91 cases included in this study, infiltrating ductal carcinoma (IDC) was the largest group, accounting for 84% (76/91) of all the cases. This group included three mixed ductal and lobular cases and two cases with associated Paget's disease. The second largest group, composed of seven cases, was lobular carcinoma. The remaining cases included five cases of mucinous carcinoma, one case of ductal carcinoma in situ, one case of infiltrating ductal carcinoma with medullary features, and one case of metaplastic carcinoma. Among the infiltrating ductal carcinomas, there were only 3 grade 1 cases, 34 grade 2 cases, 37 grade 3 cases, and 2 cases were not graded. The median age of all the cases was 47.5 years ranging from 20 to 75 years. Fifty (57%) of 88 cases with known age were below 50 years of age, whereas 38 cases were 50 years or older.
The mean age of patients with positive Her-2 expression was 42 years (Table 1), which is 10 years younger than those who lack Her-2 expression. This difference is statistically significant (P = 0.0007, Student's t-test). Similarly, Her-2 expression was seen in 34% of patients less than 50 years of age as opposed to 13% in patients 50 years or older (P = 0.003, chi square test).
Table 1 Her-2 status and estrogen and progesterone receptor expression in female breast carcinoma patients in Jordan
Her-2 weak positive Her-2 strong positive Her-2 negative ER positive ER negative PR positive PR negative
Mean age (years)a 43.7 42.3 53.2 52.40 43.68 50.74 45.71
No. of patients below 50 years of age 15 17 18 21 29 24 26
No. of patients 50 years of age or older 5 5 28 26 12 22 16
Mean size (cm)b 4.9 4.72 4.02 3.57 5.3 4.06 4.79
No. of T1(up to 2 cm) tumors 4 4 7 9 6 8 7
No. of T2 (more than 2 and up to 5 cm) tumors 11 11 29 29 22 29 22
No. of T3 (more than 5 cm) tumors 6 7 7 6 14 7 13
No lymph node metastasesc 1 3 3 2 5 3 4
1–3 lymph node metastases 3 5 8 7 9 10 6
More than 3 lymph node metastases 9 10 8 13 14 13 14
aAge was unknown for three cases; bsize was unknown for five cases;
cdetermination of axillary lymph node status was limited to 50 patients who underwent lymph node dissection.
The mean age of patients with positive ER expression was 52.4 years as opposed to 43.7 among patients lacking ER expression (P = 0.001, Student's t-test). For patients less than 50 years of age 42% were ER negative, whereas 68% of patients 50 years or older were ER positive (P = 0.009, chi square test).
The mean age for PR negative cases was 45.7 years and the mean age for PR positive cases was 50.7 years (P = 0.06, Student's t-test). PR positive patients comprised 48% of patients less than 50 years of age compared to 58% of patients 50 years old or older (P = 0.36, chi square test).
Tumors with strong Her-2 expression tended to be larger than those lacking expression (scores 0 and 1), with mean sizes of 4.7 cm and 4 cm, respectively. Among patients with tumor size more than 5 cm (T3), 35% were Her-2 positive (3+) compared to 22% with tumors more than 2 and up to 5 cm in size (T2) (P = 0.13, Chi square test).
The mean size of tumors expressing ER was 3.6 cm versus 5.3 cm for those lacking ER expression (P = 0.009, Student's t-test). ER positive tumors comprised 57% of tumors between 2 and 5 cm but only 30% of those larger than 5 cm (P = 0.04, chi square test).
For PR positive cases, the mean tumor size was 4.1 cm, and the average size for PR negative cases was 4.8 cm. (P = 0.3, Student's t-test). PR positive tumors comprised 57% of tumors between 2 and 5 cm compared to 35% of those larger than 5 cm (P = 0.1, chi square test).
Lymph node status was known in 50 (55%) of the 91 cases included in this study. As shown in Table 1, 56% of the Her-2 positive cases had more than three lymph node metastases, as opposed to 42% among the Her-2 negative cases (P = 0.29, chi square test).
The fraction of ER positive cases among those with up to three lymph node metastases was 39%, slightly lower than the 48% seen among those with more than three lymph node metastases, but this difference was not statistically significant. Similarly, no correlation between lymph node metastases and PR expression was detected.
Grading analysis was limited to 74 of the 76 cases of IDC as 2 cases were not graded. Her-2 overexpression was seen in similar proportions of grade 2 and 3 breast carcinomas, 26.4% and 27% of the cases, respectively. Similarly, no correlation was detected between the grade of the tumors and expression of ER and PR (P values 0.76 and 0.32, respectively).
A negative correlation between Her-2 expression and ER and PR was noted. Of the 22 Her-2 positive cases, 82% were ER negative. On the other hand, 42% of ER negative carcinomas were Her-2 positive. A negative correlation was also seen between Her-2 and PR expression. Of the PR negative carcinomas, 68% were Her-2 positive, and 65% of Her-2 negative cases were PR positive.
On the other hand, a positive correlation between ER and PR was detected. Of the 48 ER positive cases, 36 (75%) were PR positive. Similarly, 32 (74%) of the ER negative were also PR negative.
We noticed that two distinct groups of carcinomas can be distinguished: group 1 includes those with negative Her-2 and positive ER; and group 2 includes those with positive Her-2 and negative ER (Table 2). Group 1 had a relatively much smaller tumor size (3.6 cm) and an older mean age of 55 years at the time of diagnosis. Group 2 had an average tumor size of 4.9 cm and a mean age of 41 at the time of diagnosis. The difference between these groups was statistically highly significant.
Table 2 Her-2 negative and ER positive compared to Her-2 positive and ER negative breast carcinoma cases
Her-2 negative and ER positive Her-2 positive and ER negative P value
No. of patients 35 18 -
Mean age (years) 54.88 41.2 0.0003
Mean tumor size (cm) 3.6 4.9 0.088
No. of patients without axillary node dissection 21 8 0.002
No. of patients who underwent axillary nodes dissection 14 23
No. of patients with up to 3 lymph node metastases 8 11 0.48
No. of patients with more than 3 lymph node metastases 6 12
Discussion
Breast carcinoma is a disease with a tremendous heterogeneity in its clinical behavior. Clinical and pathological variables such as tumor size, histologic grade, histologic type, lymph node metastases, vascular space invasion, tumor cell proliferation, tumor necrosis, extent of ductal carcinoma in situ, age, and pregnancy may help in predicting prognosis and the need for adjuvant therapy [24]. Newer prognostic factors and predictors of response to therapy are needed, however, to distinguish subgroups with different biological features within carcinomas that otherwise appear homogenous according to classic pathological and clinical criteria. ER, PR and Her-2 represent the most acceptable factors for predicting prognosis, response or resistance to treatment, and the potential use of newer drugs such as trastuzumab in the case of Her-2 overexpression.
In this study, we found that 22 (24%) of 91 cases were Her-2 positive. Although there is a wide variation in Her-2 overexpression and amplification, our figure appears to be within the commonly accepted rate of 20% to 30% [9,11-13,26,27]. It does appear, however, to be lower than those reported in East Asia [28,29] and in neighboring countries such as Lebanon [30] and Egypt [31]. Her-2 was expressed in 28% of the infiltrating ductal carcinoma cases compared to only 14% of our seven lobular carcinoma cases. This pattern of low Her-2 expression in lobular carcinoma is in agreement with data reported in the literature [32-35]. None of the other types of breast carcinoma showed evidence of Her-2 expression.
We found a clear negative correlation between Her-2 overexpression and age in this study. The mean age of Her-2 positive patients was 11 years less than those patients lacking Her-2 expression, a statistically significant difference. Similarly, patients younger than 50 years of age were 2.6 times more likely to overexpress Her-2 than patients 50 years of age or older (34% versus 13%). It should be pointed out that higher rates of Her-2 overexpression in young patients have been documented in previous studies [31,36,37]. Our results show a tendency of Her-2 overexpression to be more associated with larger tumor size. Tumors expressing Her-2 were on the average 0.7 cm larger than those lacking Her-2 expression, although this difference was not statistically significant. Similarly, the fraction of tumors larger than 5 cm tended to have higher rates of Her-2 expression than those 2 to 5 cm in size (35% versus 22%), but this difference was not statistically significant (P = 0.13).
Other groups have shown a direct relationship between lymph node metastases and Her-2 expression [32,38-40]. Our data reveal that 56% of Her-2 overexpressing tumors had more than three lymph node metastases, as opposed to 42% of Her-2 negative cases, although this difference was not statistically significant. We believe that the low number of cases with known nodal status is responsible for the lack of significant correlation in this study; therefore, future studies with larger numbers of patients are needed to confirm the association of Her-2 expression with nodal metastases. Similarly, we were unable to show a significant relationship between Her-2 expression and the histologic grade of breast carcinoma. Other studies concluded that Her-2 expression or amplification is associated with grade [32,36,40,41]. It should be pointed out, however, that the low number of grade 1 carcinomas (three cases) in this study would not allow us to evaluate this variable with any degree of confidence.
ER was expressed in 53% of our cases. This figure is less than the number of ER positive cases reported in the literature (60% to 70%) [7]; however, there is a wide variation in ER expression and our figure would probably fall in the lower reported ranges [42-46]. In our study, there is a strong correlation between patient age and ER expression. The mean age for ER expressing carcinoma patients was nine years older than those lacking ER expression (P = 0.001). Similarly, 68% of patients 50 years or older were ER positive as opposed to 42% of patients less than 50 years old; this difference was statistically significant (P = 0.01). These findings are in agreement with other reports in the literature, which show an association between ER expression in breast carcinoma patients and age at the time of diagnosis [42,47,48]. In our cases, we also found that ER expressing breast carcinomas were, on average, 1.6 cm smaller than carcinomas lacking ER expression (P = 0.009). Similarly, 57% of T2 tumors (2 to 5 cm) were ER positive as opposed to only 30% of T3 tumors (P = 0.04). These figures are also in agreement with data reported in the literature [38,43,49]. Our data also show a strong inverse correlation between Her-2 and ER expression, which is in agreement with data reported by others [10,39,50,51]. Unlike reported data that shows a correlation between ER expression and tumor grade [41,43,47,49], however, we were unable to confirm such a correlation in our cases. The lack of association between ER expression and lymph node status in our study supports the findings of Chariyalertsak et al. [50] who found no correlation between ER expression and lymph node status in their breast carcinomas cases.
A strong correlation between ER and PR expression was noted in our series. Unlike ER, however, there was only a low tendency for PR to be associated with smaller carcinomas and with older patients, but this tendency was not statistically significant.
Conclusion
We have shown for the first time that Her-2 is expressed in approximately one-fourth of breast carcinomas in Jordan. This expression is strongly associated with some known bad pathological and clinical prognostic factors, such as young age, large tumor size and lack of ER and PR expression. In contrast, ER expression was seen in older patients, and was associated with small tumor size and lack of Her-2 expression. One-fifth of the breast carcinoma cases in this study have a profile characterized by overexpression of Her-2 and lack of ER expression. This subgroup of patients appears to represent an aggressive form of breast carcinoma characterized by young age presentation (median 40 years), large tumor size (5.4 cm), and high rates of axillary lymph node metastases (four or more).
Abbreviations
BC = breast carcinoma; ER = estrogen receptor; Her-2 = human epidermal growth factor receptor 2; IS = intensity score; JUST = Jordan University of Science and Technology; PBS = phosphate-buffered saline; PR = progesterone receptor; PS = proportion score; TS = total score.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
NMA designed the study, reviewed the histopathology of breast carcinomas and graded all the cases histologically, interpreted the results of Her-2, ER and PR expression, and drafted the manuscript. MH prepared the histological slides and carried out the immunoperoxidase stains on the cases.
==== Refs
Jordan Cancer Registry, Ministry of Health Cancer Incidence in Jordan 1997 Report. Amman 2000
Dajani YF Al-Jitawi SA A study of 405 breast tumors in Jordanians using the revised WHO classification Trop Geogr Med 1987 39 182 186 2820095
Amr SS Breast diseases in Jordanian females: a study of 1000 cases Eur J Surg Oncol 1985 11 257 262 2993034
Kamal MF Cancer patterns in Jordan 1975–1979 Annals Chirurgiae Gunaecologiae 1987 76 191 196
Abalkhail AA Zahawi HM Almasri NM Hameed OK The role of young population structure in determining age distribution of breast cancer in Jordan J Bahrain Med Society 2003 15 28 33
Early Breast Cancer Trialists' Collaborative Group Tamoxifen for early breast cancer: an overview of the randomised trials Lancet 1998 351 1451 1467 10.1016/S0140-6736(97)11423-4 9605801
Allred DC Harvey JM Berardo M Clark GM Prognostic and predictive factors in breast cancer by immunohistochemical analysis Mod Pathol 1998 11 155 168 9504686
Slamon DJ Clark GM Wong SG Levin WJ Ullrich A McGuire WL Human breast cancer: correlation of relapse and survival with amplification of the HER-2/neu oncogene Science 1987 235 177 182 3798106
Slamon DJ Godolphin W Jones LA Holt JA Wong SG Keith DE Levin WJ Stuart SG Udove J Ullrich A Studies of the HER-2/neu proto-oncogene in human breast and ovarian cancer Science 1989 244 707 712 2470152
Revillion F Bonneterre J Peyrat JP ERBB2 oncogene in human breast cancer and its clinical significance Eur J Cancer 1998 34 791 808 10.1016/S0959-8049(97)10157-5 9797688
Pauletti G Godolphin W Press MF Slamon DJ Detection and quantitation of HER-2/neu gene amplification in human breast cancer archival material using fluorescence in situ hybridization Oncogene 1996 13 63 72 8700555
Rajkumar T Gullick WJ The type I growth factor receptors in human breast cancer Breast Cancer Res Treat 1994 29 3 9 10.1007/BF00666177 7912566
De Potter CR Schelfhout AM The neu-protein and breast cancer Virchows Arch 1995 426 107 115 7757280
Ross JS Fletcher JA Linette GP Stec J Clark E Ayers M Symmans WF Pusztai L Bloom KJ The Her-2/neu gene and protein in breast cancer 2003: biomarker and target of therapy Oncologist 2003 8 307 325 10.1634/theoncologist.8-4-307 12897328
Hanna W Kahn HJ Trudeau M Evaluation of Her-2/neu (erbB-2) status in breast cancer: from bench to bedside Mod Pathol 1999 12 827 834 10463486
Stal O Sullivan S Wingren S Skoog L Rutqvist LE Carstensen JM Nordenskjold B c-erbB-2 expression and benefit from adjuvant chemotherapy and radiotherapy of breast cancer Eur J Cancer 1995 31A 2185 2190 10.1016/0959-8049(95)00344-4 8652240
Muss HB Thor AD Berry DA Kute T Liu ET Koerner F Cirrincione CT Budman DR Wood WC Barcos M c-erbB-2 expression and response to adjuvant therapy in women with node-positive early breast cancer N Engl J Med 1994 330 1260 1266 10.1056/NEJM199405053301802 7908410
Huston JS George AJ Engineered antibodies take center stage Hum Antibodies 2001 10 127 142 11847424
Hortobagyi GN Overview of treatment results with trastuzumab (Herceptin) in metastatic breast cancer Semin Oncol 2001 28 43 47 10.1053/sonc.2001.29710 11774205
McKeage K Perry CM Trastuzumab: a review of its use in the treatment of metastatic breast cancer overexpressing HER2 Drugs 2002 62 209 243 11790161
Shawver LK Slamon D Ullrich A Smart drugs: tyrosine kinase inhibitors in cancer therapy Cancer Cell 2002 1 117 123 10.1016/S1535-6108(02)00039-9 12086869
Ligibel JA Winer EP Trastuzumab/chemotherapy combinations in metastatic breast cancer Semin Oncol 2002 29 38 43 10.1053/sonc.2002.34054 12138396
Slamon DJ Leyland-Jones B Shak S Fuchs H Paton V Bajamonde A Fleming T Eiermann W Wolter J Pegram M Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2 N Engl J Med 2001 344 783 792 10.1056/NEJM200103153441101 11248153
Tavassoli FA Devilee P Eds World Health Organization Classification of Tumours. Pathology and Genetics of Tumours of the Breast and Female Genital Organs 2004 Lyon: IARC Press
Elston CW Ellis JO Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: Experience from a long study with long-term follow-up Histopathology 1991 19 403 410 1757079
Pauletti G Dandekar S Rong H Ramos L Peng H Seshadri R Slamon DJ Assessment of methods for tissue-based detection of the HER-2/neu alteration in human breast cancer: a direct comparison of fluorescence in situ hybridization and immunohistochemistry J Clin Oncol 2000 18 3651 3664 11054438
Varshney D Zhou YY Geller SA Alsabeh R Determination of HER-2 status and chromosome 17 polysomy in breast carcinomas comparing HercepTest and PathVysion FISH assay Am J Clin Pathol 2004 121 70 77 10.1309/FUQH-92B0-3902-5LHG 14750243
Choi DH Shin DB Lee MH Lee DW Dhandapani D Carter D King BL Haffty BG A comparison of five immunohistochemical biomarkers and HER-2/neu gene amplification by fluorescence in situ hybridization in white and Korean patients with early-onset breast carcinoma Cancer 2003 98 1587 1595 10.1002/cncr.11703 14534873
Yang J Xing T Yao X Hu R Relationship of C-erbB-2 oncogene overexpression to estrogen progesterone receptors in breast cancer and its prognostic significance Hua Xi Yi Ke Da Xue Xue Bao 1997 28 214 217 10683938
Abadjian G Antoun R Breast carcinoma: evaluation of hormone receptors and pS2, erb-B2, P-glycoprotein and Ki-67 markers J Med Liban 1996 44 10 15 8965315
el-A Helal T Khalifa A Kamel AS Immunohistochemical expression of p53 and c-erbB2 proteins in breast cancer in Egypt Anticancer Res 2000 20 2145 2150 10928168
Hoff ER Tubbs RR Myles JL Procop GW HER2/neu amplification in breast cancer: stratification by tumor type and grade Am J Clin Pathol 2002 117 916 921 10.1309/4NTU-N6K4-F8JF-EWRX 12047143
Arpino G Bardou VJ Clark GM Elledge RM Infiltrating lobular carcinoma of the breast: tumor characteristics and clinical outcome Breast Cancer Res 2004 6 R149 156 10.1186/bcr767 15084238
Rosenthal SI Depowski PL Sheehan CE Ross JS Comparison of HER-2/neu oncogene amplification detected by fluorescence in situ hybridization in lobular and ductal breast cancer Appl Immunohistochem Mol Morphol 2002 10 40 46 10.1097/00022744-200203000-00007 11893034
Porter PL Garcia R Moe R Corwin DJ Gown AM C-erbB-2 oncogene protein in in situ and invasive lobular breast neoplasia Cancer 1991 68 331 334 1676930
Taucher S Rudas M Mader RM Gnant M Dubsky P Bachleitner T Roka S Fitzal F Kandioler D Sporn E Do we need HER-2/neu testing for all patients with primary breast carcinoma? Cancer 2003 98 2547 2553 10.1002/cncr.11828 14669272
Eppenberger-Castori S Moore DH Jr Thor AD Edgerton SM Kueng W Eppenberger U Benz CC Age-associated biomarker profiles of human breast cancer Int J Biochem Cell Bio 2002 34 1318 1330 10.1016/S1357-2725(02)00052-3 12200028
Berger U Wilson P McClelland RA Davidson J Coombes RC Correlation of immunocytochemically demonstrated estrogen receptor distribution and histopathologic features in primary breast cancer Hum Pathol 1987 18 1263 1267 2824329
Paik S Bryant J Park C Fisher B Tan-Chiu E Hyams D Fisher ER Lippman ME Wickerham DL Wolmark N erbB-2 and response to doxorubicin in patients with axillary lymph node-positive, hormone receptor-negative breast cancer J Natl Cancer Inst 1998 90 1361 1370 10.1093/jnci/90.18.1361 9747867
Hartmann LC Ingle JN Wold LE Farr GH Jr Grill JP Su JQ Maihle NJ Krook JE Witzig TE Roche PC Prognostic value of c-erbB2 overexpression in axillary lymph node positive breast cancer. Results from a randomized adjuvant treatment protocol Cancer 1994 74 2956 2963 7954259
Stierer M Rosen H Weber R Hanak H Spona J Tuchler H Immunohistochemical and biochemical measurement of estrogen and progesterone receptors in primary breast cancer. Correlation of histopathology and prognostic factors Ann Surg 1993 218 13 21 8328824
Ruder AM Lubin F Wax Y Geier A Alfundary E Chetrit A Estrogen and progesterone receptors in breast cancer patients. Epidemiologic characteristics and survival differences Cancer 1989 64 196 202 2731116
Zafrani B Aubriot MH Mouret E De Cremoux P De Rycke Y Nicolas A High sensitivity and specificity of immunohistochemistry for the detection of hormone receptors in breast carcinoma: comparison with biochemical determination in a prospective study of 793 cases Histopathology 2000 37 536 545 10.1046/j.1365-2559.2000.01006.x 11122436
Ferno M Andersson C Fallenius G Idvall I Oestrogen receptor analysis of paraffin sections and cytosol samples of primary breast cancer in relation to outcome after adjuvant tamoxifen treatment. The South Sweden Breast Cancer Group Acta Oncol 1996 35 17 22 8619935
Layfield LJ Saria E Mooney EE Liu K Dodge RR Tissue heterogeneity of immunohistochemically detected estrogen receptor. Implications for image analysis quantification Am J Clin Pathol 1998 110 758 764 9844588
Esteban JM Ahn C Mehta P Battifora H Biologic significance of quantitative estrogen receptor immunohistochemical assay by image analysis in breast cancer Am J Clin Pathol 1994 102 158 162 8042582
Mohammed RH Lakatua DJ Haus E Yasmineh WJ Estrogen and progesterone receptors in human breast cancer. Correlation with histologic subtype and degree of differentiation Cancer 1986 58 1076 1081 3015374
Ashba J Traish AM Estrogen and progesterone receptor concentrations and prevalence of tumor hormonal phenotypes in older breast cancer patients Cancer Detect Prev 1999 23 238 244 10.1046/j.1525-1500.1999.99019.x 10337003
Anderson WF Chu KC Chatterjee N Brawley O Brinton LA Tumor variants by hormone receptor expression in white patients with node-negative breast cancer from the surveillance, epidemiology, and end results database J Clin Oncol 2001 19 18 27 11134191
Chariyalertsak S Chariyalertsak S Cheirsilpa A Sugano K Ohkura H Immunohistochemical detection of c-erbB-2 oncoprotein in patients with breast cancer J Med Assoc Thai 1996 79 715 721 8997010
May E Mouriesse H May-Levin F Qian JF May P Delarue JC Human breast cancer: identification of populations with a high risk of early relapse in relation to both oestrogen receptor status and c-erbB-2 overexpression Br J Cancer 1990 62 430 435 1976381
|
16168103
|
PMC1242123
|
CC BY
|
2021-04-27 23:14:50
|
no
|
Breast Cancer Res. 2005 May 24; 7(5):R598-R604
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1200
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12031616810810.1186/bcr1203Research ArticleEarly detection of breast cancer based on gene-expression patterns in peripheral blood cells Sharma Praveen [email protected] Narinder S [email protected] Robert [email protected] Per [email protected] Petter [email protected] Hege [email protected] Marianne [email protected] Lena [email protected] Cecilie [email protected] Pradeep [email protected] Alia [email protected] Jarle [email protected] Torill [email protected] Lars A [email protected] Ellen [email protected]ørresen-Dale Anne-Lise [email protected]önneborg Anders [email protected] DiaGenic ASA, Oslo, Norway2 Departments of Health, Research and Policy, and Statistics, Stanford University, Stanford, CA, USA3 Department of Radiology, Ullevål University Hospital, Oslo, Norway4 Department of Clinical Chemistry, Ullevål University Hospital, Oslo, Norway5 Department of Pathology, The Gade Institute, Haukeland University Hospital, Norway6 Department of Pathology, Ullevål University Hospital, Oslo, Norway7 Department of Surgery, Ullevål University Hospital, Oslo, Norway8 Department of Genetics, The Norwegian Radium Hospital; and University of Oslo, Faculty division, The Norwegian Radium Hospital, Oslo Norway2005 14 6 2005 7 5 R634 R644 11 4 2005 28 4 2005 Copyright © 2005 Sharma et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Existing methods to detect breast cancer in asymptomatic patients have limitations, and there is a need to develop more accurate and convenient methods. In this study, we investigated whether early detection of breast cancer is possible by analyzing gene-expression patterns in peripheral blood cells.
Methods
Using macroarrays and nearest-shrunken-centroid method, we analyzed the expression pattern of 1,368 genes in peripheral blood cells of 24 women with breast cancer and 32 women with no signs of this disease. The results were validated using a standard leave-one-out cross-validation approach.
Results
We identified a set of 37 genes that correctly predicted the diagnostic class in at least 82% of the samples. The majority of these genes had a decreased expression in samples from breast cancer patients, and predominantly encoded proteins implicated in ribosome production and translation control. In contrast, the expression of some defense-related genes was increased in samples from breast cancer patients.
Conclusion
The results show that a blood-based gene-expression test can be developed to detect breast cancer early in asymptomatic patients. Additional studies with a large sample size, from women both with and without the disease, are warranted to confirm or refute this finding.
See related letter by Li at
==== Body
Introduction
Early detection of breast cancer can improve the chances of successful treatment and recovery. To date, mammographic screening is the most reliable method to detect breast cancer in asymptomatic patients. Although highly effective, it has significant limitations, so that the development of more accurate, convenient, and objective detection methods is needed. In the absence of microcalcification, mammography often fails to detect tumors that are less than 5 mm in size, and also mammograms of women with dense breast tissue are difficult to interpret. For example, in a study of over 11,000 women with no clinical symptoms of breast cancer, the sensitivity of mammography was only 48% for the subset of women with extremely dense breasts, compared with 78% sensitivity for the entire sample of women in the study [1]. In addition, when an abnormality has been detected, further tests involving invasive steps must complement mammography to establish whether the detected abnormality is a cancer.
A vast amount of literature is already available describing the potential use of large-scale gene expression analysis in disease diagnosis, including breast cancer [2-8]. However, most published work with implications in cancer diagnosis has involved clinical samples comprising either diseased tissues or cells. Obtaining such samples for clinical purposes requires a prior knowledge of both their presence and their location in the body. A gene-expression-based test to detect cancers that does not rely upon the availability of tissues or cells from the diseased area has not yet been described.
It has recently been suggested that circulating leukocytes can be viewed as scouts, continuously maintaining a vigilant and comprehensive surveillance of the body for signs of infection or other threats, including cancer [9]. In line with this view, we show that peripheral blood can be used to develop a gene-expression-based test for early detection of breast cancer. The rationale for using blood cells as monitors for a malignant disease elsewhere in the body is based on the hypothesis that a malignant growth will cause characteristic changes in the biochemical environment of blood. These changes will affect the expression pattern of certain genes in blood cells.
In this pilot study, we have analyzed gene-expression patterns in peripheral blood cells of women diagnosed with breast cancer and women with no signs of this disease. We have identified a panel of genes with distinct expression patterns in cancer versus noncancer samples. The results indicate that breast cancer causes characteristic changes in the biochemical environment of blood already during early stages of disease development. Blood cells sense and respond to the change by decreasing the expression of genes involved in protein synthesis and increasing the expression of defense-related genes. We show that the expression pattern of the identified genes can be used to discriminate and predict the class of breast cancer and non-breast-cancer samples with high accuracy. Our findings should pave way for the development of a blood-based gene-expression test for early detection of breast cancer.
Materials and methods
Blood samples
Blood samples were collected from donors with their informed consent under an approval from Regional Ethical Committee of Norway (331-99-99138). All donors were treated anonymously during analysis. Blood was drawn from women with a suspect initial mammogram, prior to any knowledge of whether the abnormality observed during first screening was benign or malignant. In all cases, the blood samples were drawn between 8 a.m. and 4 p.m. From each woman, 10 ml blood was drawn by skilled personnel either in vacutainer tubes containing ethylenediaminetetraacetic acid (EDTA) as anticoagulant (Becton Dickinson, Baltimore, MD, USA) or directly in PAXgene™ tubes (PreAnalytiX, Hombrechtikon, Switzerland). Blood collected in EDTA-containing tubes was immediately stored at -80°C, while PAX tubes were left overnight at room temperature and then stored at -80°C until use.
Preparation of cDNA arrays
One thousand four hundred thirty-five cDNA clones were randomly picked from a plasmid library constructed from whole blood of 550 healthy individuals (Clontech, Palo Alto, CA, USA). Based on the sequence analysis of more than 500 cDNAs, redundancy among the randomly picked clones was estimated to be about 20%. For amplification of inserts, bacterial clones were grown in microtiter plates containing 150 μl Luria Broth media with 50 μg/ml carbenicillin, and incubated overnight with agitation at 37°C. To lyse the cells, 5 μl of each culture was diluted with 50 μl dH2O and incubated for 12 min at 95°C. Of this mixture, 2 μl were subjected to a PCR reaction using 40 μmol of 5' – and 3' – sequencing primers in the presence of 1.5 mM MgCl2. PCR reactions were performed with the following cycling protocol: 4 min at 95°C, followed by 25 cycles of 1 min at 94°C, 1 min at 60°C, and 3 min at 72°C either in a RoboCycler Temperature Cycler (Stratagene, La Jolla, CA, USA) or DNA Engine Dyad Peltier Thermal Cycler (MJ Research Inc, Waltham, MA, USA). The amplified products were denatured with NaOH (0.2 M, final concentration) for 30 min and spotted onto Hybond-N+ membranes (Amersham Pharmacia Biotech, Little Chalfont, UK), using a MicroGrid II workstation in accordance with the manufacturer's instructions (BioRobotics Ltd, Cambridge, UK). The immobilized cDNAs were fixed using a UV cross-linker (Hoefer Scientific Instruments, San Francisco, CA, USA).
The printed arrays also contained controls for assessing background level, consistency, and sensitivity of the assay. These were spotted at multiple positions in addition to the 1,435 cDNAs, and included controls such as PCR mix (without any insert); controls of the SpotReport™ 10-array validation system (Stratagene), and cDNAs corresponding to constitutively expressed genes such as β-actin, γ-actin, glyceraldehyde-3-phosphate dehydrogenase, human ornithine decarboxylase and cyclophilin.
RNA extraction, probe synthesis, and hybridization
Blood collected in EDTA tubes was thawed at 37°C and transferred to PAX tubes, and total RNA was purified in accordance with the supplier's instructions (PreAnalytiX). From blood collected directly in PAX tubes, total RNA was extracted in the tubes as above without any transfer to new tubes. Contaminating DNA was removed from the isolated RNA by DNAase I treatment using a DNA-free kit (Ambion Inc, Austin, TX, USA). RNA quality was determined visually by inspecting the integrity of 28S and 18S ribosomal bands after agarose-gel electrophoresis. Only samples from which good-quality RNA was extracted were used in this study. In our experience, blood collected in EDTA tubes often resulted in poor-quality RNA, whereas blood collected in PAX tubes almost always yielded good-quality RNA. The concentration and purity of extracted RNA were determined by measuring the absorbance at 260 nm and 280 nm. From the total RNA, mRNA was isolated using Dynabeads in accordance with the supplier's instructions (Dynal AS, Oslo, Norway).
Labeling and hybridization experiments were performed in 16 batches. The number of samples assayed in each batch varied from six to nine. To minimize the noise due to batch-to-batch variation in printing, only the arrays manufactured during the same print run were used in each batch. When samples were assayed more than once (replicates), aliquots from the same mRNA pool were used for probe synthesis. For probe synthesis, aliquots of mRNA corresponding to 4 to 5 μg of total RNA were mixed together with oligodT25NV (0.5 μg/μl) and mRNA spikes of the SpotReport™ 10-array validation system (10 pg; Spike 2, 1 pg), heated to 70°C, and then chilled on ice. The probes were synthesized by reverse transcription in 35 μl reaction mix in the presence of 50 μCi [α33P]dATP, 3.5 μM dATP, 0.6 mM each of dCTP, dTTP, dGTP, 200 units of SuperScript II reverse transcriptase (Invitrogen, Life Technologies, Carlsbad, CA, USA), and 0.1 M DTT labeling for 1.5 hours at 42°C. After synthesis, the enzyme was deactivated for 10 min at 70°C and mRNA removed by incubating the reaction mix for 20 min at 37°C in 4 units of Ribo H (Promega, Madison, WI, USA). Unincorporated nucleotides were removed using ProbeQuant G 50 columns (Amersham Biosciences, Piscataway, NJ, USA).
The membranes were equilibrated in 4 × standard saline citrate (SSC) (1 × SSC, 0.15 M NaCl, 0.015 M sodium citrate, pH 7.0) for 2 hours at 30°C and prehybridized overnight at 65°C in 10 ml prehybridization solution (4 × SSC, 0.1 M NaH2PO4, 1 mM EDTA, 8% dextran sulfate, 10 × Denhardt's solution, 1% SDS). Freshly prepared probes were added to 5 ml of the same prehybridization solution, and hybridization continued overnight at 65°C. The membranes were washed at 65°C with increasing stringency (2 × 30 min each in 2 × SSC, 0.1% SDS; 1 × SSC, 0.1% SDS; 0.1 × SSC, 0.1% SDS).
Quantification of hybridization signals
The hybridized membranes were exposed to Phosphoscreens (super resolution) and an image file generated using PhosphoImager (Cyclone, Packard, Meriden, CT, USA). The identification and quantification of the hybridization signals, as well as subtraction of local background values, were performed using Phoretix™ software (Nonlinear Dynamics, Newcastle upon Tyne, UK). For background subtraction, the median of the line of pixels around each spot outline was subtracted from the intensity of the signals assessed in each spot.
Data analysis
From the background-subtracted data for 1,435 genes, 1.25% of the lowest and 1.25% of the highest signals were trimmed from each membrane. Since the cDNAs with signals falling within this range varied between membranes, values of 67 cDNAs in total were removed from all membranes, and the expression data for only 1,368 genes were further analyzed. The data were normalized by dividing the value of each spot by the mean of signals in each array followed by a cube-root transformation. Supplementary Fig. 1 (left panel) (Additional file 1) shows a clear batch effect in the cube-root-normalized data (similar effects were also visible in the raw data). A simple one-way analysis of variance (ANOVA) was performed to adjust for the batch effects. Supplementary Fig. 1 (right panel) (Additional file 1) shows that the systematic batch effects were removed by the ANOVA adjustment. The batch-adjusted data were then analyzed using the nearest-shrunken-centroid method [10].
In this method, standard 'external' cross-validation is used to determine the optimal shrinkage threshold. This optimal threshold is then used with the full training set to construct the centroid. As a result, for each value of the threshold, the estimate of cross-validation error obtained is approximately unbiased for the true test-error rate.
The leave-one-out cross-validation approach was used in this work. The data were divided into M nonoverlapping subsets (M = number of unique blood samples present). The model was then trained M-1 times on these subsets combined, each time leaving out one of the subsets (unique blood sample) from the training data, but using only the omitted subset to compute the prediction error. The errors obtained on all parts were added together and used to compute the overall misclassification error. It is well known that leave-one-out cross-validation provides an approximately unbiased and reliable estimate of the misclassification rate that would be obtained from an independent sample of patients [11,12]. In the terminology of Ambroise and McLachlan [12], we used external cross-validation (as they recommend).
The raw and the batch-adjusted data for 1,368 genes in an Excel file is provided in Supplementary Table 1 (Additional file 2) and Supplementary Table 2 (Additional file 3).
Results
We analyzed gene-expression patterns in 60 blood samples obtained from 56 different women (Table 1). The experiments were performed in 16 batches. To investigate the reproducibility of results, 13 samples from women with breast cancer and 23 samples from women with no breast cancer were analyzed in different batches using aliquots from the same mRNA pool, giving a total of 102 experimental samples.
The generated expression data was preprocessed and then analyzed by the nearest-shrunken-centroid method [10]. A standard leave-one-out cross-validation approach was used to determine the optimal amount of shrinkage threshold. Since we had 60 unique blood samples and for some of them experiments were replicated more than once, for cross-validation the data were divided into 60 nonoverlapping subsets, where each subset represented a unique blood sample and included all the replicates present in the data set. A sample was judged as correctly classified only when a majority of members in the corresponding cross-validation segment were correctly classified. The minimum overall misclassification error was observed at a threshold value of 2.28, yielding a subset of 37 genes (Fig. 1). At this threshold, 10 of the 57 samples were misclassified and 3 samples were judged nondecisions, because there was no majority for either the breast-cancer or non-breast-cancer class (Table 2). A detailed prediction result is presented in Table 1.
The prediction was highly accurate for samples from women with early stages of breast cancer, stage 0 and stage I. Among the 14 samples representing early stages, there was one nondecision and 11 of 13 samples were correctly predicted. Five of seven stage II and one of two stage III samples were correctly predicted.
Most of the cancer samples (22 of 24) analyzed in this study were obtained from women who had cancer of ductal origin. One woman, the origin of whose cancer was not known, had a previous history of breast cancer and at the time of blood collection the cancer had spread to supraclavicular and infraclavicular nodes. Another sample that did not belong to the ductal group was obtained from a woman who had invasive lobular carcinoma in one breast and a tubular adenocarcinoma in the other. Unlike ductal carcinoma, which originates from cells lining ducts, lobular carcinoma originates from cells lining lobules. Both samples were incorrectly predicted. It is possible that cancer of other than ductal origin affects the expression pattern of the selected 37 genes in blood cells differently than ductal carcinomas.
Seventeen of 19 samples obtained from women with a suspect first mammogram were correctly predicted (Table 1, subgroup A2), indicating the expression profile of the selected 37 genes to be highly efficient in discriminating between cancerous and noncancerous breast abnormalities. In two samples, we were not able to make any diagnostic decision.
Among the 17 samples from women with no reported breast abnormality, 13 were correctly predicted (Table 1, subgroup A3). These included samples from breast-feeding women as well as those drawn at different times in the menstrual cycle from one woman. However, the three samples from pregnant women and a sample from a woman with acute bacterial infection at the time of blood collection were all incorrectly predicted. The woman with acute bacterial infection was, in addition, chronically infected with Epstein–Barr virus. It is known that both pregnancy and chronic infection may elicit responses that can mimic breast cancer. During late pregnancy, similar to breast cancer, cells of mammary epithelial buds divide to form ducts infiltrating breast stroma and build a local blood supply. Also, both breast cancer and chronic infections are known to induce inflammatory responses in the body.
We also calculated the misclassification error, taking an average of the class probability for each sample in all 60 cross-validation segments as compared with our previous approach in which a sample was judged as correctly classified only when a majority of members in the corresponding cross-validation segment were correctly classified. Thus, each segment represented an average class probability for each sample, and we predicted each sample to the class with the highest average probability. The main purpose of adopting this approach was to be able to make a unanimous decision with respect to class membership. The minimum error rate using the average-class approach was obtained at a threshold value of 2.42 and involved a subset of only 25 genes, giving a further reduction of 12 genes (Supplementary Fig. 2) (Additional file 4). Also, 10 (7 breast cancer and 3 non-breast-cancer samples) of the 60 samples were misclassified, which is a slightly better result than that obtained with 37 genes, where there were 3 nondecisions (Supplementary Fig. 3) (Additional file 5).
Table 3 shows the shrunken t-statistic scores of the selected 37 predictive genes for comparing breast-cancer class to non-breast-cancer class, the genes in the public databases to which they show sequence similarity, and their putative biological function. The relative expression of 12 predictive genes with highest scores is presented in Fig. 2. The majority of the predictive genes (29 of 37) had a decreased expression (positive score) in the samples from breast cancer patients. The identity of predictive genes was determined by partially sequencing the corresponding spotted cDNA clones and searching for gene similarities in public databases.
Sequence analysis revealed that 8 of 35 predictive genes contained redundant information. Since the arrayed cDNAs were derived from randomly picked clones from a library constructed from whole blood from 550 healthy individuals, we had expected a redundancy of about 20% among the selected genes. Of the 35 genes, 18 (51%) encoded ribosomal proteins. In comparison, the frequency of cDNAs representing ribosomal proteins was estimated to be only about 8% among the arrayed cDNAs. All genes encoding ribosomal proteins had reduced expression in samples from breast cancer patients, indicating a decrease in ribosome production in the blood cells of these patients. Also, genes encoding a translation elongation factor, eEF1 and RACK1 (receptor for activated C kinase), were expressed at a lower level in samples from cancer patients, indicating reduced protein translation activity in these samples. RACK1 plays a key role in the joining of 60S and 40S subunits into a functionally active 80S ribosome complex [13].
Among the eight predictive genes with increased expression in samples from breast cancer patients, two encoded histone replacement protein H3.3, which is thought to be involved in chromatin remodelling [14], and six encoded proteins that may play a role in defense-related functions. Four genes with increased expression encoded ferritin and calgranulin B. Ferritin is involved in intracellular storage and sequestration of iron. Increased expression of ferritin has been shown to reduce the accumulation of reactive oxygen species in response to oxidant challenge in HeLa cells [15]. Calgranulin B is expressed by blood cells both during infection and during inflammation and may play a role in host defense [16]. Interferon-induced transmembrane protein 2 has been implicated in the immune response, while human granule proteoglycan peptide core is assumed to form stable complexes with proteases and other granule-localized proteins to prevent their intragranular autolysis and facilitate their concerted action extracellularly [17]. Interestingly, most predictive genes identified in this study belonged to the family of genes that exhibited altered expression in neutrophils after stimulation by nonvirulent and virulent bacterial stimuli [18,19].
Discussion
This is a first report demonstrating that breast cancer affects gene-expression patterns in peripheral blood cells during early stages of disease development. The results presented represent an initial phase in the development of a blood-based gene-expression test for breast cancer detection. A larger number of samples, from both women with and women without the disease, should be further analyzed before the clinical efficacy of our finding can be evaluated. However, the results clearly show that by analyzing the expression pattern of selected genes in blood cells, a diagnostic test for breast cancer detection can be efficiently developed.
In the present study, we examined gene-expression patterns in peripheral blood cells as a whole, rather than specific cellular subsets. It has recently been shown that individual variations in gene-expression pattern in peripheral blood could be traced to altered relative proportions of the specific blood cell subsets [9]. If there were systematic differences in the relative proportions of peripheral blood cell types in women with breast cancer and those without this disease, such differences might explain the observed gene-expression patterns. Interestingly, Whitney and colleagues [9] found that transcripts involved in protein synthesis were over-represented in lymphocytes and monocytes as compared with granulocytes. The reduced expression of transcripts involved in protein synthesis and the increased expression of transcripts involved in defense responses in breast cancer patients may reflect a systematic shift in favor of granulocytes as compared with lymphoid cells in the peripheral blood of breast cancer patients. However, to our knowledge, no such systematic shift during breast cancer development has been reported, and the subject requires further investigation. Alternatively, changes in the expression pattern of genes involved in protein synthesis, chromatin remodelling, and defense-related genes in the blood samples of breast cancer patients may indicate systematic activation of certain blood cell subsets such as neutrophils in these patients.
Our ability to correctly assign the class of samples from women with Crohn's disease, rheumatic disease, or diabetes as non-breast-cancer suggests that breast cancer affects the expression pattern of identified predictive genes differently from some of the diseases associated with anemia and chronic inflammation. The correct prediction of two samples from a woman with ductal carcinoma in situ further suggested that malignant lesions, though confined within the breast duct, may induce similar changes in the expression pattern of these genes to the changes seen during the more advanced stages of breast cancer (stages I to III). However, incorrect prediction of a sample obtained from a woman with invasive lobular carcinoma and tubular adenocarcinoma and from a woman where the cancer had spread to supraclavicular and infraclavicular nodes indicates that malignancy in itself is not a prerequisite condition for the observed changes in the expression pattern of the identified predictive genes.
The efficient prediction of samples derived from patients whose cancer had not yet spread to lymph nodes shows that a blood-based gene-expression test can be developed for breast cancer detection in asymptomatic patients. As compared with existing methods, an accurate method for breast cancer detection based on peripheral blood as a clinical sample will be highly desirable because of the easy accessibility and the less invasive procedure for obtaining samples. The test could be integrated as an adjunct to already established methods and be used to improve their efficacy. For example, a blood-based gene-expression test could assist mammography in discriminating between benign and malignant breast abnormalities. It could become a part of routine screening programs, especially when the patient has an increased risk for breast cancer.
It is important that any test intended for use in breast cancer diagnosis has a low rate of both false positives and false negatives. Based on the expression pattern of identified 37 genes, the prediction achieved corresponded to a false positive rate of 0.12 and false negative rate of 0.26. Since, the main goal of this work was to see whether the information about breast cancer is present in peripheral blood samples in the form of changed gene-expression patterns, we analyzed only a limited number of gene candidates in this study. The genes analyzed corresponded to clones that were randomly picked from a plasmid library constructed from whole blood of 550 individuals. The motivation for this approach for selecting gene candidates was based on the assumption that if the expression pattern of certain genes in blood cells is affected during early stages of breast cancer, the genes affected would most likely include ones involved in cell maintenance and general metabolism. Since such genes are expressed at high level in a cell, they would be frequently represented in a cDNA library and selected preferentially when randomly picked. It is our view that expression techniques such as microarrays, where the expression of thousands of genes can be monitored simultaneously, can further be used to screen for better predictive genes and develop more accurate diagnostic models.
We envisage blood-based gene-expression tests to have the potential of becoming a versatile and powerful tool for detection of disease, including other forms of cancers. As with breast cancer, other diseases may also cause characteristic changes in the biochemical environment of blood and affect the gene-expression patterns in blood cells. Specific gene-expression-based models can then be developed and used for diagnostic purposes.
Conclusion
The results presented show that breast cancer even during early stages of disease development affects the expression pattern of certain genes in peripheral blood cells. By identifying these genes and analyzing their expression pattern, it is possible to develop a blood-based gene-expression test for early detection of breast cancer. Additional studies with a large sample size, both from women with and without the disease, are warranted to confirm or refute this finding.
Abbreviations
ANOVA = analysis of variance; EDTA = ethylenediaminetetraacetic acid; eEF = eukaryotic elongation factor; RACK1 = receptor for activated C kinase 1; SSC = standard saline citrate (1 × SSC, 0.15 M NaCl, 0.015 M sodium citrate, pH 7.0).
Competing interests
PvS, NSS, HB, MJ, LK, CM, PdS, AZ, and AL are employees of DiaGenic. None of the other authors have any competing interests.
Authors' contributions
PvS and AL conceived the experiments. PvS, AL, and NSS designed the experiments. HB, MJ, CM, AZ, LK, PdS, PvS, and AL performed the experiments. PSk, PU, ES, TS, JA, and LAA provided the samples and their clinical details. RT and NSS performed the statistical analysis. PvS wrote the paper. RT, ALBD, AL, NSS, and PSk provided helpful comments during preparation of the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Supplementary Figure 1, a pdf showing batch adjustment. (Left) Normalized data before batch adjustment; (right) normalized data after batch adjustment by ANOVA.
Click here for file
Additional File 2
Supplementary Table 1, an Excel file showing the raw data for 1,368 genes. C, breast-cancer class; N, non-breast-cancer class.
Click here for file
Additional File 3
Supplementary Table 2, an Excel file showing the batch-corrected data for 1,368 genes. C, breast-cancer class; N, non-breast-cancer class.
Click here for file
Additional File 4
Supplementary Figure 2, pdf showing misclassification rate as a function of threshold value and the number of genes involved when the error is calculated by taking an average of the class probability for each sample in all 60 cross-validation segments. The upper graph shows that the minimum overall misclassification error is observed at a threshold value of 2.42. The lower graph shows the profile for the misclassification error for breast-cancer (C) and non-breast-cancer (N) samples as a function of threshold value and the number of genes involved.
Click here for file
Additional File 5
Supplementary Figure 3, a pdf showing estimated cross-validated probabilities of 60 different blood samples. Red circles represent breast-cancer class (C) and green circles represent non-breast-cancer class (N). Each sample has two probabilities, one for the breast-cancer class and the other for the non-breast-cancer class. The sample is classified in the class whose probability is >0.5.
Click here for file
Acknowledgements
The experimental work was supported by DiaGenic ASA. ALBD was supported by a grant under the Functional Genomics (FUGE) programme (159188/S10) from the Research Council of Norway.
Figures and Tables
Figure 1 Misclassification rate as a function of threshold value and the number of genes involved. The error was calculated using the majority rule. A nondecision was counted as an error. The upper graph shows that the minimum overall misclassification error was observed at a threshold value of 2.28. The lower graph shows the profile for misclassification error for breast-cancer (C) and non-breast-cancer (N) samples as a function of threshold value and the number of genes involved.
Figure 2 Relative expression of 13 predictive genes with the highest scores in breast-cancer and non-breast-cancer samples. Red circles represent samples from women with breast cancer and green circles represent samples from women with no signs of breast cancer. The number on the upper axis represents the position ID of predictive genes in the array (Table 3).
Table 1 Gene-expression patterns in 60 blood samples obtained from 56 different women
Subgroup A1: Women with breast cancer
Sample ID Age (y) Stage Histology Grade Size (mm) Nodes Comments/ other disease if present Times assayed Prediction (37 genes)
3 54 I IDC 1 11 0 # 2 +
5 67 0 DCIS 2 20 0 # 3 +
7 51 II IDC 3 20 1/7 # 2 +
8 84 II IDC 1 22 2/2 # 2 +
15 66 I IDC 2 15 0 Rheumatic disease 3 +
16 68 I IDC 1 7 0 # 1 +
17 66 II IDC 1 26 0 Epilepsy 1 -
27 48 I IDC 2 4 0 # 2 ND
31 47 I IDC 2 15 0 # 2 -
35 44 II IDC 2 25 0 # 1 +
36 50 I Multifocal IDC 1 5 × 14 0 # 1 -
38 n.a. 0 DCIS 2 9 0 # 1 +
39 65 I IDC 1 15 0 # 1 +
40 n.a. I IDC 2 14 0 Psoriasis 1 +
42 71 I IDC 1 8 0 # 1 +
44 55 III IDC 1 35 0 # 1 +
45 63 II IDC 3 23 0 # 1 -
48 65 IV - - - Metastases in supra- and infra- clavicular nodes Breast cancer, 1982 1 -
49 65 I IDC 1 11 0 Type 2 diabetes 3 +
50 69 III ILC 2 50 2/19 # 2 -
51 50 II IDC 2 24 0 # 2 +
53 60 II IDC 2 23 0 # 2 +
59 63 I IDC 1 10 0 # 2 +
60 52 I IDC 1 3 0 # 2 +
Subgroup A2: Women with abnormal first mammography
Sample ID Age (y) Breast abnormality Comments / other disease if present Times assayed Prediction (37 genes)
1 44 Benign density # 2 +
2 53 Benign microcalcifications Encapsulated cyst in left knee 2 +
4 45 Benign density # 2 +
11 46 Benign density Ulcerative colitis since 1983 2 +
12 44 Benign density # 2 +
13 50 Benign density Type 1 diabetes 2 +
14 47 Benign microcalcifications # 2 +
19 46 Benign density, cyst Crohn's disease 2 +
20 n.a. Benign density Rheumatic disease 1 +
28 44 Benign microcalcifications # 2 +
29 63 Benign density, cyst Fibromyalgia 2 ND
30 46 Benign density # 2 +
32 59 Benign tumor, fibroadenoma # 2 +
34 45 Benign density Type 2 diabetes 2 +
41 50 Fibrosis, benign Size histology 60 mm 1 +
43 51 Radial scar Size histology 10 mm 1 +
52 47 Benign density # 2 ND
54 52 Benign microcalcifications Cancer, large intestine, 1992 1 +
58 46 Benign density # 2 +
Subgroup A3: Women with no reported breast abnormality
Sample ID Age (y) Comments Times assayed Prediction (37 genes)
6 42 # 3 +
9 30 Breast feeding 2 +
10 34 Breast feeding 3 +
21 26 # 1 +
22 - # 1 +
18* 18 Week 1 2 +
23* Week 2 1 +
24* Week 3 1 +
26* Week 4 2 +
25* Week 5 1 +
33 34 Pregnant, 8 months 3 -
37 51 Acute bacterial infection in addition to chronic Epstein–Barr virus infection 1 -
46 27 Pregnant, 6 months 1 -
47 29 Pregnant, 9 months 1 -
55 43 # 1 +
56 43 # 2 +
57 22 # 2 +
Sample detail. Stage 0, in situ carcinoma; Stage I, invasive carcinoma with tumor size <20 mm; Stage II, invasive carcinoma with tumor size >20-50 mm; Stage III, invasive carcinoma with tumor size >50 mm. Stage IV, cancer spread to distant parts. *, Blood samples taken on five consecutive weeks from the same woman; -, incorrectly predicted; #, no relevant information available; +, correctly predicted; DCIS, ductal carcinoma in situ; IDC, invasive ductal carcinoma; ILC, invasive lobular carcinoma; n.a., not available; ND, nondecision.
Table 2 Confusion matrix of prediction results using 37 genesa
True/Predicted C N Error rateb
C 17 6 0.26
N 4 30 0.12
aWhen there was no majority for either the breast-cancer or non-breast-cancer class, the prediction was regarded as a nondecision. bTotal error rate = 0.18; 3 nondecisions. C, breast-cancer samples; N, non-breast-cancer samples.
Table 3 Details of the identified 37 predictive genes
Accession no. Gene similarity Putative cellular function Position ID Scorea
BC000514 Ribosomal protein L13a Ribosome production 19AM 0.8377
BC007512 Ribosomal protein L18a Ribosome production 31AJ 0.7321
BC019093 Guanine nucleotide binding protein, beta polypeptide 2-like; RACKs (receptors for activated C kinase) Protein translation 12AM 0.6972
BC009696 Interferon induced transmembrane protein 2 Cell – environment interaction, Immune response 12Q -0.6962
BC047681 S100 calcium binding protein A9 (Calgranulin B) Defence; inhibition of casein kinase II 31J -0.6444
BC066901 H3 histone, family 3B (H3.3B) Chromatin remodelling 5AK -0.6394
BC034149 Ribosomal protein S3 Ribosome production 23V 0.639
AK026634 Highly similar to HUMTI227HC, mRNA for TI-227H - 21AH 0.627
BC047681 S100 calcium binding protein A9 (Calgranulin B) Defence; inhibition of casein kinase II 24AQ -0.627
BC001126 Ribosomal protein S14 Ribosome production 28T 0.6231
NM_000980 Ribosomal protein L18a Ribosome production 31AF 0.6215
AY495316 Cytochrome c oxidase subunit, COX 1 Mitochondrial electron transport chain 15AK 0.6112
NM_001016 Ribosomal protein S12 Ribosome production 22S 0.6102
- - - 20AG 0.5839
BC016378 Ribosomal protein S11 Ribosome production 8S 0.5827
AY495316 Cytochrome c oxidase subunit, COX 1 Mitochondrial electron transport chain 27AG 0.5729
AF077043 Ribosomal protein L36 Ribosome production 3AR 0.5699
AF346981 Mitochondrial 16S rRNA Ribosome production 25P 0.5507
BC013857 H3 histone, family 3A Chromatin remodelling 3T -0.5496
M22146 Ribosomal protein S4 Ribosome production 31U 0.5176
BC016857 Ferritin, heavy polypeptide 1 Iron storage; defence against ROS 6N -0.5134
BC053370 Ribosomal protein SA Ribosome production 2G 0.5113
BC010165 Ribosomal protein S2 Ribosome production 2V 0.5071
BC009689 Cyclin D-type binding protein E2F-mediated transcription 21O 0.4978
BC018641 Eukaryotic translation elongation factor 1α (eEF1A) Protein translation 4AA 0.4974
D87735 Ribosomal protein L14 Ribosome production 19H 0.486
- - - 6AQ 0.4837
BC016857 Ferritin, heavy polypeptide 1 Iron storage; defence against ROS 3AB -0.481
BC012146 Ribosomal protein L3 Ribosome production 32AM 0.4776
BC001126 Ribosomal protein S14 Ribosome production 25R 0.4759
BC006784 Ribosomal protein S14 Ribosome production 24AJ 0.4695
J03223 Human secretory granule proteoglycan peptide core Defence (may neutralize hydrolytic enzymes) 11H -0.4681
AY147037 Myeloid/lymphoid or mixed-lineage leukemia 5 cDNA Chromatin remodeling and cellular growth suppression 30AP 0.4669
CD246392, EST Agencourt_14095501 NIH_MGC_172 cDNA - 8AK 0.4666
AY339570 Cytochrome c oxidase subunit, COX 1 Mitochondrial electron transport chain 2E 0.4662
U43701 Human ribosomal protein L23a Ribosome production 8G 0.4629
AY495252 Mitochondrial 16S rRNA Ribosome production 8AF 0.4625
The position of genes in the array is shown as well as their scores, the accession number of sequences in public databases that match them, and their known or putative cellular function. aThe score is a shrunken t-statistic for comparing breast-cancer class to non-breast-cancer class. A positive score means that expression was greater in the noncancer sample than the cancer sample; a negative score means that expression was greater in the cancer sample than the noncancer sample. -, no information available; ROS, reactive oxygen species.
==== Refs
Kolb TM Lichy J Newhouse JH Comparison of the performance of screening mammography, physical examination, and breast US and evaluation of factors that influence them: an analysis of 27,825 patient evaluations Radiology 2002 225 165 175 12355001
Bertucci F Nasser V Granjeaud S Eisinger F Adelaide J Tagett R Loriod B Giaconia A Benziane A Devilard E Gene expression profiles of poor-prognosis primary breast cancer correlate with survival Hum Mol Genet 2002 11 863 872 11971868 10.1093/hmg/11.8.863
Ellis M Davis N Coop A Liu M Schumaker L Lee RY Srikanchana R Russell CG Singh B Miller WR Development and validation of a method for using breast core needle biopsies for gene expression microarray analyses Clin Cancer Res 2002 8 1155 1166 12006532
Perou CM Sorlie T Eisen MB van de Rijn M Jeffrey SS Rees CA Pollack JR Ross DT Johnsen H Akslen LA Molecular portraits of human breast tumours Nature 2000 406 747 752 10963602 10.1038/35021093
Sørlie T Perou CM Tibshirani R Aas T Geisler S Johnsen H Hastie T Eisen MB van de Rijn M Jeffrey SS Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications Proc Natl Acad Sci USA 2001 98 10869 10874 11553815 10.1073/pnas.191367098
Sørlie T Tibshirani R Parker J Hastie T Marron JS Nobel A Deng S Johnsen H Pesich R Geisler S Repeated observation of breast tumor subtypes in independent gene expression data sets Proc Natl Acad Sci USA 2003 100 8418 8423 12829800 10.1073/pnas.0932692100
van 't Veer LJ Dai H van de Vijver MJ He YD Hart AA Mao M Peterse HL van der Kooy K Marton MJ Witteveen AT Gene expression profiling predicts clinical outcome of breast cancer Nature 2002 415 530 536 11823860 10.1038/415530a
West M Blanchette C Dressman H Huang E Ishida S Spang R Zuzan H Olson JA JrMarks JR Nevins JR Predicting the clinical status of human breast cancer by using gene expression profiles Proc Natl Acad Sci USA 2001 98 11462 11467 11562467 10.1073/pnas.201162998
Whitney AR Diehn M Popper SJ Alizadeh AA Boldrick JC Relman DA Brown PO Individuality and variation in gene expression patterns in human blood Proc Natl Acad Sci USA 2003 100 1896 1901 12578971 10.1073/pnas.252784499
Tibshirani R Hastie T Narasimhan B Chu G Diagnosis of multiple cancer types by shrunken centroids of gene expression Proc Natl Acad Sci USA 2002 99 6567 6572 12011421 10.1073/pnas.082099299
Hastie T Tibshirani R Friedman J The Elements of Statistical Learning 2001 New York: Springer
Ambroise C McLachlan GJ Selection bias in gene extraction on the basis of microarray gene-expression data Proc Natl Acad Sci USA 2002 99 6562 6566 11983868 10.1073/pnas.102102699
Ceci M Gaviraghi C Gorrini C Sala LA Offenhauser N Marchisio PC Biffo S Release of elF6 (p27-BBP) from the 60S subunit allows 80S ribosome assembly Nature 2003 426 579 584 14654845 10.1038/nature02160
Ahmad K Henikoff S Histone H3 variants specify modes of chromatin assembly Proc Natl Acad Sci USA 2002 99 16477 16484 12177448 10.1073/pnas.172403699
Orino K Lehman L Tsuji Y Ayaki H Torti SV Torti FM Ferritin and the response to oxidative stress Biochem J 2001 357 241 247 11415455 10.1042/0264-6021:3570241
Nisapakultorn K Ross KF Herzberg MC Calprotectin expression inhibits bacterial binding to mucosal epithelial cells Infect Immun 2001 69 3692 3696 11349032 10.1128/IAI.69.6.3692-3696.2001
Nicodemus CF Avraham S Austen KF Purdy S Jablonski J Stevens RL Characterization of the human gene that encodes the peptide core of secretory granule proteoglycans in promyelocytic leukemia HL-60 cells and analysis of translated product J Biol Chem 1990 265 5889 5896 2180935
Zhang X Kluger Y Nakayama Y Poddar R Whitney C DeTora A Weismann SM Newburger PE Gene expression in mature neutrophils: early responses to inflammatory stimuli J Leukoc Biol 2004 75 358 372 14634056 10.1189/jlb.0903412
Subrahmanyam YV Yamga S Prashar Y Lee HH Hoe NT Kluger Y Gerstein M Goguen JD Newburger PE Weismann SM RNA expression patterns change dramatically in human neutrophils exposed to bacteria Blood 2001 97 2457 2468 11290611 10.1182/blood.V97.8.2457
|
16168108
|
PMC1242124
|
CC BY
|
2021-01-04 16:04:33
|
no
|
Breast Cancer Res. 2005 Jun 14; 7(5):R634-R644
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1203
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12591616810210.1186/bcr1259Research ArticleDifferential responses to doxorubicin-induced phosphorylation and activation of Akt in human breast cancer cells Li Xinqun [email protected] Yang [email protected] Ke [email protected] Bolin [email protected] Zhen [email protected] Department of Experimental Therapeutics, The University of Texas MD Anderson Cancer Center, Houston, TX, USA2005 24 5 2005 7 5 R589 R597 24 1 2005 11 2 2005 18 4 2005 29 4 2005 Copyright © 2005 Li et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
We have shown previously that overexpression of constitutively active Akt or activation of Akt caused by constitutively active Ras or human epidermal growth factor receptor-2 (HER2) confers on breast cancer cells resistance to chemotherapy or radiotherapy. As an expanded study we here report differential responses in terms of phosphorylation and activation of Akt as a result of treatment with doxorubicin in a panel of breast cancer cell lines.
Methods
The levels of Akt phosphorylation and activity were measured by Western blot analysis with an anti-Ser473-phosphorylated Akt antibody and by in vitro Akt kinase assay using glycogen synthase kinase-3 as a substrate.
Results
Within 24 hours after exposure to doxorubicin, MCF7, MDA468 and T47D cells showed a drug-dose-dependent increase in the levels of phosphorylated Akt; in contrast, SKBR3 and MDA231 cells showed a decrease in the levels of phosphorylated Akt, and minimal or no changes were detected in MDA361, MDA157 and BT474 cells. The doxorubicin-induced Akt phosphorylation was correlated with increased kinase activity and was dependent on phosphoinositide 3-kinase (PI3-K). An increased baseline level of Akt was also found in MCF7 cells treated with ionizing radiation. The cellular responses to doxorubicin-induced Akt phosphorylation were potentiated after the expression of Akt upstream activators including HER2, HER3 and focal adhesion kinase.
Conclusion
Taken together with our recent published results showing that constitutive Akt mediates resistance to chemotherapy or radiotherapy, our present data suggest that the doxorubicin-induced phosphorylation and activation of Akt might reflect a cellular defensive mechanism of cancer cells to overcome doxorubicin-induced cytotoxic effects, which further supports the current efforts of targeting PI3-K/Akt for enhancing the therapeutic responses of breast cancer cells to chemotherapy and radiotherapy.
==== Body
Introduction
Cancer cells with an inherent or acquired capability to resist induction of apoptosis at some point(s) in the signal cascade pathway leading to cell death generally tend to be resistant to chemotherapy or radiotherapy. The serine–threonine protein kinase Akt has received much interest in recent years because it suppresses apoptosis induced by chemotherapy or radiotherapy through interaction with several critical molecules that regulate or execute apoptosis. For instance, after activation, Akt could do the following: it phosphorylates the proapoptotic protein Bcl-2 partner, Bad, which binds to and blocks the activity of Bcl-x, a factor in cell survival [1]; it inactivates caspase-9, which initiates the caspase cascade leading to apoptosis [2]; it represses the forkhead transcription factor FKHRL-1, which regulates the expression of the apoptosis-inducing Fas ligand [3]; and it phosphorylates IκB, thereby promoting the degradation of IκB and increasing the activity of the nuclear factor κB (NFκB) [3,4].
The kinase activity of Akt is triggered after the interaction of its pleckstrin homology domain with the lipid second messenger phosphatidylinositol 3,4,5-trisphosphate, which is generated by phosphoinositide 3-kinase (PI3-K). This interaction recruits Akt from the cytoplasm to the inner cytoplasmic membrane, where Akt undergoes conformational changes and is phosphorylated by the phosphatidylinositol-dependent kinases. The activated Akt is then relocated to the cytoplasm and may be transported further to the nucleus, phosphorylating a wide spectrum of substrates including the molecules mentioned above that are involved in the regulation of cell survival. PI3-K itself is activated by multiple mechanisms, including the activation of growth factor receptor tyrosine kinases [5,6] and G protein-coupled receptors [7,8], integrin-mediated cell adhesion [7,8], and the actions of oncogene products such as Ras [9,10] and hormones such as estrogen [11-13]. By controlling the levels of lipid second messengers, PI3-K regulates various cellular processes, including growth, differentiation, survival, migration and metabolism [14,15].
We have recently shown that expression of a constitutively active Akt, or an increased activity of the human epidermal growth factor receptor-2 (HER2)/PI3-K/Akt or Ras/PI3-K/Akt pathway, leads to multidrug or radiation resistance in human breast cancer cells [16-18]. In those studies we assessed the sensitivity to chemotherapy (including doxorubicin) or radiotherapy of breast cancer cells that contain a higher level of Akt activity due to the overexpression of HER2, constitutively active Ras or constitutively active Akt. To expand our previous studies, we report here a differential pattern of responses of breast cancer cell lines in terms of Akt phosphorylation and activity as a result of treatment with doxorubicin. Depending on the cell types, treatment of breast cancer cells with doxorubicin may trigger a transient phosphorylation and activation of Akt. This therapeutic intervention-triggered activation of Akt depends on an inherent activity of PI3-K, and the capability of the response is potentiated after the expression of Akt upstream regulators including HER2, HER3 or the focal adhesion kinase (FAK), but not by all the signals that are known to affect Akt activity, an example of which is the estrogen-mediated signal. Deprivation of the effect of estrogen did not alter the responsiveness of MCF7 cells to doxorubicin-induced Akt phosphorylation. Our data suggest that the therapeutic intervention-triggered activation of Akt might contribute to the resistance of breast cancer cells to doxorubicin. These results provide further experimental evidence that justifies targeting the PI3-K/Akt pathway to enhance the efficacy of breast cancer chemotherapy or radiotherapy.
Materials and methods
Cell lines and cell cultures
Eight breast cancer cell lines used in this study (MCF7, MDA468, SKBR3, MDA157, MDA231, MDA361, BT474 and T47D) were originally purchased from American Type Culture Collection (Manassas, VA, USA). The cells were grown and routinely maintained in Dulbecco's modified Eagle's medium/F12 medium supplemented with 10% fetal bovine serum (FBS), 2 mM glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. MCF7HER2 cells were described previously [19]. All cells were grown in a 37°C incubator supplied with 5% CO2 and 95% air.
Western blot antibodies and other reagents
Antibodies directed against Akt, Ser473-phosphorylated Akt1 (p-Akt), Ser21/9-phosphorylated glycogen synthase kinase-3 (GSK3), Ser136-phosphorylated Bad and anti-HER2 monoclonal antibody were obtained from Cell Signaling Technology (Beverly, MA, USA). Anti-HER3 antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti-His tag monoclonal antibody was ordered from Upstate Biotechnology (Charlottesville, VA, USA), as was the anti-FAK antibody that recognizes both FAK and FAK-related non-kinase (FRNK), a dominant-negative mutant of FAK [20,21].
The humanized anti-HER2 monoclonal antibody trastuzumab (Herceptin) was made by Genentech (San Francisco, CA, USA). PI3-K-specific inhibitor LY294002 was obtained from CalBiochem (San Diego, CA, USA), and the estrogen receptor (ER) antagonist ICI 182,780 was purchased from Tocris (Ballwin, MO, USA). Doxorubicin (Adriamycin) was ordered from the pharmacy of MD Anderson Cancer Center. All other reagents were purchased from Sigma-Aldrich (St Louis, MO, USA).
cDNA and transient expression
The pcDNA3 expression construct containing HER3 was provided by Dr Xiaofeng Le (MD Anderson Cancer Center), and the expression constructs of FAK and FRNK (pCMV-Myc) were kindly provided by Dr Thomas Parsons (University of Virginia, Charlottesville, VA, USA). Transient transfection was performed with the FuGENE 6 transfection kit, in accordance with instructions provided by the manufacturer (Roche Diagnostic, Indianapolis, IN, USA).
Western blot analysis and Akt kinase assay
Western blot analysis and Akt kinase assay were performed as described previously [16,19].
Cytoplasmic and nuclear fractionation
The method for cytoplasmic and nuclear fractionation was adopted from the literature [22,23] with minor modifications. In brief, pellets containing 2 × 107 cells were resuspended into 800 μl of buffer A (50 mM NaCl, 10 mM HEPES pH 8.0, 500 mM sucrose, 1 mM EDTA, 0.5 mM spermidine, 0.15 mM spermine, 0.2% Triton X-100, 1 mM phenylmethylsulphonyl fluoride, 2 mM Na3VO4, 25 μg/ml leupeptin, 25 μg/ml aprotinin). After incubation on ice for 10 min, the cells were homogenized with 10 strokes in a Dounce homogenizer. A small aliquot of the cell homogenates was then examined under a microscope to confirm that more than 98% of cells were lysed. After brief centrifugation of the cell homogenates at 4°C, the supernatant (cytoplasmic fraction) was collected and the pellet was washed twice with 400 μl of buffer B (50 mM NaCl, 10 mM HEPES pH 8, 25% glycerol, 0.1 mM EDTA, 0.5 mM spermidine, 0.15 mM spermine) and then resuspended in 150 μl of buffer C (350 mM NaCl, 10 mM HEPES pH 8.0, 25% glycerol, 0.1 mM EDTA, 0.5 mM spermidine, 0.15 mM spermine) with gentle rocking for 30 min at 4°C [22]. After centrifugation, the supernatant (nuclear fraction) was collected. The amounts of protein in the cytoplasmic and nuclear fractions were determined with the Bradford method (Bio-Rad, Hercules, CA, USA).
Ionizing radiation
Cells grown on Petri dishes were irradiated with γ-rays from a high-dose-rate 137Cs unit (4.5 Gy/min) at room temperature (25 – 27°C), as described previously [17,19]. After irradiation, the cells were harvested by trypsinization.
Results
Differential responses in the baseline levels of Akt phosphorylation and kinase activity in a panel of breast cancer cell lines after treatment with doxorubicin
To assess the cellular responses in breast cancer cells in the baseline levels of Akt phosphorylation and activity as a result of doxorubicin treatment, we first examined the level of Akt phosphorylation and activation in MCF7 breast cancer cells after treatment with doxorubicin. Figure 1a shows a time-dependent induction in the levels of p-Akt with reference to the total levels of Akt in MCF7 cells treated with 1 μM doxorubicin, a dose that we have shown previously to induce apoptosis in the cells [16,18]. An increase in p-Akt level was detected as early as after 1 hour of exposure of the cells to doxorubicin, and a robust increase in the level of p-Akt was observed 24 hours after treatment.
We next expanded the experiment in a panel of eight breast cancer cell lines treated with increasing concentrations (0.125 to 1 μM) of doxorubicin for 24 hours in culture medium supplemented with 10% FBS. Interestingly, it was found that the changes in the levels of p-Akt varied between the cell lines after the treatment. In comparison with control cells, which were kept untreated for 24 hours in the same type of culture medium, MCF7, MDA468 and T47D cells showed a dose-dependent increase in p-Akt levels; in contrast, SKBR3 and MDA231 cells showed a dose-dependent decrease, and no or minimal change was detected in MDA361, MDA157 and BT474 cells (Fig. 1b). As expected, no changes in total Akt expression were found in the cell lines after the treatment. These results suggest that genetic context among individual cell lines might have a role in determining the cellular responses to the treatment.
To confirm that the phosphorylation of Akt induced by doxorubicin was associated with an increased Akt kinase activity, we assessed Akt activity by in vitro Akt kinase assay on two known Akt substrates, Bad and GSK3, in MCF7 cells. Figure 2a shows that, in comparison with untreated MCF7 cells and with the cells treated with type 1 insulin-like growth factor (IGF-1), the cells treated with doxorubicin contained an increased level of p-Akt, which was comparable to the increase of p-Akt level stimulated by IGF-1. Treatment of the cells with ionizing radiation induced a similar increase in the level of p-Akt. The increases in p-Akt level induced by doxorubicin or radiation were associated with increased Akt kinase activities measured by the Akt in vitro kinase assay (Fig. 2b). We found that the Akt protein immunoprecipitated from doxorubicin-treated or γ-ray-irradiated cells phosphorylated both Bad and GSK3 as strongly as the Akt protein from the IGF-1-treated cells.
As another measure of the functional status of Akt after treatment with doxorubicin or ionizing radiation, we also examined the translocation of Akt from the cytoplasm to the nucleus. To allow the detection of the signals of Akt from cytoplasmic to nuclear translocation, we raised the level of Akt expression in MCF7 cells by transient transfection of the cells with a His-tagged Akt1 expression construct 48 hours before harvest. Both the doxorubicin-induced and radiation-induced increases in Akt phosphorylation were associated with increased translocation of Akt from the cytoplasm to the nucleus (Fig. 2c).
To determine the extent to which the doxorubicin-induced activation of Akt is regulated by the PI3-K pathway, we explored this question with MCF7 cells, which express a relatively low baseline level of p-Akt, and MDA468 cells, which express a relatively high baseline level of p-Akt because of the mutation status of PTEN (phosphatase and tensin homolog deleted on chromosome ten) in the cells [24]. We found that an overnight (16 hours) exposure of the cells to LY294002, a PI3-K-specific inhibitor, remarkably abolished the increase in Akt phosphorylation after treatment with doxorubicin in both MCF7 and MDA468 cells (Fig. 2d). The results indicate that the doxorubicin-induced phosphorylation and activation of Akt were mediated through a PI3-K dependent pathway.
Roles of HER family members in doxorubicin-induced activation of Akt
Because the doxorubicin-induced activation of Akt is dependent on PI3-K activity, we proposed that the breast cancer cells with compelling molecular components of the PI3-K pathway might show an enhanced cellular response to doxorubicin-induced activation of Akt. The HER family members are important upstream regulators of the PI3-K/Akt pathway and are known to be important in the progression of breast cancer and its resistance to chemotherapy or radiotherapy [25,26]. To determine the extent to which HER family members might potentiate the cellular response to doxorubicin-induced activation of Akt in breast cancer cells, we assessed the effect of treatment with doxorubicin (0.5 to 1 μM) on p-Akt levels in MCF7 cells transfected with a HER2 expression construct (MCF7HER2 cells). In comparison with control vector-transfected MCF7 cells (MCF7neo), MCF7HER2 cells showed not only a higher baseline level of p-Akt (Fig. 3, compare lanes 1 and 7) but also an enhanced response to the doxorubicin-induced increase in Akt phosphorylation (Fig. 3, compare lanes 1 to 3 with lanes 7 to 9). A caveat is that it is unlikely that the enhancement was caused by an additive effect of Akt phosphorylation by doxorubicin treatment and HER2 overexpression in the cells, because treatment of MCF7neo cells with trastuzumab also decreased the level of doxorubicin-induced phosphorylation of Akt. As expected, we detected no changes in the level of total Akt. The increase in the levels of p-Akt in MCF7neo and MCF7HER2 cells by doxorubicin was markedly diminished by pretreatment with trastuzumab, which downregulates HER2 in these cells (Fig. 3, compare lanes 1 to 3 with lanes 4 to 6, and lanes 7 to 9 with lanes 10 to 12). Taken together, these results indicate that the higher level of HER2 in MCF7HER2 cells potentiates the response of the cells to doxorubicin-induced activation of Akt.
Interestingly, some cell lines including SKBR3 cells showed a decline in the level of p-Akt after treatment with doxorubicin (Fig. 1), despite the fact that SKBR3 cells express an appreciable level of HER2 [27,28]. A notable difference between MCF7 and SKBR3 cells is that the former expresses HER3 whereas the latter has no detectable level of HER3 expression [16]. Of the HER family members, HER3 contains the most PI3-K-binding sites, but it is kinase-deficient and is mainly activated though heterodimerization with other HER members [29]. We proposed that an insufficient level of HER3 expression might affect the response of SKBR3 cells to treatment with doxorubicin. To test this hypothesis we transiently transfected SKBR3 cells with a HER3 expression construct. Figure 4 shows that, in comparison with control vector-transfected SKBR3 cells, transient expression of HER3 prevented the decline in the level of p-Akt after doxorubicin treatment in SKBR3 cells. It is noteworthy that, in this particular experiment, HER3 was only transiently transfected into the SKBR3 cells, with an estimated 10 to 15% transfection efficiency. Given the result from the mixed (transfected and untransfected) cells, it is reasonable to speculate that selected clonal or pooled HER3-expressing SKBR3 cells would exhibit a pattern of response similar to that observed in MCF7 cells.
Exposure of the transiently transfected cells to doxorubicin also led to a decrease in the level of HER3, the mechanism of which is unknown. We speculate that it might be related to a degradation of the protein after heterodimerization with HER2. Nevertheless, the transient expression of HER3 in only a small fraction of the cell population (10 to 15%; data not shown) prevented the decline in p-Akt after treatment with doxorubicin in a HER2-overexpressing cell line (SKBR3) suggests a potential cooperative role of HER2 and HER3 in the increase in Akt activity after treatment with doxorubicin. Thus, the ability of HER2 to potentiate the cellular response of Akt phosphorylation or activation after treatment with doxorubicin depends on the cell types.
Involvement of FAK in doxorubicin-triggered phosphorylation and activation of Akt
To broaden the implication of our findings, we sought to assess possible roles of other signal pathways that might also potentiate the cellular response of Akt phosphorylation of MCF7 cells after treatment with doxorubicin. In addition to the HER family members, the FAK pathway is also known to modulate the PI3-K pathway. The FAK pathway is regulated by the interaction between extracellular matrix receptors and integrins, and is often augmented in human breast cancer cells [30,31]. We therefore transiently transfected MCF7 cells with an expression construct of FAK or its dominant-negative counterpart, FRNK. In comparison with control vector-transfected cells, which exhibited a LY294002-sensitive increase in the level of p-Akt over baseline, FAK-transfected cells had a higher p-Akt level both at baseline and after treatment with doxorubicin and were sensitive to LY294002 (Fig. 5). In contrast, transfection of MCF7 cells with FRNK led to a lower phosphorylation level of Akt after treatment with doxorubicin. Irrespective of the expression of FAK or FRNK, the level of total Akt remained unchanged. Taken together, these results suggest that molecular components, such as FAK and HER3, that enhance the cellular sensitivity and responsiveness for PI3-K activation might potentiate the cellular responses of Akt phosphorylation to treatment with doxorubicin.
Effects of estrogen on doxorubicin-induced phosphorylation and activation of Akt
To determine whether the signaling pathways known to modulate the activity of PI3-K/Akt might unanimously potentiate the cellular response of Akt phosphorylation to treatment with doxorubicin, we examined the effect of doxorubicin on the level of p-Akt in MCF7 cells cultured in medium supplemented with an ER antagonist or in estrogen-depleted medium. Estrogen is known to be involved in the regulation of Akt phosphorylation in both ER-positive and ER-negative breast cancer cells [12,32,33]. In comparison with vehicle-treated cells, MCF7 cells stimulated with estrogen (estradiol) showed a higher level of p-Akt, which was decreased when an ER antagonist (ICI 182,780) was present in the culture medium (Fig. 6a). In contrast with the results shown in Figs 4 and 5, we observed no difference in the levels of p-Akt after doxorubicin treatment in MCF7 cells cultured in regular 0.5% FBS medium, charcoal-stripped FBS (that is, estrogen-depleted) medium, or regular 0.5% FBS medium plus ICI 182,780 (Fig. 6b). These results suggested that at least the PI3-K signaling regulated by estrogen does not potentiate the cellular responsiveness to doxorubicin-induced phosphorylation of Akt.
Discussion
In our present study we found that the activity of Akt, an important signal molecule that promotes cell survival and confers cellular resistance to chemotherapy and radiotherapy as shown by us [16,18,19] and others [34,35], was transiently elevated in a subset of breast cancer cell lines as a result of exposure to doxorubicin, a chemotherapeutic agent commonly used to treat patients with breast cancers. Activation of Akt in MCF7 cells after exposure to doxorubicin was reported earlier, but the mechanism was not explored in detail [34,35]. We noted here that, in comparison with resting (non-stimulated) cells, in which most Akt was found in the cytoplasm, exposure of the cells to doxorubicin or ionizing radiation led to a relocation of Akt to the nucleus. It is noteworthy that several antiapoptotic substrates of Akt are nuclear proteins. This subcellular translocation of Akt is important for cells to overcome the death signals initiated by treatment with doxorubicin or ionizing radiation. Taken together with our previous results [16-18], the present results suggest that doxorubicin-triggered activation of Akt has a role in the resistance of breast cancer cells to this drug and that the same might apply to radiotherapy.
Because the overall cellular sensitivity of breast cancer cells to chemotherapy or radiotherapy is attributed to multiple intrinsic and extrinsic factors, such as p53 status, Bcl-2/Bax levels, expression of multiple drug resistance proteins, and hypoxic status, a caveat is that our data do not necessarily imply that one group of breast cancer cells showing increases in the level of p-Akt after chemotherapy or radiotherapy would absolutely be more chemoresistant or radioresistant than the another group of breast cancer cells without showing such a response. Rather, the data indicate that the activation and phosphorylation of Akt triggered by chemotherapy or radiotherapy contribute to the overall cellular sensitivity to these conventional therapies.
Several questions remain to be fully answered. First, why was Akt activation after treatment with doxorubicin found in only some of the breast cancer cell lines we tested? Apparently, cells must be equipped with certain molecular components that enable them to react to signals induced by chemotherapy or radiotherapy. We found that the drug-triggered activation of Akt depends on the activity of PI3-K, which can be activated by several known pathways, some of which we have explored in the present study. Which pathway is activated depends on the genetic context and functional status of the signal transduction network in individual cell types. In our study, MCF7 cells transiently expressing a high level of HER2 potentiated the response of the cells to the doxorubicin-induced activation of Akt. This result is consistent with those shown recently by us [16,19] and others [36-38] indicating that HER2 expression in breast cancer cells might render them more resistant to chemotherapy or radiotherapy.
However, a high level of HER2 expression alone might not be sufficient to mediate this response. For example, we detected no change in the level of p-Akt in BT474 breast cancer cells after treatment with doxorubicin, even though they expressed a high level of HER2. SKBR3, another breast cancer cell line that expresses high levels of HER2, even showed a reduced level of p-Akt after treatment with doxorubicin. Expression of a transient transfected HER3 in the SKBR3 cells prevented this decline, indicating that heterodimerization and crosstalk between HER2 and HER3 might be important in mediating the downstream pathway that leads to Akt activation in breast cancer cells after treatment with doxorubicin. This might explain the negative findings from a recent clinical study reporting that HER2 overexpression does not seem to predispose to locoregional recurrence for breast cancer patients treated with neoadjuvant doxorubicin-based chemotherapy, mastectomy and radiotherapy [39].
A second question is what molecular executioner leads to the activation of Akt after chemotherapy or radiotherapy. Are any soluble factors or non-secreted membrane-bound ligands involved, or is the PI3-K/Akt pathway activated directly and autonomously? In our study, we demonstrated that several different mechanisms, two of which are the expression of HER2 and of FAK, might enhance the doxorubicin-induced activation of Akt. Each mechanism activates PI3-K but does so through different ligands. Interference with these pathways by the anti-HER2 monoclonal antibody trastuzumab or by a dominant-negative mutant FAK (FRNK) abolished the drug-triggered activation of Akt mediated by HER2 and FAK, respectively. An interesting finding from our studies is that not all stimuli that lead to PI3-K activation enhance the drug-triggered activation of Akt. For example, abnormal estrogen exposure is associated with an increased risk of breast cancer, and estrogen is known to activate Akt via a non-nuclear estrogen-signaling pathway involving the direct interaction of ER with PI3-K [40].
The ER isoform ERα binds to the p85α regulatory subunit of PI3-K in a ligand-dependent manner. Stimulation with estrogen increases ERα-associated PI3-K activity, leading to the activation of Akt. This interaction between ER and p85α is independent of gene transcription and does not involve phosphotyrosine adapter molecules or Src-homology domains of p85α [40]. We found that the ER antagonist ICI 182,780 blocked estrogen-induced Akt activation in the ER-positive MCF7 cells but did not affect doxorubicin-induced Akt activation. Depletion of estrogen from the culture medium did not affect the doxorubicin-induced activation of Akt either. These data suggest that estrogen-induced signals, whether dependent on ER or not, are not involved in the pathway that enhances the doxorubicin-induced activation of Akt.
In fact, this atypical activation of Akt seems not to be limited to doxorubicin or ionizing radiation. We have observed that treatment of MCF7 cells with several different drugs (paclitaxel, 5-flurouracil and gemcitabine) that act through different mechanisms can also induce Akt phosphorylation, although the response and the timing and dose required for this effect varied between the drugs tested (data not shown). Cellular stress such as hypoxia and ultraviolet radiation has been reported by others to induce PI3-K-dependent Akt activation [41-43]. Thus, inherent properties of individual cell types, rather than specific cell death signals, might determine whether Akt is activated after cells are exposed to stresses. Cancer cells with functional aberrations, such as overexpression of HER family members or increased cell adhesion potential, are probably more capable than noncancerous cells of activating Akt as a defensive mechanism against external detrimental stimuli, which justifies a novel approach of targeting the PI3-K/Akt for chemosensitization or radiosensitization.
In summary, doxorubicin might cause a PI3-K-dependent increase of Akt activity in breast cancer cells. Together with other recent results of ours [16-18], the present observations suggest that clinical benefits in treating patients with breast cancer could be obtained with appropriate combinations of novel Akt inhibitors and conventional chemotherapeutic drugs or ionizing radiation.
Conclusion
We found that the activities of Akt are increased in selected cell lines treated with doxorubicin, which is a PI3-K-dependent process and is potentiated after overexpression of HER2/HER3 receptor tyrosine kinases or FAK nonreceptor tyrosine kinase. This therapeutic intervention (doxorubicin)-triggered activation of Akt might have a role in affecting the overall therapeutic responses of cancer cells to the treatment. Clinical benefits in the treatment of breast cancer patients could be obtained with appropriate combinations of novel Akt inhibitors and conventional chemotherapeutic drugs or ionizing radiation. Our observations further justify the efforts of targeting PI3-K/Akt for enhancing the therapeutic responses of breast cancer cells to the conventional therapies.
Abbreviations
ER = estrogen receptor; FBS = fetal bovine serum; FAK = focal adhesion kinase; FRNK = FAK-related non-kinase; GSK3 = glycogen synthase kinase-3; HER2 = human epidermal growth factor receptor-2; IGF-I = type 1 insulin-like growth factor; p-Akt = Ser473-phosphorylated Akt; PI3-K = phosphoinositide 3-kinase.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
The authors' contributions to this research are reflected in the order shown, with the exception of ZF, who supervised all aspects of this research and prepared the manuscript. XL did most Western blot analyses and sample preparations; YL contributed the experiments of cell transfection and some Western blot analyses; KL performed radiation of the cells; BL participated in the overall design of experiments and data interpretation. All authors read and approved the final manuscript.
Acknowledgements
This study was supported in part by a research award from The Breast Cancer Research Foundation (New York).
Figures and Tables
Figure 1 Differential responses to doxorubicin-induced phosphorylation of Akt in a panel of human breast cancer cells. (a) MCF7 cells were exposed to 1 μM doxorubicin (Doxo) for the indicated periods in culture medium supplemented with 10% fetal bovine serum (FBS). The cells were then harvested and prepared for Western blot analyses with antibodies directed against Ser473-phosphorylated Akt (p-Akt) and total Akt. The densitometric levels of p-Akt at each time point were normalized to the corresponding levels of total Akt and are shown in the bar graph. (b) A panel of eight indicated breast cancer cell lines were treated with 0.125 to 1 μM doxorubicin for 24 hours in culture medium supplemented with 10% FBS. After treatment the individual cell lines were harvested, lysed and subjected to Western blot analyses with antibodies against p-Akt and total Akt.
Figure 2 Functional analysis of Akt phosphorylation in human breast cancer cells after treatment with doxorubicin. (a) MCF7 cells were exposed to 1 μM doxorubicin (Doxo) for 24 hours, or irradiated (XRT) with X-rays (5 Gy) and then cultured for a further 24 hours in culture medium supplemented with 10% fetal bovine serum (FBS). MCF7 cells stimulated for 15 min with 10 nM type 1 insulin-like growth factor (IGF-1) were used as a positive control. Untreated (Untx) and the treated MCF7 cells were harvested, lysed and subjected to Western blot analyses with antibodies against Ser473-phosphorylated Akt (p-Akt) and total Akt. (b) MCF7 cells transiently transfected with a His-tagged Akt1 expression construct were left untreated or treated as in (a). Akt1 was immunoprecipitated by using an anti-His tag monoclonal antibody and then assayed for in vitro kinase activities for phosphorylation on Bad and glycogen synthase kinase-3 (GSK3). (c) Cytoplasmic and nuclear fractionations from untreated, doxorubicin-treated and γ-ray-irradiated MCF7 cells were separated as described in Materials and Methods. Equal amounts (40 μg) of cytoplasmic fraction (Cy) and nuclear fraction (Nu) from each sample were subjected to Western blot analysis to determine the distribution of Akt after treatment with doxorubicin or ionizing radiation. β-Actin and poly(ADP-ribose) polymerase (PARP) were used as markers of the cytoplasmic and nuclear fractions, respectively. (d) MCF7 or MDA468 cells were pre-exposed to 10 μM LY294002 for 2 hours before treatment with 1 μM doxorubicin for 24 hours. After treatment the cells were harvested, lysed and subjected to Western blot analyses with antibodies against p-Akt and total Akt.
Figure 3 Potentiation to doxorubicin-induced Akt phosphorylation in MCF7 cells after expressing high levels of HER2. Control vector-transfected MCF7 cells (MCF7neo) and MCF7 transfectants expressing high levels of human epidermal growth factor receptor-2 (MCF7HER2) were left untreated or were treated with 0.5 or 1 μM doxorubicin (Doxo) for 24 hours. Separate dishes of MCF7neo and MCF7HER2 cells were pretreated with 20 nM trastuzumab (Herceptin) 24 hours before treatment with doxorubicin. After treatment the cells were harvested, lysed and subjected to Western blot analyses with antibodies against HER2, Ser473-phosphorylated Akt (p-Akt) and total Akt. The densitometric levels of p-Akt were normalized to the respective levels of total Akt in each lane and are shown in the bar graph.
Figure 4 Effect of HER3 expression on decrease in Akt phosphorylation in SKBR3 cells after doxorubicin treatment. SKBR3 cells were transiently transfected with a control vector (pcDNA3) or a HER3 expression construct for 24 hours, and then exposed to 0.125 to 1 μM doxorubicin for a further 24 hours. After treatment the cells were harvested, lysed and subjected to Western blot analyses with antibodies against HER3, Ser473-phosphorylated Akt (p-Akt) and total Akt. The densitometric levels of p-Akt were normalized to the respective levels of total Akt in each lane and are shown in the bar graph.
Figure 5 Potentiation of cellular response to doxorubicin-induced Akt phosphorylation in MCF7 cells by FAK. MCF7 cells were transiently transfected for 24 hours with a control vector (pcDNA3.1), a focal adhesion kinase (FAK) expression construct or a FAK-related non-kinase (FRNK) expression construct, followed by exposure to 1 μM doxorubicin (Doxo) for a further 24-hour of culture, with or without the addition of the phosphoinositide 3-kinase-specific inhibitor LY294002 (20 μM) 16 hours before the end of exposure to doxorubicin. After treatment the cells were harvested, lysed and subjected to Western blot analyses with antibodies against FAK, FRNK, Ser473-phosphorylated Akt (p-Akt) and total Akt. The densitometric levels of p-Akt were normalized to the respective levels of total Akt in each lane and are shown in the bar graph.
Figure 6 Potential involvement of estrogen-mediated signaling on doxorubicin-induced Akt phosphorylation. (a) MCF7 cells grown to subconfluence were exposed to 1 μM ICI 182,780 (an estrogen receptor antagonist) or the vehicles (dimethyl sulfoxide [DMSO] and ethanol [EtOH]) for 30 min, and then treated with or without 10 nM estradiol (E2) for a further 30 min. Cell lysates were prepared for Western blot analysis with antibodies against Ser473-phosphorylated Akt (p-Akt) and total Akt. (b) MCF7 cells were cultured in regular 0.5% fetal bovine serum (FBS) medium, charcoal-stripped (CS) FBS medium, or regular 0.5% FBS plus 1 μM ICI 182,780. The cells were then treated with 0.25 to 1 μM doxorubicin (Doxo) for 24 hours. After treatment the cells were harvested, lysed and subjected to Western blot analyses with antibodies against p-Akt and total Akt.
==== Refs
del Peso L Gonzalez-Garcia M Page C Herrera R Nunez G Interleukin-3-induced phosphorylation of BAD through the protein kinase Akt Science 1997 278 687 689 9381178 10.1126/science.278.5338.687
Cardone MH Roy N Stennicke HR Salvesen GS Franke TF Stanbridge E Frisch S Reed JC Regulation of cell death protease caspase-9 by phosphorylation Science 1998 282 1318 1321 9812896 10.1126/science.282.5392.1318
Ozes ON Mayo LD Gustin JA Pfeffer SR Pfeffer LM Donner DB NF-κB activation by tumour necrosis factor requires the Akt serine-threonine kinase Nature 1999 401 82 85 10485710 10.1038/43466
Romashkova JA Makarov SS NF-κB is a target of AKT in anti-apoptotic PDGF signalling Nature 1999 401 86 90 10485711 10.1038/43474
Fry MJ Phosphoinositide 3-kinase signalling in breast cancer: how big a role might it play? Breast Cancer Res 2001 3 304 312 11597319 10.1186/bcr312
Cantrell DA Phosphoinositide 3-kinase signalling pathways J Cell Sci 2001 114 1439 1445 11282020
Xia H Nho RS Kahm J Kleidon J Henke CA Focal adhesion kinase is upstream of phosphatidylinositol 3-kinase/Akt in regulating fibroblast survival in response to contraction of type I collagen matrices via a β1 integrin viability signaling pathway J Biol Chem 2004 279 33024 33034 15166238 10.1074/jbc.M313265200
Banfic H Tang X Batty IH Downes CP Chen C Rittenhouse SE A novel integrin-activated pathway forms PKB/Akt-stimulatory phosphatidylinositol 3,4-bisphosphate via phosphatidylinositol 3-phosphate in platelets J Biol Chem 1998 273 13 16 9417038 10.1074/jbc.273.1.13
Rodriguez-Viciana P Warne PH Dhand R Vanhaesebroeck B Gout I Fry MJ Waterfield MD Downward J Phosphatidylinositol-3-OH kinase as a direct target of Ras Nature 1994 370 527 532 8052307 10.1038/370527a0
Rodriguez-Viciana P Warne PH Vanhaesebroeck B Waterfield MD Downward J Activation of phosphoinositide 3-kinase by interaction with Ras and by point mutation EMBO J 1996 15 2442 2451 8665852
Sun M Paciga JE Feldman RI Yuan Z Coppola D Lu YY Shelley SA Nicosia SV Cheng JQ Phosphatidylinositol-3-OH Kinase (PI3K)/AKT2, activated in breast cancer, regulates and is induced by estrogen receptor α (ERα) via interaction between ERα and PI3K Cancer Res 2001 61 5985 5991 11507039
Hisamoto K Ohmichi M Kurachi H Hayakawa J Kanda Y Nishio Y Adachi K Tasaka K Miyoshi E Fujiwara N Taniguchi N Murata Y Estrogen induces the Akt-dependent activation of endothelial nitric-oxide synthase in vascular endothelial cells J Biol Chem 2001 276 3459 3467 11044445 10.1074/jbc.M005036200
Yu X Rajala RV McGinnis JF Li F Anderson RE Yan X Li S Elias RV Knapp RR Zhou X Cao W Involvement of insulin/phosphoinositide 3-kinase/Akt signal pathway in 17 beta-estradiol-mediated neuroprotection J Biol Chem 2004 279 13086 13094 14711819 10.1074/jbc.M313283200
Vanhaesebroeck B Leevers SJ Panayotou G Waterfield MD Phosphoinositide 3-kinases: a conserved family of signal transducers Trends Biochem Sci 1997 22 267 272 9255069 10.1016/S0968-0004(97)01061-X
Vanhaesebroeck B Alessi DR The PI3K-PDK1 connection: more than just a road to PKB Biochem J 2000 346 561 576 10698680 10.1042/0264-6021:3460561
Knuefermann C Lu Y Liu B Jin W Liang K Wu L Schmidt M Mills GB Mendelsohn J Fan Z HER2/PI-3K/Akt activation leads to a multidrug resistance in human breast adenocarcinoma cells Oncogene 2003 22 3205 3212 12761490 10.1038/sj.onc.1206394
Liang K Jin W Knuefermann C Schmidt M Mills GB Ang KK Milas L Fan Z Targeting the phosphatidylinositol 3-kinase/Akt pathway for enhancing breast cancer cells to radiotherapy Mol Cancer Ther 2003 2 353 360 12700279
Jin W Wu L Liang K Liu B Lu Y Fan Z Roles of the PI-3K and MEK pathways in Ras-mediated chemoresistance in breast cancer cells Br J Cancer 2003 89 185 191 12838322 10.1038/sj.bjc.6601048
Liang K Lu Y Jin W Ang KK Milas L Fan Z Sensitization of breast cancer cells to radiation by trastuzumab Mol Cancer Ther 2003 2 1113 1120 14617784
Golubovskaya V Beviglia L Xu LH Earp HS Craven R IIICance W Dual inhibition of focal adhesion kinase and epidermal growth factor receptor pathways cooperatively induces death receptor-mediated apoptosis in human breast cancer cells J Biol Chem 2002 277 38978 38987 12167618 10.1074/jbc.M205002200
Schaller MD Borgman CA Parsons JT Autonomous expression of a noncatalytic domain of the focal adhesion-associated protein tyrosine kinase pp125FAK Mol Cell Biol 1993 13 785 791 8423801
Zhou BP Liao Y Xia W Spohn B Lee MH Hung MC Cytoplasmic localization of p21Cip1/WAF1 by Akt-induced phosphorylation in HER-2/neu-overexpressing cells Nat Cell Biol 2001 3 245 252 11231573 10.1038/35060032
Asada M Yamada T Ichijo H Delia D Miyazono K Fukumuro K Mizutani S Apoptosis inhibitory activity of cytoplasmic p21Cip1/WAF1 in monocytic differentiation EMBO J 1999 18 1223 1234 10064589 10.1093/emboj/18.5.1223
Lu Y Lin YZ LaPushin R Cuevas B Fang X Yu SX Davies MA Khan H Furui T Mao M The PTEN/MMAC1/TEP tumor suppressor gene decreases cell growth and induces apoptosis and anoikis in breast cancer cells Oncogene 1999 18 7034 7045 10597304 10.1038/sj.onc.1203183
Albanell J Baselga J The ErbB receptors as targets for breast cancer therapy J Mammary Gland Biol Neoplasia 1999 4 337 351 10705918 10.1023/A:1018762216147
Abd El-Rehim DM Pinder SE Paish CE Bell JA Rampaul RS Blamey RW Robertson JF Nicholson RI Ellis IO Expression and co-expression of the members of the epidermal growth factor receptor (EGFR) family in invasive breast carcinoma Br J Cancer 2004 91 1532 1542 15480434 10.1038/sj.bjc.6602184
Jallal B Schlessinger J Ullrich A Tyrosine phosphatase inhibition permits analysis of signal transduction complexes in p185HER2/neu-overexpressing human tumor cells J Biol Chem 1992 267 4357 4363 1347042
D'Souza B Berdichevsky F Kyprianou N Taylor-Papadimitriou J Collagen-induced morphogenesis and expression of the alpha 2-integrin subunit is inhibited in c-erbB2-transfected human mammary epithelial cells Oncogene 1993 8 1797 1806 8099725
Olayioye MA Neve RM Lane HA Hynes NE The ErbB signaling network: receptor heterodimerization in development and cancer EMBO J 2000 19 3159 3167 10880430 10.1093/emboj/19.13.3159
Weiner TM Liu ET Craven RJ Cance WG Expression of focal adhesion kinase gene and invasive cancer Lancet 1993 342 1024 1025 8105266 10.1016/0140-6736(93)92881-S
Cance WG Harris JE Iacocca MV Roche E Yang X Chang J Simkins S Xu L Immunohistochemical analyses of focal adhesion kinase expression in benign and malignant human breast and colon tissues: correlation with preinvasive and invasive phenotypes Clin Cancer Res 2000 6 2417 2423 10873094
Ahmad S Singh N Glazer RI Role of AKT1 in 17beta-estradiol- and insulin-like growth factor I (IGF-I)-dependent proliferation and prevention of apoptosis in MCF-7 breast carcinoma cells Biochem Pharmacol 1999 58 425 430 10424760 10.1016/S0006-2952(99)00125-2
Tsai EM Wang SC Lee JN Hung MC Akt activation by estrogen in estrogen receptor-negative breast cancer cells Cancer Res 2001 61 8390 8392 11731414
Tari AM Mehta A Lopez-Berestein G Modulation of Akt activity by doxorubicin in breast cancer cells J Chemother 2001 13 334 336 11450894 10.1159/000049727
Clark AS West K Streicher S Dennis PA Constitutive and inducible Akt activity promotes resistance to chemotherapy, trastuzumab, or tamoxifen in breast cancer cells Mol Cancer Ther 2002 1 707 717 12479367
Pietras RJ Poen JC Gallardo D Wongvipat PN Lee HJ Slamon DJ Monoclonal antibody to HER-2/neureceptor modulates repair of radiation-induced DNA damage and enhances radiosensitivity of human breast cancer cells overexpressing this oncogene Cancer Res 1999 59 1347 1355 10096569
Yu D Liu B Tan M Li J Wang SS Hung MC Overexpression of c-erbB-2/neu in breast cancer cells confers increased resistance to Taxol via mdr-1-independent mechanisms Oncogene 1996 13 1359 1365 8808711
Yu D Jing T Liu B Yao J Tan M McDonnell TJ Hung MC Overexpression of ErbB2 blocks Taxol-induced apoptosis by upregulation of p21Cip1, which inhibits p34Cdc2 kinase Mol Cell 1998 2 581 591 9844631 10.1016/S1097-2765(00)80157-4
Buchholz TA Huang EH Berry D Pusztai L Strom EA McNeese MD Perkins GH Schechter NR Kuerer HM Buzdar AU Her2/neu-positive disease does not increase risk of locoregional recurrence for patients treated with neoadjuvant doxorubicin-based chemotherapy, mastectomy, and radiotherapy Int J Radiat Oncol Biol Phys 2004 59 1337 1342 15275718 10.1016/j.ijrobp.2004.02.018
Simoncini T Hafezi-Moghadam A Brazil DP Ley K Chin WW Liao JK Interaction of oestrogen receptor with the regulatory subunit of phosphatidylinositol-3-OH kinase Nature 2000 407 538 541 11029009 10.1038/35035131
Alvarez-Tejado M Naranjo-Suarez S Jimenez C Carrera AC Landazuri MO del Peso L Hypoxia induces the activation of the phosphatidylinositol 3-kinase/Akt cell survival pathway in PC12 cells: protective role in apoptosis J Biol Chem 2001 276 22368 22374 11294857 10.1074/jbc.M011688200
Beitner-Johnson D Rust RT Hsieh TC Millhorn DE Hypoxia activates Akt and induces phosphorylation of GSK-3 in PC12 cells Cell Signal 2001 13 23 27 11257444 10.1016/S0898-6568(00)00128-5
Alvarez-Tejado M Alfranca A Aragones J Vara A Landazuri MO del Peso L Lack of evidence for the involvement of the phosphoinositide 3-kinase/Akt pathway in the activation of hypoxia-inducible factors by low oxygen tension J Biol Chem 2002 277 13508 13517 11815624 10.1074/jbc.M200017200
|
16168102
|
PMC1242125
|
CC BY
|
2021-01-04 16:04:33
|
no
|
Breast Cancer Res. 2005 May 24; 7(5):R589-R597
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1259
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12601616810410.1186/bcr1260Research ArticleGreatly increased occurrence of breast cancers in areas of mammographically dense tissue Ursin Giske [email protected] Linda [email protected] Yuri R [email protected] Malcolm C [email protected] Anna H [email protected] Department of Preventive Medicine, Keck School of Medicine, University of Southern California/Norris Comprehensive Cancer Center, Los Angeles, California, USA2 Department of Radiology, Keck School of Medicine, University of Southern California/Norris Comprehensive Cancer Center, Los Angeles, California, USA2005 8 6 2005 7 5 R605 R608 24 1 2005 8 3 2005 23 4 2005 29 4 2005 Copyright © 2005 Ursin et al. licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Mammographic density is a strong, independent risk factor for breast cancer. A critical unanswered question is whether cancers tend to arise in mammographically dense tissue (i.e. are densities directly related to risk or are they simply a marker of risk). This question cannot be addressed by studying invasive tumors because they manifest as densities and cannot be confidently differentiated from the densities representing fibrous and glandular tissue. We addressed this question by studying ductal carcinoma in situ (DCIS), as revealed by microcalcifications.
Method
We studied the cranio-caudal and the mediolateral-oblique mammograms of 28 breasts with a solitary DCIS lesion. Two experienced radiologists independently judged whether the DCIS occurred in a mammographically dense area, and determined the density of different areas of the mammograms.
Results
It was not possible to determine whether the DCIS was or was not in a dense area for six of the tumors. Of the remaining 22 lesions, 21 occurred in dense tissue (test for difference from expected taken as the percentage of density of the 'mammographic quadrant' containing DCIS; P < 0.0001). A preponderance of DCIS (17 out of 28) occurred in the mammographic quadrant with the highest percentage density.
Conclusion
DCIS occurs overwhelmingly in the mammographically dense areas of the breast, and pre-DCIS mammograms showed that this relationship was not brought about by the presence of the DCIS. This strongly suggests that some aspect of stromal tissue comprising the mammographically dense tissue directly influences the carcinogenic process in the local breast glandular tissue.
==== Body
Introduction
On a mammogram fat appears radiolucent or dark, whereas connective and epithelial tissue appear radiodense or white. The amount of mammographic density is a strong independent predictor of breast cancer risk [1-4]. For technical reasons, mammographic density has usually been expressed as mammographic percentage density (MPD; i.e. the ratio of the area of the breast that is dense to the total area of the breast on the mammogram). There is an approximately fivefold increased breast cancer risk in women with 60% or more MPD as compared with women with under 10% MPD, with a steady increase in risk with increasing MPD. Studies in which MPD is used directly have found very similar effects [5-8].
The biological basis for this increased risk for breast cancer associated with increased mammographic densities is not understood, and the detailed nature of densities has not been studied extensively. A fundamental question that has yet to be answered is whether, within a particular breast, cancers tend to occur in dense areas or not. Invasive carcinoma is usually identified as a spiculated dense mass on the mammogram, and even an expert mammographer is frequently unable to distinguish this mass from normal dense breast tissue, making it difficult, and often impossible, to decide whether the tumor arose in a dense area. We used mammograms obtained from patients with ductal carcinoma in situ (DCIS), as evidenced mammographically by microcalcifications, to address this question. DCIS is a 'nonobligate but definite local precursor of invasive carcinoma' [9], which commonly manifests as calcifications on the mammogram and usually can easily be differentiated from dense breast tissue.
Materials and methods
We retrospectively identified consecutive women diagnosed with DCIS at the Henrietta C Lee Breast Center at the USC/Norris Comprehensive Cancer Center, for whom diagnostic mammograms were available at the Breast Center. There were 31 such patients. The study protocol was approved by the Institutional Review Board of the University of Southern California School of Medicine; patient informed consent was not required for this retrospective study, with minimal risk to participant.
For each participant we obtained the mammogram(s) of the affected breast(s) at the time of the DCIS diagnosis, and whenever available we also obtained the most recent pre-DCIS mammogram(s). We eliminated four of these women from further consideration at this stage: one because the DCIS lesions occurred over a wide area; one because the radiologists could not decide on the precise location of the DCIS (see below); and two because the single DCIS lesion occurred in the subareolar area – an area that is uniformly dense. Of the remaining 27 women, one had a single DCIS lesion in both breasts and 26 had a single DCIS lesion in one breast. These 28 lesions comprise the subject of this report. The DCIS tumors ranged in size (length) from 0.5 cm to 9.3 cm, and all but four were less than 4 cm. Eleven tumors were grade 3, eight were grade 2, one was grade 1, and seven were of unknown grade.
For each affected breast, the cranio-caudal (CC) and mediolateral-oblique (MLO) mammographic views were studied. Based on these two mammograms for each DCIS lesion, the two expert mammography radiologists (LHL and YRP) independently coded the DCIS lesion as to whether it was in an area of mammographically dense tissue (yes/no/cannot determine).
The breast images were also divided equally into a lateral and medial part based on the CC image (CC-L and CC-M, respectively); similarly, the MLO view was divided into a superior and inferior part (MLO-S and MLO-I, respectively). The radiologists independently visually assessed the percentage of mammographically dense tissue for each of the four 'mammographic areas' (CC-L, CC-M, MLO-S and MLO-I). The MPD for the whole breast was taken as the average of these four values. The percentage density of the part of the breast ('mammographic quadrant') containing the DCIS lesion was estimated as the average of the two density assessments of the mammographic areas containing the lesion. For instance, if the DCIS lesion occurred in the upper outer clinical quadrant of the breast, then it would be observed in the superior MLO view and the lateral part of the CC view, and the percentage density of the DCIS-containing mammographic quadrant would then be the average density of the CC-L and MLO-S mammographic areas. The percentage density of the part of the breast not containing the DCIS lesion was taken as the average of the remaining two density assessments. It is possible that more optimal measures of the density of that part of the breast containing, and not containing, the DCIS could be made, but this simple averaging measure is sufficient for our purposes here and it is not clear how one would determine a better measure.
Each DCIS lesion was classified as being in dense tissue (score of 1) or in nondense tissue (score of 0) or not determined (scored as a missing value). The expected score for a lesion was taken to be the PMD of the part of the breast containing the DCIS as described above (this was done separately for the estimated percentage density, as recorded by the radiologists individually and as their average values). The total score for the lesions was compared with the total expected score using exact methods for combinations of independent binomial distributions with known expected values. We used the paired t-test method to test the significance of the differences between the densities in affected and nonaffected DCIS areas in a single breast. All P values are two sided.
Results
Table 1 shows the results for our main question of interest, namely whether DCIS occurs preferentially in mammographically dense tissue. Of the 28 images, radiologists 1 and 2 agreed on the rating for 24 of them: 21 as being in dense tissue, one as being in nondense tissue, and two as being associated with a part of the mammogram where the lesion was too closely associated with both dense and nondense tissue to be able to make an assignment confidently. Of the remaining four lesions, radiologist 2 was not willing to make a call on three of them whereas radiologist 1 felt that the lesion was in dense tissue, and there was a single lesion in which the two radiologists disagreed (radiologist 1 called it in nondense tissue whereas radiologist 2 called it as being in dense tissue). Of the 22 informative lesions with agreement between radiologists, 21 were called as being in dense areas (observed 21; expected 10.75 based on the assessment by radiologist 1, 10.77 based on the assessment by radiologist 2, and 10.76 for their average; P < 0.0001 using any of the three expected values).
Table 2 shows the average density (for the two radiologists combined) and the number of DCIS lesions by mammographic area (CC-M, CC-L, MLO-S and MLO-I). Of the 28 lesions, 17 occurred in the lateral-superior mammographic quadrant, whereas only two occurred in the medial-inferior mammographic quadrant. This correlated strongly with the average percentage density in the different mammographic quadrants, which varied from 55.8% in the lateral-superior mammographic quadrant (i.e. average of densities in the CC-L and MLO-S) to 38.3% in the medial-inferior mammographic quadrant (i.e. average of densities in the CC-M and MLO-I). Not surprisingly, we also found significant differences in densities between the lateral (n = 21) and medial (n = 7) areas of the mammographic CC view (51.3% for CC-L versus 33.5% for CC-M), and between the superior (n = 22) and the inferior (n = 6) areas of the MLO view (47.6% for MLO-S versus 34.7% for MLO-I). These two mammographic views are highly correlated; for example, the correlation between the percentage densities of the CC-L area and the MLO-S area was 0.93.
We also considered tumor grade and tumor size in our analysis. We found no clear association between tumor grade and percentage density, and the results were essentially unchanged when we excluded the four women with tumors greater than 4 cm (data not shown).
Discussion
The findings of this study show that DCIS occurs overwhelmingly in the areas of the breast that are mammographically dense. We identified previous mammograms showing no DCIS for 13 of the 21 mammograms with DCIS in a dense area; all 13 showed that the areas subsequently showing DCIS were clearly dense at the time of the earlier mammogram.
Our results further show that DCIS occurs in the part of the breast that has the highest percentage density on the mammogram, namely the lateral-superior mammographic view. Although the difference in densities between the lateral-superior and the medial-lower mammographic quadrants (55.8% versus 38.3%) is highly statistically significant (P < 0.0001), the magnitude of the difference does not mirror the difference in frequency of DCIS and suggests that the amount of mammographically dense tissue is not the only component of the breast that plays a key role in the carcinogenic process.
Mammographic density is a very strong risk factor for breast cancer. The findings reported here suggest that this is a direct causal connection between the dense tissue and the breast glandular tissue. One possible biological basis for this would be that breast glandular tissue is overwhelmingly concentrated in the (mammographically) dense areas of the breast. Although, in our experience, this is generally believed to be true, there are surprisingly few data on this, and a recent study [10] found no correlation between the amount of glandular tissue in a breast and the mammographic density of the breast. There is much evidence from studies in rodents of a complex interaction between breast stroma and epithelial tissue [11], and increased epithelial cell proliferation secondary to increased density could explain why mammographic density is associated with an increased risk for epithelial malignancy.
Mammographic densities are themselves changed by interventions that affect breast cancer risk. Selective estrogen receptor modulators reduce breast cancer risk and reduce densities [12-14], and estrogen–progestin replacement therapy increases breast cancer risk and increases densities [15,16]. The interaction between the stroma and the epithelium may therefore be two way.
Conclusion
Much remains to be learned, but this study shows that it is the nature of the interaction between stromal and epithelial tissue that should be the major focus of breast cancer research.
Abbreviations
CC = cranio-caudal; DCIS = ductal carcinoma in situ; MLO = mediolateral-oblique; MPD = mammographic percentage density.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All authors participated in the design of the study. GU coordinated the study, drafted the manuscript and contributed to the statistical analysis. LHL and YRP analyzed the mammograms. MCP contributed to the statistical analysis. AHW helped draft the manuscript. AHW and MCP conceived of the study. All authors read and approved the final manuscript.
This work was supported by the USC/Norris Comprehensive Cancer Center Core Grant P30 CA14089 (MCP). The funding source had no role in this report.
Figures and Tables
Table 1 Relation of DCIS lesion to dense tissue
Radiologist 2
Dense Nondense Cannot determine
Radiologist 1 Dense 21 1 0
Nondense 0 1 0
Cannot determine 3 0 2
DCIS, ductal carcinoma in situ.
Table 2 Average mammographic density and location of DCIS by 'mammographic quadrant' of the breast
Mammographic quadrant
CC MLO Density (%; mean ± SE) Number of DCIS lesions
L S 55.8 ± 4.2 17
M S 47.9 ± 4.0 5
L I 46.2 ± 4.0 4
M I 38.3 ± 4.0 2
CC, cranio-caudal; DCIS, ductal carcinoma in situ; I, inferior; L, lateral; M, medial; MLO, mediolateral-oblique; S, superior; SE, standard error.
==== Refs
Saftlas AF Szklo M Mammographic parenchymal patterns and breast cancer risk Epidemiol Rev 1987 9 146 174 3315715
Oza AM Boyd NF Mammographic parenchymal patterns: a marker of breast cancer risk Epidemiol Rev 1993 15 196 208 8405204
Warner E Lockwood G Tritchler D Boyd NF The risk of breast cancer associated with mammographic parenchymal patterns: a meta-analysis of the published literature to examine the effect of method of classification Cancer Detect Prev 1992 16 67 72 1532349
Boyd NF Lockwood GA Byng JW Tritchler DL Yaffe MJ Mammographic densities and breast cancer risk Cancer Epidemiol Biomarkers Prev 1998 12 1133 1144
Brisson J Merletti F Sadowsky NL Twaddle JA Morrison AS Cole P Mammographic features of the breast and breast cancer risk Am J Epidemiol 1982 115 428 437 7064977
Boyd NF Byng JW Jong RA Fishell EK Little LE Miller AB Lockwood GA Tritchler DL Yaffe MJ Quantitative classification of mammographic densities and breast cancer risk: results from the Canadian National Breast Screening Study J Natl Cancer Inst 1995 87 670 675 7752271
Byrne C Schairer C Wolfe J Parekh N Salane M Brinton LA Hoover R Haile R Mammographic features and breast cancer risk: effects with time, age and menopause status J Natl Cancer Inst 1995 87 1622 1629 7563205
Ursin G Ma H Wu AH Bernstein L Salane M Parisky YR Astrahan M Siozon CC Pike MC Mammographic density and breast cancer in three ethnic groups Cancer Epidemiol Biomarkers Prev 2003 12 332 338 12692108
Page DL Rogers LW Schuyler PA Dupont WD Jensen RA Silverstein MJ, Recht A, Lagios MD The natural history of ductal carcinoma in situ of the breast Ductal Carcinoma In Situ of the Breast 2002 Philadelphia, PA: Lippincott, Williams and Wilkins 17 21
Alowami S Troup S Al-Haddad S Kirkpatrick I Watson PH Mammographic density is related to stroma and stromal proteoglycan expression Breast Cancer Res 2003 5 R129 R135 12927043 10.1186/bcr622
Barcellos-Hoff MH Medina D New highlights on stroma-epithelial interactions in breast cancer Breast Cancer Res 2005 7 33 36 15642180 10.1186/bcr972
Fisher B Costantino JP Wickerham DL Redmond CK Kavanah M Cronin WM Vogel V Robidoux A Dimitrov N Atkins J Tamoxifen for the prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 study J Natl Cancer Inst 1998 90 1371 1388 9747868 10.1093/jnci/90.18.1371
Brisson J Brisson B Cote G Maunsell E Berube S Robert J Tamoxifen and mammographic breast densities Cancer Epidemiol Biomarkers Prev 2000 9 911 915 11008908
Cuzick J Warwick J Pinney E Warren RM Duffy SW Tamoxifen and breast density in women at increased risk of breast cancer J Natl Cancer Inst 2004 96 621 628 15100340
Writing Group for the Women's Health Initiative Investigators Risks and benefits of estrogen plus progestin in healthy postmenopausal women: Principal results from the Womens Health Initiative randomized controlled trial JAMA 2002 288 321 333 12117397 10.1001/jama.288.3.321
Greendale GA Reboussin BA Slone S Wasilauskas C Pike MC Ursin G Postmenopausal hormone therapy and change in mammographic density J Natl Cancer Inst 2003 95 30 37 12509398
|
16168104
|
PMC1242126
|
CC BY
|
2021-01-04 16:54:51
|
no
|
Breast Cancer Res. 2005 Jun 8; 7(5):R605-R608
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1260
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12611616811110.1186/bcr1261Research ArticleModulation of monocyte matrix metalloproteinase-2 by breast adenocarcinoma cells Szabo Kristina A [email protected] Gurmit [email protected] Juravinski Cancer Centre, Hamilton, Ontario, Canada2 Department of Pathology and Molecular Medicine, McMaster University, Hamilton, Ontario, Canada2005 14 6 2005 7 5 R661 R668 31 1 2005 3 3 2005 24 3 2005 4 4 2005 Copyright © 2005 Szabo et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
The presence of monocyte and macrophage cells in growing breast tumors, and the positive relationship between the degree of immune cell infiltration and tumor growth, suggest a possible paracrine growth regulatory function of immune cells in breast cancer.
Method
To better understand the interaction between monocytes and breast cancer cells, in vitro matrix metalloproteinase and tissue inhibitor of metalloproteinase activity was assessed from the THP-1 myeloid cell line in response to conditioned media from two breast cancer cell lines, MCF-7 and MDA-MB-231.
Results
Enzymography and immunoblotting revealed increased MMP-2 as well as increased levels of TIMP-1 and TIMP-2. Furthermore, a significant increase in the invasive potential of MCF-7 and MDA-MB-231 cells was noted in response to THP-1 cell-conditioned media.
Conclusion
These data demonstrate that monocyte cells in the breast tumor microenvironment play an important role in the modulation of MMPs, which may have a significant effect on the control of tumor growth and metastatic spread.
==== Body
Introduction
There is a growing body of evidence to suggest that the tumor microenvironment is immunosuppressive [1,2]. This is perhaps as a result of selection for such an environment, which is a process recently termed immunoediting [3]. The growth of solid tumors has been likened to an aberration of the normal process of wound healing and, in consequence, immune cells may inadvertently aid tumor growth. The tumor microenvironment often contains a number of migratory haematopoietic cells that play pivotal roles in the progression and metastasis of tumors [4-8]. Monocyte and macrophage cells are prominent in the inflammatory infiltrate of tumors, often in considerable numbers [9-12], and it has been suggested that the presence of these cells may independently influence the metastatic potential of certain tumors [13].
The clinical significance of the mononuclear infiltrate that is often seen in breast cancer has remained the subject of continuous debate. In invasive breast carcinoma, the neoplastic cell population is often outnumbered by stromal cells such as tumor-associated macrophages (TAMs), which can comprise as much as 80% of the entire tumor-associated leukocyte (TAL) population [14], and more than 50% of the total tumor mass [15]. It appears that TAMs are actually required for the tumor to survive [16-19]. Other studies have also reported an overwhelming predominance of TAMs within the TAL population of both primary [12,20,21] and metastatic [22] breast carcinoma. Moreover, positive relationships between the presence of macrophages and lymph node metastasis [23], c-erbB2 [24], and increased expression of urokinase plasminogen activator (uPA) [25] in breast cancer have been reported.
The majority of studies that have attempted to correlate macrophage content with tumor severity have noted higher numbers of macrophages in conjunction with increasing tumor malignancy potential. Among the stromal cells, the presence of macrophage cells is frequently noted in aggressive malignant tumors, indicating a relation between macrophages and the degree of tumor cell differentiation [26-28]. These data are particularly compelling for breast, prostate, ovarian and cervical cancers. There is clinical data correlating a poor prognosis with the extent of macrophage infiltrate in breast cancer patients [26] as well as the differential cytotoxicity of macrophages from regressing and progressing tumors [29]. Monocytes represent precursor cells that serve as a source for the constant renewal of tissue macrophages on demand as well as in steady-state conditions. It is thought that monocytes in the peripheral circulation are recruited to the tumor site [30] by the release of chemotactic cytokines, and once recruited, the monocytes differentiate to become TAMs.
Matrix metalloproteinases (MMPs) comprise a family of zinc-containing endopeptidases that share structural domains and have the capacity to degrade extracellular matrix (ECM) components as well as to alter biological functions of ECM macromolecules [31]. The specific proteolytic targets of MMPs may include many other proteases, protease inhibitors, clotting factors, chemotactic molecules, latent growth factors, growth factor binding proteins, cell surface receptors, as well as cell-cell and cell-matrix adhesion molecules. Under physiological conditions, the activity of MMPs is tightly regulated to prevent excessive proteolytic activity and tissue destruction. Important sources of MMPs are immune cells, which utilize these enzymes to mediate extravasation into tissues during inflammation. Although initially it was assumed that the tumor cell was the origin of MMPs found in this environment, in situ hybridization techniques have shown that while some MMPs are expressed by tumor cells, MMPs are predominantly produced by adjacent host stromal and inflammatory cells in response to factors released by tumors [32,33]. The tumor cell MMPs may contribute to the invasive growth of the tumor while the stromal elements contribute to the remodelling process and the desmoplastic reaction that occurs in the tissue adjacent to the tumor [34].
MMPs produced by monocyte and macrophage cells, which have been implicated in tumor cell invasion and metastasis, may be enhanced by soluble factors from breast cancer cells. While tumor cells have been extensively studied, the roles of monocytes in the tumor microenvironment have not been well characterized. In an attempt to further understand the roles of monocytes in the tumor microenvironment, in vitro MMP and tissue inhibitor of metalloproteinase (TIMP) protein levels and enzymatic activity from THP-1 monocyte cells were assessed in response to the breast cancer cell lines MCF-7 and MDA-MB-231. An assessment of the invasive potential of these breast cancer cell lines in response to the monocyte cells was also conducted. Our results indicate that monocytes act as important regulators of ECM breakdown during tumor invasion and metastasis as a result of their ability to regulate MMP and TIMP production following their migration to the tumor site. Therefore, the basic mechanisms that regulate monocyte recruitment from the circulation into a tumor site and their production of MMPs and TIMPs are likely to be significant to breast cancer research.
Materials and methods
Cell lines and cell culture
The monocyte cell line THP-1, the estrogen receptor positive human breast cancer cell line MCF-7 and the estrogen receptor negative breast cancer cell line MDA-MB-231 were obtained from the American Type Culture Collection (Manassas, VA, USA). All cell lines were cultured under standard tissue culture conditions and tested negative for mycoplasma contamination. These cell lines were maintained in RPMI 1640 medium (Gibco, Grand Island, NY, USA), supplemented with 10% w/v fetal bovine serum (FBS) (Gibco), 1 mM sodium pyruvate, penicillin (100 units/ml) and streptomycin (100 units/ml) (Gibco) in a 5% CO2 incubator at 37°C. When experimental conditions called for the use of phenol red-free and serum-free media, the same RPMI medium was used as above but without the phenol red and FBS. The cells were only used a maximum of 15 passages and cells were grown to 80% confluence prior to experimentation. All media and reagents contained less than 0.06 endotoxin units/ml, confirmed by testing in our laboratory using the Limulus Amebocyte Lysate gel clot assay (Cambrex, East Rutherford, NJ, USA).
Collection of conditioned media
Cultured cells were washed three times with phenol red-free, serum-free RPMI and incubated in this media for 48 h. Following incubation, the cells were harvested, the recovered cell number determined, and cell viability was assessed at the end of each culture period by Trypan Blue exclusion. Samples of tumor cell-derived conditioned media (CM) were collected and centrifuged to remove cell debris. Harvested CM was concentrated using the Amicon Ultra-4 centrifugal filter units with a nominal molecular weight limit of 10 kDa (Millipore, Bedford, MA, USA). The CM was immediately frozen at -80°C until enzymography or immunoblotting was performed.
Treatment of THP-1 cells with conditioned media
For CM studies, THP-1 cells were grown to subconfluency (80% to 85%) in serum-supplemented media. The cells were then pelleted and washed three times in phenol red-free and serum-free RPMI before resuspension in this same media. Prior to experimentation, the concentration of the THP-1 cells was adjusted to 5 × 105 cells/ml by suspension in RPMI and the cells were seeded into six-well, flat bottom plates. The total volume of each well was 2.5 ml, with cells exposed to increasing volumes of MCF-7 and MDA-MB-231 breast cancer cell CM or incubated with control volumes of concentrated serum-free media for 48 h. After the 48 h incubation time, cells and debris were removed from the THP-1 supernatant by centrifugation. Following incubation, the recovered cell number was determined and cell viability established using Trypan Blue exclusion. Under the conditions in this study, the treatment of THP-1 cells with CM did not reduce the viability of the THP-1 cells.
Western blotting
The CM protein levels were quantified using a protein assay (Bio-Rad, Hercules, CA, USA) and the results were compared with a standard curve of BSA concentrations. The protein concentrations were used to normalize the amount of CM loaded onto the gel; thus, the total protein loaded was equivalent in each lane. Loading buffer (5% w/v SDS, 0.225 M Tris-Cl pH 6.8, 50% w/v glycerol, 0.05% w/v bromophenol blue, 0.25 M dithiothreitol) was added to the CM and the mixture was denatured by boiling for 5 minutes. Samples were loaded onto a 10% SDS polyacrylamide gel with 10 μl of a wide-range colored protein molecular weight marker (Invitrogen, Carlsbad, CA, USA) also loaded onto the gel. The proteins were subsequently transferred from the gel onto a nitrocellulose membrane (Amersham Biosciences, Piscataway, NJ, USA). The membrane was incubated with mouse monoclonal anti-MMP-2 antibodies diluted 1:100 (Oncogene, Cambridge, MA, USA), and mouse monoclonal anti-TIMP-2, anti-TIMP-2 and anti-uPA antibodies diluted 1:400 (Oncogene). The membrane was subsequently incubated with mouse anti-rabbit IgG antibodies conjugated to horseradish peroxidase (HRP) or goat anti-mouse IgG antibodies HRP diluted 1:5000 (Santa Cruz Biotechnology, Santa Cruz, CA, USA). ECL chemiluminescent detection (Amersham Biosciences) was used to visualize the proteins.
Enzymography
Aliquots of CM were mixed with sample buffer (0.5 M Tris-HCl pH 6.8, 10% w/v glycerol, 10% w/v SDS, and 0.1% w/v bromophenol blue) and were subjected to electrophoresis on a 10% SDS polyacrylamide gel containing 1 mg/ml gelatin (Sigma, St. Louis, MO, USA) or 1 mg/ml casein (Sigma). After electrophoresis, the gel was washed in 2.7% (v/v) Triton X-100 for 1 h at 37°C. The gels were then incubated in a developing buffer (50 mM Tris Base, 40 mM 6N HCl, 200 mM NaCl, 5 mM CaCl2·H2O, 0.02% v/v Brij 35) for 15 minutes at room temperature followed by an overnight incubation on a shaker at 37°C in the same buffer to allow digestion of the substrate. After digestion, the gels were rinsed briefly with deionized water and stained with 0.5% (w/v) Coomassie Brilliant Blue R-250 in 40% (v/v) ethanol and 10% (v/v) acetic acid for 1 h, followed by destaining in a mixture of 30% (v/v) ethanol and 10% (v/v) acetic acid. MMP proteolytic bands were identified by examining unstained regions on the substrate-stained background; intact protein substrate stained blue, leaving enzyme degraded bands transparent.
Reverse enzymography
Protein standardized samples of CM were resolved in 10% SDS-PAGE containing 1 mg/ml gelatin (Sigma) and 1.5 μg/ml MMP-9 (Sigma). After gel electrophoresis, the gels were treated in the same manner as described for enzymography. For the reverse zymogram, the stained areas of the gel indicate enzymatic activity corresponding to the TIMP at that molecular weight.
Motility and invasion assays
Cell motility was assessed using 24-well Matrigel Insert Chambers (Becton Dickinson Labware, Franklin Lakes, NJ, USA) with polycarbonate filters containing 8 μm pores coated with growth factor reduced matrigel matrix (50 μg/filter). MCF-7 and MDA-MB-231 cells were seeded at 5 × 104 cells/well in the upper compartment of each invasion chamber. The lower chambers contained serum-free media, serum-free media containing 10% v/v FBS, as well as THP-1 CM, 5 × 104 and 10 × 104 MCF-7, MDA-MB-231 and THP-1 cells per chamber. After 48 h, the top surface of the membrane was gently scrubbed with a cotton bud and cells on the undersurface were fixed and stained with the DiffQuick staining kit (Baxter, McGaw Park, Il, USA). The number of cells that had migrated to the undersurface of the filters was counted in five separate high-powered fields for each membrane. Values from the separate experiments containing all conditions were averaged and plotted on a bar graph ± their corresponding standard error.
Statistical analysis of the data
All experiments were repeated a minimum of three times and representative results are shown. Student's t-test was used to compare results from the treated cells with the results from untreated control cells. Data are expressed as the mean ± standard error of the mean. Results were considered statistically significant at p < 0.05.
Results
Induction of MMP activity from THP-1 monocytes
MCF-7 (Fig. 1a) and MDA-MB-231 (Fig. 2a) CM elicited enhanced MMP-2 activity by THP-1 monocytes, as measured by gelatin zymography, when compared to either unstimulated controls or to the levels of the enzyme in the CM alone. Distinct gelatinolytic bands were observed at calculated Mr 72,000, and 92,000, corresponding to the latent 72 kDa type IV collagenase (MMP-2), and the latent form of 92 kDa type IV collagenase (MMP-9), respectively. At the highest concentration of CM (300 μl) the enzymatic effect was the greatest. This coincided with an increase in expression of MMP-2 (Figs 1b and 2b) protein levels produced by the monocyte cells as determined by immunoblot analysis. In addition, the presence of casein-degrading MMPs was analysed by enzymography, although there were no appreciable levels of MMP-1 or MMP-7 produced. Western blot analysis confirmed this absence of MMP-7 and, furthermore, the basal level MMP-9 and uPA produced by the monocyte cells was not changed upon exposure to the breast cancer CM (results not shown).
Induction of TIMP activity from THP-1 monocytes
In addition to the contribution by MMPs, the degree of connective tissue destruction is also influenced by TIMPs. Reverse enzymography using MMP-9 as the degrading source revealed TIMP-1 and TIMP-2 activity (Figs 1c and 2c) at Mr 28,000 and 21,000, respectively. There was no difference in the level of monocyte TIMP-1 or TIMP-2 inhibition between the control and MCF-7 CM treated cells. Upon exposure to the MDA-MB-231 CM, however, there was an increased level of TIMP-2 inhibition (Fig. 2c). In order to compare the inhibitory activity of TIMP-1 and TIMP-2 with the protein levels, immunoblot analysis was performed on the samples at the 300 μl optimal concentration. The THP-1 monocytes produced enhanced TIMP-1 (Fig. 3a,b) and TIMP-2 (Fig. 3c,d) protein levels; thus a small increase in TIMP-1 expression as well as a more substantial increase in TIMP-2 was observed.
Motility and invasive potential of MCF-7 and MDA-MB-231 cells in response to monocytes
The previous results, and our interest in the ability of monocytes to degrade the ECM, raised the question of whether or not monocytes increase the invasive potential of breast cancer cells. The presence of THP-1 CM led to a significant increase (p < 0.05) in MCF-7 (Fig. 4a) cell invasion in comparison to the serum-free media control. The presence of 5 × 104 THP-1 cells in the lower chamber also led to a significant increase (p < 0.05) in the number of MCF-7 (Fig. 4a) and MDA-MB-231 (Fig. 4b) cells that invaded through the matrigel membrane. The invasive potential of the breast cancer cells was maintained, however, as the concentration of the monocyte cells was doubled.
Discussion
The expression of components of the matrix degrading protease system by tumor stromal cells indicates that the tumor stroma does not merely play a passive role in cancer progression. Rather, it may in fact actively participate in the process of cancer invasion. We propose that it is the mixed population of cancer cells and recruited stromal cells that produce the matrix degrading protease components in order to facilitate the destruction of the surrounding normal tissue, which allows for malignant cell invasion. Our present data may partially explain why MMPs are often predominantly expressed by the stromal cells that surround invasive neoplastic cells. Specifically, monocytes may play an important role in tumor invasion and metastasis through their MMP proteolytic activity.
MMPs are usually expressed and secreted by cells as inactive enzymes, and further proteolytic processing is necessary to convert them into their active forms. The degradation of the ECM, especially basement membrane type IV collagen, is considered a key event for tumor cell invasion and metastasis. Specifically in colorectal cancer, MMP-9 is derived principally from stromal monocytes [35,36] and a high level of MMP-9 in tumor versus paired normal mucosa is an independent predictor of poor prognosis [37]. Moreover, there is a connection between the expression of TIMP-2 and MMP-2 under many physiological and pathological conditions, suggesting that TIMP-2 regulates MMP-2 activity [38,39]. Furthermore, TIMP-1 and TIMP-2 over-expression has been associated with malignant breast tumor behaviour in vivo [40,41]. Immunohistochemical studies have indicated that MMP-2 is highly expressed in more invasive and metastatic cancer tissues [42]. The fact that MMP-2 is often associated with adjacent normal tissues rather than the tumor cells themselves suggests that neoplastic cells can use MMPs produced by normal cells to facilitate their egress from the tumor mass and potentially their entry into new sites [43,44]. Thus, the breast cancer microenvironment may affect the monocyte MMP/TIMP balance and, consequently, play a role in the ECM breakdown.
Considering the high concentration of infiltrating immune cells in the breast cancer microenvironment, we examined whether breast cancer cells modulate the reactivity of monocyte cells by altering their production of MMPs and TIMPs. The data presented here suggest that monocyte-derived MMP, notably MMP-2, may play an important role in invasive processes. Both the protein levels and enzymatic activity of MMP-2 were elevated in response to concentrated MCF-7 and MDA-MB-231 CM. TIMP-1 and TIMP-2 protein levels were both increased in the CM treated cells. A study examining the effect of MCF-7 cells on human dermal fibroblasts found a similar observation, in that the breast cancer cells augmented the production of proMMP-1, -2, and -3 as well as TIMP-1 by fibroblast cells [45]. In addition, others have shown that contact between MDA-MB-231 cells and bone marrow derived fibroblasts resulted in an increase in the concentration of MMP-2 in the culture supernatant [46]. These findings are consistent with the observation that fibroblast cells promote tumor progression in animal models through their production of MMPs [47-49] and the findings emphasize the importance of tumor-host interactions during cancer progression. A recent study by Blot and colleagues [50] provides additional evidence on the role of monocytes in breast cancer. The authors cultured peripheral monocytes with MDA-MB-231 breast cancer cells and noticed an increase in MMP-9 expression [50]. Thus, the ability of MCF-7 and MDA-MB-231 cells to augment the production of MMPs in surrounding normal cells is likely one of the important properties for cell invasion and metastasis by these two breast cancer cell lines. However, since the THP-1 monocyte cell line differs in many respects from TAMs, further study is needed to assess the in vivo activities of monocyte cells and to determine if MMP-2 is upregulated by monocytes in the breast cancer microenvironment in vivo.
Collectively, these findings shed new light on the role of tumor-associated monocytes in the regulation of breast tumor development, invasion and metastasis. Increased MMP production has been shown to break down the basement membrane around pre-invasive tumors, thereby enhancing the ability of tumor cells to escape into the surrounding stroma. Thus, it is becoming apparent that stromal cells in the tumor microenvironment play an important role in allowing the tumor to express its full neoplastic phenotype. Consequently, further studies could reveal many new pathophysiological implications of MMPs in their regulation of immune cells, cytokine and chemokine networks, and matrix proteolysis during the metastatic process.
Conclusion
This study supports the hypothesis that tumor cell-host stromal cell interactions play a critical role in the proteolytic cascade required for tumor progression. Here we have focussed on one type of solid tumor, carcinoma of the breast. Monocyte MMP-2 enzymatic activity and MMP-2 as well as TIMP-1 and TIMP-2 protein levels were increased in response to soluble factors from MCF-7 and MDA-MB-231 breast cancer cells. Furthermore, the breast cancer cells displayed significantly enhanced in vitro invasion in response to monocyte CM. These data provide evidence for a potentially important function of monocytes in the modulation of MMPs and degradation of connective tissue in neoplastic disease.
Abbreviations
BSA = bovine serum albumin; CM = conditioned media; ECM = extracellular matrix; FBS = fetal bovine serum; HRP = horse radish peroxidase; MMP = matrix metalloproteinase; TAL = tumor-associated leukocyte; TAM = tumor-associated macrophage; TIMP = tissue-inhibitor of matrix metalloproteinase; uPA = urokinase plasminogen activator.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
KS conceived and designed the study, performed analysis, interpreted the data and drafted the article. GS coordinated the study and contributed to the design of the study, also taking a role in supervising and final approval of the manuscript.
Acknowledgements
This work was supported by an operating grant from the Canadian Breast Cancer Research Alliance (CBCRA) to Gurmit Singh.
Figures and Tables
Figure 1 The effects of MCF-7 breast adenocarcinoma CM on MMP and TIMP activity from THP-1 monocytes. (a) Gelatin zymogram and (b) MMP-2 immunoblot showing the upregulation of monocyte MMP-2 after exposure to MCF-7 CM, and (c) a reverse zymogram to show that the activity of monocyte TIMP-1 and -2 remained the same upon treatment. Lanes 1, 4, and 7 contain CM from THP-1 cells in serum-free media grown in the presence of 100 μl, 200 μl and 300 μl concentrated MCF-7 CM, respectively. Lanes 2, 5, and 8 contain CM from THP-1 cells in serum-free media grown in the presence of 100 μl, 200 μl and 300 μl concentrated serum-free media, respectively. Lanes 3, 6, and 9 contain serum-free media with 100 μl, 200 μl and 300 μl, respectively, of MCF-7 CM in the absence of THP-1 cells. These figures are representative of three independent experiments carried out in duplicate, each of which demonstrates similar results.
Figure 2 The effects of MDA-MB-231 breast adenocarcinoma CM on MMP and TIMP activity from THP-1 monocytes. (a) Gelatin zymogram and (b) MMP-2 immunoblot showing the upregulation of monocyte MMP-2 after exposure to MDA-MB-231 CM, and (c) a reverse zymogram to show that the activity of monocyte TIMP-2 was also upregulated upon exposure to MDA-MB-231 CM, although TIMP-1 activity remained the same upon treatment. Lanes 1, 4, and 7 contain CM from THP-1 cells in serum-free media grown in the presence of 100 μl, 200 μl and 300 μl concentrated MDA-MB-231 CM, respectively. Lanes 2, 5, and 8 contain CM from THP-1 cells in serum-free media grown in the presence of 100 μl, 200 μl and 300 μl concentrated serum-free media, respectively. Lanes 3, 6, and 9 contain serum-free media with 100 μl, 200 μl and 300 μl, respectively, of MDA-MB-231 CM in the absence of THP-1 cells. These figures are representative of three independent experiments carried out in duplicate, each of which demonstrates similar results.
Figure 3 MCF-7 and MDA-MB-231 upregulation of TIMP-1 and TIMP-2 protein levels from THP-1 monocytes. Immunoblots representative of three independent experiments carried out in duplicate using monoclonal (a,b) TIMP-1 and (c,d) TIMP-2 antibodies to show an increased production of both proteins by monocyte cells in response to MCF-7 and MDA-MB-231 CM. Lane 1 contains CM from THP-1 cells in serum-free media grown in the presence of 300 μl concentrated (a,c) MCF-7 CM and (b,d) MDA-MB-231 CM. Lane 2 contains CM from THP-1 cells in serum-free media grown in the presence of 300 μl concentrated serum-free media. Lane 3 contains serum-free media with 300 μl of (a,c) MCF-7 CM and (b,d) MDA-MB-231 CM in the absence of THP-1 cells. Lane 4 contains unconcentrated (a) MCF-7 and (b,d) MDA-MB-231 CM.
Figure 4 The effects of monocyte cells and CM on the invasive potential of breast adenocarcinoma cells. Graphs showing an increase in the number of (a) MCF-7 and (b) MDA-MB-231 cells that migrated through matrigel-coated membranes in response to monocyte cell CM. The following conditions were used in the lower chamber: 1, serum-free media; 2, 10% FBS in serum-free media; 3, THP-1 cell-conditioned media; 4, (a) 5 × 104 MCF-7 cells, (b) 5 × 104 MDA-MB-231 cells; 5, 5 × 104 THP-1 cells; 6, (a) 10 × 104 MCF-7 cells, (b) 10 × 104 MDA-MB-231 cells; 7, 10 × 104 THP-1 cells. *p < 0.05 compared to condition 2; **p < 0.05 compared to condition 4; ***p < 0.05 compared to condition 6. The values are the means of three experiments and the error bars are the standard error of the mean.
==== Refs
Elgert KD Alleva DG Mullins DW Tumor-induced immune dysfunction: the macrophage connection J Leukoc Biol 1998 64 275 290 9738653
Ohm JE Carbone DP VEGF as a mediator of tumor-associated immunodeficiency Immunol Res 2001 23 263 272 11444391 10.1385/IR:23:2-3:263
Dunn GP Bruce AT Ikeda H Old LJ Schreiber RD Cancer immunoediting: from immunosurveillance to tumor escape Nat Immunol 2002 3 991 998 12407406 10.1038/ni1102-991
Balkwill F Mantovani A Inflammation and cancer: back to Virchow? Lancet 2001 357 539 545 11229684 10.1016/S0140-6736(00)04046-0
Brigati C Noonan DM Albini A Benelli R Tumors and inflammatory infiltrates: friends or foes? Clin Exp Metastasis 2002 19 247 258 12067205 10.1023/A:1015587423262
Coussens LM Werb Z Inflammation and cancer Nature 2002 420 860 867 12490959 10.1038/nature01322
Leek RD Harris AL Tumor-associated macrophages in breast cancer J Mammary Gland Biol Neoplasia 2002 7 177 189 12463738 10.1023/A:1020304003704
Lin EY Gouon-Evans V Nguyen AV Pollard JW The macrophage growth factor CSF-1 in mammary gland development and tumor progression J Mammary Gland Biol Neoplasia 2002 7 147 162 12465600 10.1023/A:1020399802795
Lwin KY Zuccarini O Sloane JP Beverley PC An immunohistochemical study of leukocyte localization in benign and malignant breast tissue Int J Cancer 1985 36 433 438 3930410
Steele RJ Brown M Eremin O Characterization of macrophages infiltrating human mammary carcinomas Br J Cancer 1985 51 135 138 3155620
McBride WH Phenotype and functions of intratumoral macrophages Biochim Biophys Acta 1986 865 27 41 3524684
Kelly PM Davison RS Bliss E McGee JO Macrophages in human breast disease: a quantitative immunohistochemical study Br J Cancer 1988 57 174 177 2833921
Lauder I Aherne W Stewart J Sainsbury R Macrophage infiltration of breast tumors: a prospective study J Clin Pathol 1977 30 563 568 874114
Gottlinger HG Rieber P Gokel JM Lohe KJ Riethmuller G Infiltrating mononuclear cells in human breast carcinoma: predominance of T4+ monocytic cells in the tumor stroma Int J Cancer 1985 35 199 205 3918945
O'Sullivan C Lewis CE Tumor-associated leukocytes: friends or foes in breast carcinoma J Pathol 1994 172 229 235 8195926
Evans R Macrophage requirement for growth of a murine fibrosarcoma Br J Cancer 1978 37 1086 1089 307956
Sunderkotter C Steinbrink K Goebeler M Bhardwaj R Sorg C Macrophages and angiogenesis J Leukoc Biol 1994 55 410 422 7509844
Fauve RM Some aspects of the macrophage-cancer relationship Res Immunol 1993 144 265 268 8378592 10.1016/0923-2494(93)80103-6
Mantovani A Bottazzi B Colotta F Sozzani A Ruco L The origin and function of tumor-associated macropahges Immunol Today 1992 13 265 270 1388654 10.1016/0167-5699(92)90008-U
Van Netten JP Ashmead BJ Parker RL Thornton IG Fletcher C Cavers D Coy P Brigden ML Macrophage-tumor cell associations: a factor in metastasis of breast cancer? J Leukoc Biol 1993 54 360 362 8409759
Van Ravenswaay Claasen HH Kluin PM Fleuren GJ Tumor infiltrating cells in human cancer. On the possible role of CD16+ macrophages in antitumor cytotoxicity Lab Invest 1992 67 166 174 1501443
Horst HA Horny HP Characterization and frequency distribution of lymphoreticular infiltrates in axillary lymph node metastases of invasive ductal carcinoma of the breast Cancer 1987 60 3001 3007 2824022
Zuk JA Walker RA Immunohistochemical analysis of HLA antigens and mononuclear infiltrates of benign and malignant breast J Pathol 1987 152 275 285 3312546
Pupa SM Bufalino R Invernizzi AM Andreaola S Rilke F Lombardi L Colnaghi MI Menard S Macrophage infiltrate and prognosis in c-erbB-2-overexpressing breast carcinomas J Clin Oncol 1996 14 85 94 8558226
Hildenbrand R Dilger I Horlin A Stutte HJ Urokinase and macrophages in tumor angiogenesis Br J Cancer 1995 72 818 823 7547226
Leek RD Lewis CE Whitehouse R Greenall M Clarke J Harris AL Association of macrophage infiltration with angiogenesis and prognosis in invasive breast carcinoma Cancer Res 1996 56 4625 4629 8840975
Ring P Johansson K Hoyhtya M Rubin K Lindmark G Expression of tissue inhibitor of metalloproteinases TIMP-2 in human colorectal cancer – a predictor of tumor stage Br J Cancer 1997 76 805 811 9310250
Jaalinoja J Herva R Korpela M Hoyhtya M Turpeenniemi-Hujanen T Matrix metalloproteinase 2 (MMP-2) immunoreactive protein is associated with poor grade and survival in brain neoplasms J Neurooncol 2000 46 81 90 10896208 10.1023/A:1006421112839
Russell SW McIntosh AT Macrophages isolated from regressing Moloney sarcomas are more cytotoxic than those recovered from progressing sarcomas Nature 1977 268 69 71 887150
Szabo KA Singh G Matrix metalloproteinases and the immune response Clin Appl Immunol Rev 2004 4 295 319 10.1016/j.cair.2004.02.001
Nagase H Woessner JF Jr Matrix metalloproteinases J Biol Chem 1999 274 21491 21494 10419448 10.1074/jbc.274.31.21491
Nelson AR Fingleton B Rothenberg ML Matrisian LM Matrix metalloproteinases: biologic activity and clinical implications J Clin Oncol 2000 18 1135 1149 10694567
Hoekstra R Eskens FA Verweij E Matrix metalloproteinase inhibitors: current developments and future perspectives Oncologist 2001 6 415 427 11675519 10.1634/theoncologist.6-5-415
John A Tuszynski G The role of matrix metalloproteinases in tumor angiogenesis and tumor metastasis Pathol Oncol Res 2001 7 14 23 11349215
Zeng ZS Guillem JG Distinct pattern of matrix metalloproteinase 9 and tissue inhibitor of metalloproteinase 1 mRNA expression in human colorectal cancer and liver metastases Br J Cancer 1995 72 575 582 7669564
Swallow CJ Murray MP Guillem JG Metastatic colorectal cancer cells induce matrix metalloproteinase release by human monocytes Clin Exp Metastasis 1996 14 3 11 8521614
Zeng ZS Huang Y Cohen AM Guillem JG Prediction of colorectal cancer relapse and survival via tissue RNA levels of matrix metalloproteinase-9 J Clin Oncol 1996 14 3133 3140 8955659
Strongin AY Collier I Bannikov G Marmer BL Grant GA Goldberg GI Mechanism of cell surface activation of 72-kDa type IV collagenase. Isolation of the activated form of the membrane metalloproteinase J Biol Chem 1995 270 5331 5338 7890645 10.1074/jbc.270.10.5331
Butler GS Butler MJ Atkinson SJ Will H Tamura T van Westrum SS Crabbe T Clements J d'Ortho MP Murphy G The TIMP2 membrane type 1 metalloproteinase "receptor" regulates the concentration and efficient activation of progelatinase A. A kinetic study J Biol Chem 1998 273 871 880 9422744 10.1074/jbc.273.2.871
McCarthy K Maguire T McGreal G McDermott E O'Higgins N Duffy MJ High levels of tissue inhibitor of metalloproteinase-1 predict poor outcome in patients with breast cancer Int J Cancer 1999 84 44 48 9988231 10.1002/(SICI)1097-0215(19990219)84:1<44::AID-IJC9>3.0.CO;2-P
Visscher DW Hoyhtya M Ottosen SK Liang CM Sarkar FH Crissman JD Fridman R Enhanced expression of tissue inhibitor of metalloproteinase-2 (TIMP-2) in the stroma of breast carcinomas correlates with tumor recurrence Int J Cancer 1994 59 339 344 7927938
Monteagudo C Merino MJ San-Juan J Liotta LA Stetler-Stevenson WG Immunohistochemical distribution of type IV collagenase in normal, benign, and malignant breast tissue Am J Pathol 1990 136 585 592 2156430
Heppner KJ Matrisian LM Jensen RA Rodgers WH Expression of most matrix metalloproteinase family members in breast cancer represents a tumor-induced host response Am J Pathol 1996 149 273 282 8686751
Boyd RS Balkwill FR MMP-2 release and activation in ovarian carcinoma: the role of fibroblasts Br J Cancer 1999 80 315 321 10408832 10.1038/sj.bjc.6690357
Ito A Nakajima S Sasaguri Y Nagase H Mori Y Co-culture of human breast adenocarcinoma MCF-7 cells and human dermal fibroblasts enhances the production of matrix metalloproteinases 1, 2, and 3 in fibroblasts Br J Cancer 1995 71 1039 1045 7734296
Saad S Bendall LJ James A Gottlieb DJ Bradstock KF Induction of matrix metalloproteinases MMP-1 and MMP-2 by co-culture of breast cancer cells and bone marrow fibroblasts Breast Cancer Res Treat 2000 63 105 115 11097087 10.1023/A:1006437530169
Noel A De Pauw-Gillet MC Purnell G Nusgens B Lapiere CM Foidart JM Enhancement of tumorigenicity of human breast adenocarcinoma cells in nude mice by matrigel and fibroblasts Br J Cancer 1993 68 909 915 8217606
Noel A Polette M Lewalle JM Munaut C Emonard HP Birembaut P Foidart JM Coordinate enhancement of gelatinase A mRNA and activity levels in human fibroblasts in response to breast-adenocarcinoma cells Int J Cancer 1994 56 331 336 8314319
Masson R Lefebvre O Noel A Fahime ME Chenard MP Wendling C Kebers F LeMeur M Dierich A Foidart JM In vivo evidence that the stromelysin-3 metalloproteinase contributes in a paracrine manner to epithelial cell malignancy J Cell Biol 1998 140 1535 1541 9508784 10.1083/jcb.140.6.1535
Blot E Chen W Vasse M Paysant J Denoyelle C Pille J-Y Vincent L Vannier J-P Soria J Soria C Cooperation between monocytes and breast cancer cells promotes factors involved in cancer aggressiveness Br J Cancer 2003 88 1207 1212 12698185 10.1038/sj.bjc.6600872
|
16168111
|
PMC1242127
|
CC BY
|
2021-01-04 16:04:33
|
no
|
Breast Cancer Res. 2005 Jun 14; 7(5):R661-R668
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1261
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12621616810510.1186/bcr1262Research ArticleSomatic mutation and gain of copy number of PIK3CA in human breast cancer Wu Guojun [email protected] Mingzhao [email protected] Elizabeth [email protected] Xin 3Liu Junwei 1Guo Zhongmin [email protected] Aditi [email protected] David 1Gollin Susanne M 3Sukumar Saraswati [email protected] Barry [email protected] David [email protected] Department of Otolaryngology-Head and Neck Surgery, Head and Neck Cancer Research Division, Johns Hopkins University School of Medicine, Baltimore, MD, USA2 Division of Endocrinology and Metabolism, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, USA3 Department of Human Genetics, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA, USA4 Breast cancer program, The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, Baltimore, MD, USA2005 31 5 2005 7 5 R609 R616 7 9 2004 27 10 2004 21 3 2005 4 5 2005 Copyright © 2005 Wu et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Phosphatidylinositol 3-kinases (PI3Ks) are a group of lipid kinases that regulate signaling pathways involved in cell proliferation, adhesion, survival, and motility. Even though PIK3CA amplification and somatic mutation have been reported previously in various kinds of human cancers, the genetic change in PIK3CA in human breast cancer has not been clearly identified.
Methods
Fifteen breast cancer cell lines and 92 primary breast tumors (33 with matched normal tissue) were used to check somatic mutation and gene copy number of PIK3CA. For the somatic mutation study, we specifically checked exons 1, 9, and 20, which have been reported to be hot spots in colon cancer. For the analysis of the gene copy number, we used quantitative real-time PCR and fluorescence in situ hybridization. We also treated several breast cancer cells with the PIK3CA inhibitor LY294002 and compared the apoptosis status in cells with and without PIK3CA mutation.
Results
We identified a 20.6% (19 of 92) and 33.3% (5 of 15) PIK3CA somatic mutation frequency in primary breast tumors and cell lines, respectively. We also found that 8.7% (8 of 92) of the tumors harbored a gain of PIK3CA gene copy number. Only four cases in this study contained both an increase in the gene copy number and a somatic mutation. In addition, mutation of PIK3CA correlated with the status of Akt phosphorylation in some breast cancer cells and inhibition of PIK3CA-induced increased apoptosis in breast cancer cells with PIK3CA mutation.
Conclusion
Somatic mutation rather than a gain of gene copy number of PIK3CA is the frequent genetic alteration that contributes to human breast cancer progression. The frequent and clustered mutations within PIK3CA make it an attractive molecular marker for early detection and a promising therapeutic target in breast cancer.
==== Body
Introduction
Phosphatidylinositol 3-kinases (PI3Ks) are a group of lipid kinases composed of 85-kDa and 110-kDa subunits. The 85-kDa subunit lacks PI3K activity and acts as adaptor, coupling the 110-kDa subunit (P110) to activated protein tyrosine kinases and generating second messengers by phosphorylating membrane inositol lipids at the D3 position. The resulting phosphatidylinositol derivatives then permit activation of downstream effectors that are involved in cell proliferation, survival, metabolism, cytoskeletal reorganization, and membrane trafficking [1,2].
PIK3CA, the gene encoding the 110-kDa subunit of PI3K, was mapped to 3q26, an area amplified in various human cancers including ovarian, head and neck, breast, urinary tract, and cervical cancers [3-5]. PIK3CA was specifically found to be amplified and overexpressed in ovarian and cervical cancer [6-9]. The increased copy number of the PIK3CA gene is associated with increased PIK3CA transcription, P110-alpha protein expression, and PI3K activity in ovarian cancer [9]. Treatment with a PI3K inhibitor decreased proliferation and increased apoptosis, suggesting that PIK3CA has an important role in ovarian cancer. More recently, PIK3CA mutations were identified in different human cancers. In that report, PIK3CA was mutated in 32%, 27%, 25%, and 4% of colon, brain, gastric, and lung cancers, respectively. Only 12 cases of breast cancer were examined, of which one was found to harbor a mutation in PIK3CA [10].
In an effort to identify the genetic alterations of the PIK3CA gene in breast cancer, we determined the mutation frequency and the change in the gene copy number of PIK3CA in a set of primary breast tumors and breast cancer cell lines. We found a high frequency of these somatic alterations of PIK3CA gene in a large number of primary breast cancers. In addition, mutation of the PIK3CA gene correlated with the activation of Akt. Inhibition of PIK3CA induced significant apoptosis in cells with PIK3CA mutation.
Materials and methods
Breast cancer cell line and tumors
Of the breast cancer cell lines examined, MCF12A, Hs.578t, and MDA436 were kindly provided by Dr Nancy Davidson at Johns Hopkins University, and MDA-MB157, MDA-MB468, BT474, T47D, and UACC893 were kindly provided by Dr Fergus J Couch at Mayo Clinic. The other cell lines were obtained from the American Type Culture Collection. A total of 92 cases of breast tumor, including 33 paired primary invasive breast carcinomas and adjacent normal tissues (frozen tissue), were obtained from the Surgical Pathology archives of the Johns Hopkins Hospital, Baltimore, MD, USA, in accordance with the Institutional Review Board protocol and DNA was isolated using a standard phenol–chloroform protocol. Prof Saraswati Sukumar at the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University provided isolated DNA. Each tumor used in this study was determined to contain greater than 70% tumor cells by H&E staining. Among these specimens, 3 were stage 1, 52 were stage 2, 22 were stage 3, and 4 were stage 4. Eleven were of uncharacteristic stage status. All of the tumors were high grade.
PCR, sequencing, and mutational analysis
Cell line and tumor DNA were isolated as standard protocol. The primers we used for PCR and sequencing were as follows. For exon 1: forward, CTCCACGACCATCATCAGG, reverse, GATTACGAAGGTATTGGTTTAGACAG, and sequencing primer, ACTTGATGCCCCCAAGAATC; for exon 9: forward, GATTGGTTCTTTCCTGTCTCTG, reverse, CCACAAATATCAATTTACAACCATTG, and sequencing primer, TTGCTTTTTCTGTAAATCATCTGTG; for exon 20: forward, TGGGGTAAAGGGAATCAAAAG, reverse, CCTATGCAATCGGTCTTTGC, and sequencing primer, TGACATTTGAGCAAAGACCTG. We used the same PCR conditions for all three exons. After incubation at 95°C for 5 min, two cycles of amplification were performed at the initial annealing temperature of 62°C, with a subsequent annealing temperature decrease of 2°C for every two cycles until 54°C. Twenty-five amplification cycles were then performed. After PCR reaction, samples were subjected to automated DNA sequencing using the ABI 377 Sequencer. The positive samples were confirmed by re-PCR and sequencing using the same primers and conditions.
Western blotting
To evaluate Akt phosphorylation status, MDA231, MD361, MCF7, BT20, BT474, and T47D cells were grown in appropriate medium and cell lysates were collected in SDS lysis buffer (cell signaling). Lysates were cleared of insoluble material by microcentrifugation at 15,800 g for 15 min at 4°C, and protein concentrations were determined (protein assay kit; Bio-Rad, Hercules, CA, USA). Approximately 50 μg of total protein from each sample was denatured in loading buffer for 10 min, electrophoresed through 10% polyacrylamide gels, and electroblotted to a nylon transfer membrane (Schleicher & Schuell, Bioscience, Keene, NH USA). The membrane was incubated overnight with primary antibody Akt ser473 (antirabbit, cell signaling), Akt (anti rabbit, Cell signaling) or β-actin (antimouse antibody; Sigma, St Louis, MO, USA) at 4°C. Then the membrane was washed three times in Tris-buffered saline with 0.1% Tween 20 at room temperature and incubated for 1 hour at room temperature with horseradish-peroxidase-labeled secondary antibody (goat antirabbit IgG; or goat antimouse IgG; Sigma). Signal detection was by horseradish peroxidase chemiluminescent reaction (ECL; Amersham).
Quantitative real-time PCR
For real-time PCR, specific primers and probes were designed using software from Applied Biosystems (Foster City, CA, USA) to amplify the PIK3CA and control β-actin (sequences are available on request). Using this combination and the protocol described by Mambo and colleagues [11], the samples were run in triplicate. Primers and probes to β-actin were run in parallel to standardize the input DNA (4 ng). Standard curves were developed using serial dilutions of DNA extracted from MCF12A. PCR amplifications were performed on an ABI 7900 TaqMan (Applied Biosystems) according to the manufacturer's protocol.
Fluorescence in situ hybridization (FISH)
Bacterial artificial chromosome (BAC) clone RP11-466H15 for PIK3CA was obtained from Research Genetics (Invitrogen Corporation, Carlsbad, CA, USA). BAC DNA isolation was carried out using the standard laboratory protocol for phenol–chloroform extraction. The chromosome 3 α-satellite plasmid and BAC DNA were labeled directly in SpectrumOrange-dUTP® and SpectrumGreen-dUTP® (Vysis, Downers Grove, IL, USA), respectively, using the Vysis nick translation kit (Vysis) in accordance with the manufacturer's instructions. Slides were fixed using methanol:acetic acid (3:1), followed by pretreatment with RNase, and dual-color FISH was performed as described previously [12]. Slides were counterstained with 4',6-diamidino-2-phenylindole (DAPI; Sigma), mounted with antifade (Vysis), and stored at -20°C. At least 100 nuclei were evaluated for each sample. Analysis was carried out using an Olympus (New Hyde Park, NY, USA) BHS fluorescence microscope, and images were captured using a CytoVision Ultra (Applied Imaging, Santa Clara, CA, USA).
Apoptosis detection
We assessed cellular apoptosis using an Annexin V-FITC (fluorescein isothiocyanate) apoptosis detection kit (BD Biosciences, San Jose, CA, USA). Cells were cultured in 100-mm dishes until 50% confluent, serum-starved overnight, and then treated with LY294002, 3 μM and 10 μM, for 72 hours. Both detached and adherent cells were then collected and labeled with Annexin V-FITC and Propidium Iodide. The apoptosis was evaluated using FACScan (Becton Dickinson ImmunoSystems, Mountain View, CA, USA) flow cytometer.
Results
PIK3CA is frequently mutated in breast cancer cell lines and primary tumors
A previous report suggested that more than 80% of the mutations of the PIK3CA gene occur in three small clusters, namely in the p85 (exon 1), helical (exon 9) and kinase (exon 20) domains [10]. Based on this information, we sequenced exon 1, 9, and 20 in 15 breast cancer cell lines, 92 primary tumors, and 33 normal tissues. A total of five mutations were identified only in the 15 breast cancer cell lines (33.3%). No mutations were detected in the normal epithelial cell line MCF12A. Three of the mutations were identified in exon 9 and two were found in exon 20. No mutation was identified in exon 1. The BT20 cell line contained two different mutations, C1616G in exon 9 and A3140G in exon 20 (Fig. 1), corresponding to P539R and H1047R amino acid change, respectively.
A total of 19 mutation cases (20.6%) were identified in 92 primary tumors. Six of the 19 mutations were identified in 33 tumor samples but not in their paired normal samples. This indicated that the identified mutations are somatic mutations. Thirteen of these 19 mutations were in exon 9, and 6 were in exon 20. No mutation was identified in exon 1 in any of the 92 tumors. As shown in Table 1, the E545K mutation in exon 9 and the H1047R mutation in exon 20 were the two most frequent mutations in both breast cancer cell lines and primary tumors.
Gain of copy number of PIK3CA gene in primary breast cancer cell lines and tumors
To determine the PIK3CA gene copy number, we performed real-time quantitative PCR on 12 breast cancer cell lines, 92 primary tumors, and 33 normal controls. Standard curves for PIK3CA and β-actin amplification were generated using serially diluted MCF12A DNA, and showed linearity over the range used. Fig. 2a shows the standard curve for PIK3CA amplification with a slope of -4.044, while Fig. 2b shows the standard curve for β-actin amplification with a slope of -3.919. We did not observe any deletion of β-actin in the tumor samples. Most samples showed no difference in β-actin amplification between the paired tumor and a normal samples. A representative figure of β-actin amplification in a paired tumor and normal sample is shown in Fig. 2c. To evaluate the gene copy number in all samples, we set the cutoff line at 4 copies. Among the 33 cases with paired tissue, 8 (24.2%) showed a much higher gene copy number than normal controls (Fig. 3a). Only one case showed more than 4 copies. In a total of 92 cases of primary tumors, 8 (8.7%) had more than 4 copies, with the highest number being 7.8 copies (Fig. 3b). In addition, PIK3CA gene copy number was also determined in 12 breast cancer cell lines, and the MCF7, T47D, and BT474 cell lines had more than 4 copies (Fig. 3c). We also confirmed the gene copy number results of these 12 cell lines with FISH analysis. Representative FISH images are shown in Fig. 3d. Thus, our data of gene copy analysis indicates that gene amplification/gain of copy number of PIK3CA gene is not a frequent genetic alteration in breast cancer.
Biological effect of PIK3CA mutations in breast cancer
To determine whether the mutation of PIK3CA correlated with the activation of Akt (a downstream gene of PIK3 that mediates carcinogenic events such as proliferation), we performed western blot analysis to check the phosphorylation of Akt in several breast cancer cell lines. As shown in Fig. 4a, Akt phosphorylation was strongest in BT20 cells (which harbor two PIK3CA mutations) and MCF7 cells (which harbor PIK3CA mutation and high PIK3CA gene copy numbers). We also observed weak phosphorylation of Akt in MDA361 (which has one mutation) and in BT474 and T47D (no observable mutation but with high PIK3CA gene copy numbers). We did not observe phosphorylation of Akt in MDA231 (Fig. 4a) or in MCF12A and MDA157 cells (data not shown) that had no observable mutations and no copy number gain of PIK3CA. These data indicate that PIK3CA mutations might increase kinase activity and in turn activate the PI3K/AKT pathway.
We further investigated the biological effects of PIK3CA mutation in breast cancer cell lines by treating breast cancer cells with or without PIK3CA mutation with the PIK3CA inhibitor LY294002. As shown in Table 1, MCF7 harbors one mutation at position E545K, and BT20 harbors two mutations, which are located at positions P539R and H1047R. These somatic mutations were recently shown to have oncogenic transforming activity [13]. As shown in Fig. 4b and 4c, the fractions of apoptotic cells at 72 hours after treatment with 3 μM and 10 μM LY294002 were increased in MCF7 and BT20 cells. In addition, 3 μM and 10 μM LY294002 did not induce further apoptosis in MDA157 cells (Fig. 4c) or MDA231 cells (data not shown), even though serum starvation alone can induce more than 50% apoptosis in MDA157 cells (Fig. 4c).
Discussion
This study describes two innovations. First, we show a 20.6% mutation rate of the PIK3CA gene in breast cancer, indicating that PIK3CA mutation is a frequent genetic alteration in breast cancer. The 8% mutation rate of PIK3CA in breast cancer, reported in a previous study, was underestimated [10], probably because of the smaller number of cases examined. Another possibility might be the grade status of the tumors used, as all of the tumors in our study were of high grade. It will be useful and interesting in the future to explore whether PIK3CA mutation is correlated with tumor grade status.
Second, CGH (comparative genomic hybridization) studies have shown that 3q26 is an amplified chromosome region in various cancers, including breast cancer [4,5]. Unfortunately, it was not previously possible to identify the PIK3CA gene amplification pattern, because of the low resolution of the methods used. In our study, we used quantitative real-time PCR, a very sensitive and far more accurate technique [14,15], to specifically quantitate the genomic copy number of PIK3CA not only in primary breast tumors but also in paired tissues. Our data showed that gene amplification or gain of PIK3CA copy number is not a frequent genetic alteration event. This suggests that gene amplification is not the main molecular mechanism in activating the PIK3/AKT-driven tumorigenesis pathway in breast cancer.
Even though a complex and heterogeneous set of genetic alterations, including gene amplification/gain of copy number, deletion, and mutation, were reported to be involved in the etiology of breast cancer [16,17], our paper confirmed that gain of gene copy number and somatic mutation of one oncogene exist in parallel in breast cancer. Both amplification/gain of gene copy number and somatic mutation of PIK3CA have been shown to be associated with increased PI3K activity and might contribute to cancer through inhibition of apoptosis [6,9]. Gene amplification/gain of gene copy is well accepted as a later event in tumor progression [18,19], as is somatic mutation [10]. To determine the relation between somatic mutation and gain of gene copy number of PIK3CA gene in breast cancer, we integrated our mutation and gene copy number data. As shown in Table 2, 19 (20.6%) of 92 cases had a PIK3CA gene mutation and 4 cases did not harbor a mutation but showed a gain of gene copy number. Overall, a quarter (23 of 92) of all breast tumors examined had either a mutation or gain of copy number of the PIK3CA. In addition, 15 of 19 mutations were identified in tumors without gain of copy number of PIK3CA, suggesting that somatic mutations are a major contributory factor in the PIK3CA signaling pathway. Only four cases in the whole study had both a mutation and gain of copy number of PIK3CA. We did not observe a significant association between somatic mutation and gain of PIK3CA gene copy number in 92 cases of breast tumors (Table 2). We suggest that further studies using larger number of cases be undertaken in order to determine whether somatic mutation and gene amplification are independent genetic alterations in breast cancer.
Conclusion
The results from this study indicate that somatic mutation rather than gene amplification of PIK3CA is the main genetic alternation in breast cancer. The frequent and clustered mutations within PIK3CA make it an attractive molecular marker for early detection of breast cancer. In addition, the somatic mutations lead to activation of PIK3CA and also correlate with the activation of the PI3K/AKT pathway. Inhibition of PIK3CA can significantly induce apoptosis in cells with PIK3CA mutation. This suggests that PIK3CA might be a promising therapeutic target in breast cancer.
(During the writing of this manuscript, Bachman KE and colleagues published their results in Cancer Biology and Therapy [20]. They also reported more than 20% somatic mutations in breast cancer, a finding consistent with this study).
Abbreviations
BAC = bacterial artificial chromosome; DAPI = 4',6-diamidino-2-phenylindole; FISH = fluorescence in situ hybridization; H & E = hematoxylin and eosin; PI3K = phosphatidylinositol 3-kinase.
Authors' contributions
GW carried out PCR and sequencing reactions and prepared the manuscript. MX carried out primer design and PCR reactions. EM carried out real-time PCR reactions and prepared the manuscript. XH carried out FISH analysis. JW carried out apoptosis analysis. ZG carried out sequencing reactions. AC carried out sample preparation and real-time PCR. DG is the pathologist and carried out data analysis. SG coordinated FISH analysis. SS provided all breast cancer tissue and helped in designing the study. DS conceived of the study and participated in its design. BT coordinated all the studies. All authors read and approved the final manuscript.
Acknowledgements
We thank Dr Nancy Davidson at Johns Hopkins University and Dr Fergus J Couch at Mayo Clinic for providing some breast cancer cell lines. This work was supported by the following grants: National Cancer Institute's Lung Cancer SPORE Grant No. CA 58184-01 and the National Institute of Dental and Craniofacial Research Grant No. RO1-DE 012588-0.
Figures and Tables
Figure 1 Detection of somatic mutation of PIK3CA in breast cancer. In each case, the left sequence chromatogram was obtained from normal control and the right sequence chromatogram was obtained from tumor. Arrows indicate the location of missense mutations. The nucleotide and amino acid alterations are indicated on the left.
Figure 2 Typical real-time PCR curves generated for a cell line, a tumor, and normal tissue. Typical standard curves generated for (a) β-actin and (b) PIK3CA using serial diluted DNA from cell line MCF12A. (c) Representative real-time PCR curves for β-actin, generated using paired normal and tumor DNA from breast tissue of one individual. Each experiment was performed in triplicate and is shown by overlapping amplification curves. ΔRn = (Rn+) - (Rn-), where Rn+ is the fluorescence emission intensity of reporter/emission intensity of quencher at any time point, and Rn- is the initial emission intensity of reporter/emission intensity of quencher in the same reaction vessel before PCR amplification was initiated. Ct, cycle threshold.
Figure 3 Gain of PIK3CA gene copy number in breast cancer. (a)The change in the gene copy number of PIK3CA in 33 paired breast tissue samples. *Cases with significant change in gene copy number. The case number is indicated below the panel. (b) The change in the gene copy number of PIK3CA in 92 breast tumors and 33 normal controls. (c) Eight samples show more than 4 copies of PIK3CA. (d) The change in the gene copy number of PIK3CA in breast cancer cells using real-time PCR. Representative images of fluorescence in situ hybridization (FISH) analysis in various breast cancer cell lines. Green signals represent bacterial artificial chromosome (BAC) 466H15 probe. Red signals represent chromosome 3 centromere probe.
Figure 4 Biological effect of PIK3CA mutations in breast cancer. (a) Phosphorylation (P) of Akt in breast cancer cells. Western blotting showed stronger phosphorylation of Akt in BT20 and MCF7 cells than in MDA 231, MDA361, BT474, or T47D. β-Actin was used as a protein loading control. (b) Summary of the fraction of apoptotic cell in three cell lines with different treatments (3 μM and 10 μM LY294002). [(c) Apoptosis measurement in which cells were stained with Annexin V-FITC and for DNA content with Propidium Iodide and analyzed using flow cytometry. Apoptotic cells appear as a discrete population with elevated FITC fluorescence. FITC, fluorescein isothiocyanate; SS, serum starvation
Table 1 Somatic mutation of PIK3CA in breast cancer cell lines and primary tumors
Exon Nucleotide Amino acid Functional domain Cell lines Tumors Total
9 C1616G P539R Helical 1 (BT20) 1
9 G1624A E542K Helical 3 3
9 G1633A E545K Helical 2 (MDA361, MCF7) 8 10
9 A1634G E545G Helical 1 1
20 A3140G H1047R Kinase 2 (BT20, UACC893) 5 5
20 A3140T H1047L Kinase 1
No. with mutations 5 19 22
No. samples screened 15 92
Percent of cases with mutations 33.3 20.6
Table 2 The relation of somatic mutation and gain of copy number of PIK3CA in breast cancer
Mutation- Mutation+ Total
Gain+ 4 4 8
Gain- 69 15 84
Total 73 19 92
Fisher exact test, P = 0.054.
==== Refs
Leevers SJ Vanhaesebroeck B Waterfield MD Signalling through phosphoinositide 3-kinases: the lipids take centre stage Curr Opin Cell Biol 1999 11 219 225 10209156 10.1016/S0955-0674(99)80029-5
Vivanco I Sawyers CL The phosphatidylinositol 3-kinase AKT pathway in human cancer Nat Rev Cancer 2002 2 489 501 12094235 10.1038/nrc839
Knuutila S Bjorkqvist AM Autio K Tarkkanen M Wolf M Monni O Szymanska J Larramendy ML Tapper J Pere H DNA copy number amplifications in human neoplasms: review of comparative genomic hybridization studies Am J Pathol 1998 152 1107 1123 9588877
Rooney PH Murray GI Stevenson DA Haites NE Cassidy J McLeod HL Comparative genomic hybridization and chromosomal instability in solid tumours Br J Cancer 1999 80 862 873 10360667 10.1038/sj.bjc.6690433
Larramendy ML Lushnikova T Bjorkqvist AM Wistuba II Virmani AK Shivapurkar N Gazdar AF Knuutila S Comparative genomic hybridization reveals complex genetic changes in primary breast cancer tumors and their cell lines Cancer Genet Cytogenet 2000 119 132 138 10867149 10.1016/S0165-4608(99)00226-5
Ma YY Wei SJ Lin YC Lung JC Chang TC Whang-Peng J Liu JM Yang DM Yang WK Shen CY PIK3CA as an oncogene in cervical cancer Oncogene 2000 19 2739 2744 10851074 10.1038/sj.onc.1203597
Racz A Brass N Heckel D Pahl S Remberger K Meese E Expression analysis of genes at 3q26-q27 involved in frequent amplification in squamous cell lung carcinoma Eur J Cancer 1999 35 641 646 10492640 10.1016/S0959-8049(98)00419-5
Redon R Muller D Caulee K Wanherdrick K Abecassis J du Manoir S A simple specific pattern of chromosomal aberrations at early stages of head and neck squamous cell carcinomas: PIK3CA but not p63 gene as a likely target of 3q26-qter gains Cancer Res 2001 61 4122 4129 11358835
Shayesteh L Lu Y Kuo WL Baldocchi R Godfrey T Collins C Pinkel D Powell B Mills GB Gray JW PIK3CA is implicated as an oncogene in ovarian cancer Nat Genet 1999 21 99 102 9916799 10.1038/5042
Samuels Y Wang Z Bardelli A Silliman N Ptak J Szabo S Yan H Gazdar A Powell SM Riggins GJ High frequency of mutations of the PIK3CA gene in human cancers Science 2004 304 554 15016963 10.1126/science.1096502
Mambo E Gao X Cohen Y Guo Z Talalay P Sidransky D Electrophile and oxidant damage of mitochondrial DNA leading to rapid evolution of homoplasmic mutations Proc Natl Acad Sci USA 2003 100 1838 1843 12578990 10.1073/pnas.0437910100
Ried T Lengauer C Cremer T Wiegant J Raap AK van der Ploeg M Groitl P Lipp M Specific metaphase and interphase detection of the breakpoint region in 8q24 of Burkitt lymphoma cells by triple-color fluorescence in situ hybridization Genes Chromosomes Cancer 1992 4 69 74 1377011
Kang S Bader AG Vogt PK Phosphatidylinositol 3-kinase mutations identified in human cancer are oncogenic Proc Natl Acad Sci USA 2005 102 802 807 15647370 10.1073/pnas.0408864102
Yakes FM Van Houten B Mitochondrial DNA damage is more extensive and persists longer than nuclear DNA damage in human cells following oxidative stress Proc Natl Acad Sci USA 1997 94 514 519 9012815 10.1073/pnas.94.2.514
Ayala-Torres S Chen Y Svoboda T Rosenblatt J Van Houten B Analysis of gene-specific DNA damage and repair using quantitative polymerase chain reaction Methods 2000 22 135 147 11020328 10.1006/meth.2000.1054
Ingvarsson S Molecular genetics of breast cancer progression Semin Cancer Biol 1999 9 277 288 10448115 10.1006/scbi.1999.0124
Nathanson KL Wooster R Weber BL Breast cancer genetics: what we know and what we need Nat Med 2001 7 552 556 11329055 10.1038/87876
Lengauer C Kinzler KW Vogelstein B Genetic instabilities in human cancers Nature 1998 396 643 649 9872311 10.1038/25292
Gray JW Collins C Genome changes and gene expression in human solid tumors Carcinogenesis 2000 21 443 452 10688864 10.1093/carcin/21.3.443
Bachman KE Argani P Samuels Y Silliman N Ptak J Szabo S Konishi H Karakas B Blair BG Lin C The PIK3CA gene is mutated with high frequency in human breast cancers Cancer Biol Ther 2004 3 772 775 15254419
|
16168105
|
PMC1242128
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 May 31; 7(5):R609-R616
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1262
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12631616810710.1186/bcr1263Research ArticleObesity promotes 7,12-dimethylbenz(a)anthracene-induced mammary tumor development in female zucker rats Hakkak Reza [email protected] Andy W [email protected] Stewart L [email protected] Pippa M [email protected] George J [email protected] Chan Hee [email protected] Thomas [email protected] Soheila [email protected] Department of Dietetics and Nutrition, University of Arkansas for Medical Sciences, Little Rock, Arkansas, USA2 Department of Pediatrics, University of Arkansas for Medical Sciences, Little Rock, Arkansas, USA3 Arkansas Children's Hospital Research Institute, Little Rock, Arkansas, USA4 Arkansas Cancer Research Center, University of Arkansas for Medical Sciences, Little Rock, Arkansas, USA5 Department of Pathology, University of Arkansas for Medical Sciences, Little Rock, Arkansas, USA2005 6 6 2005 7 5 R627 R633 8 2 2005 7 4 2005 29 4 2005 6 5 2005 Copyright © 2005 Hakkak et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
High body mass index has been associated with increased risk for various cancers, including breast cancer. Here we describe studies using 7,12-dimethylbenz(a)anthracene (DMBA) to investigate the role of obesity in DMBA-induced mammary tumor susceptibility in the female Zucker rat (fa/fa), which is the most widely used rat model of genetic obesity.
Method
Fifty-day-old female obese (n = 25) and lean (n = 28) Zucker rats were orally gavaged with 65 mg/kg DMBA. Rats were weighed and palpated twice weekly for detection of mammary tumors. Rats were killed 139 days after DMBA treatment.
Results
The first mammary tumor was detected in the obese group at 49 days after DMBA treatment, as compared with 86 days in the lean group (P < 0.001). The median tumor-free time was significantly lower in the obese group (P < 0.001). Using the days after DMBA treatment at which 25% of the rats had developed mammary tumors as the marker of tumor latency, the obese group had a significantly shorter latency period (66 days) than did the lean group (118 days). At the end of the study, obese rats had developed a significantly (P < 0.001) greater mammary tumor incidence (68% versus 32%) compared with the lean group. The tumor histology of the mammary tumors revealed that obesity was associated with a significant (P < 0.05) increase in the number of rats with at least one invasive ductal and lobular carcinoma compared with lean rats.
Conclusion
Our results indicate that obesity increases the susceptibility of female Zucker rats to DMBA-induced mammary tumors, further supporting the hypothesis that obesity and some of its mediators play a significant role in carcinogenesis.
==== Body
Introduction
Obesity has been identified as an epidemic in the USA for more than two decades, yet the proportion of overweight and obese adults in the population continues to grow. The most recent data from the 1999–2000 National Health and Nutrition Examination Survey (NHANES) [1] showed that almost 65% of adults in the USA are overweight, defined as having a body mass index (BMI) greater than 25 kg/m2. This is a significant increase from the 56% for adults reported as overweight in NHANES III, which was conducted between 1988 and 1994. The prevalence of obesity, defined as a BMI of 30 kg/m2 or greater, also increased dramatically from 23% to 31% during the same period. It is estimated that the prevalence of obesity in adults will rise to 39% by the year 2008. This trend has alarming health and economic implications, because obesity is associated with major causes of morbidity and mortality such as diabetes, cardiovascular disease and several types of cancers, including breast cancer [2,3]. Breast cancer is the most common malignant tumor among women, being the second leading killer of women in the USA [4]. Epidemiological studies link breast cancer and obesity in postmenopausal women [4-6]; almost half of breast cancer cases among postmenopausal women occur in those with a BMI in excess of 29 kg/m2 [7]. Although a number of studies have shown that excess weight is a risk factor for breast and other cancers, Calle and coworkers [8] concluded that 20% of all deaths from cancer in women aged over 50 years old could be attributed to being overweight or obese.
In experimental models, higher body weight has been associated with an increase in both spontaneous and chemically induced mammary tumors in various strains of mice [7,9-12]. In order to understand better the mechanisms that are associated with obesity and cancer, we have turned our attention to Zucker rats as a model. The Zucker rat (fa/fa) is the best known, most widely used rat model of genetic obesity. Obesity in the Zucker rat is inherited as an autosomal-recessive trait caused by a mutation (fa) in the leptin receptor gene [13,14], discovered by Zucker and Zucker [15,16]. Animals homozygous for the fa allele become noticeably obese by age 3–5 weeks, and by 14 weeks of age more than 40% of their body is composed of lipids [17]. Many investigators have used this model to study the development, etiology, associated pathogenesis, possible treatment, and putative mechanisms of severe obesity [18]. Obese Zucker rats develop hyperinsulinemia and insulin resistance before they develop obesity-associated, non-insulin-dependent diabetes mellitus in a manner similar to that in humans [19]. Lean Zucker rats, in contrast, exhibit normal metabolic function and are considered ideal controls. Consequently, this model is an ideal one in which to investigate the relationship between obesity and mammary tumor development.
There has been only one study published that used female Zucker rats as a model to investigate the role of obesity in mammary tumor development [20]. That study used N-methyl-N-nitrosourea (MNU) as a carcinogen and reported that lean rats developed more mammary tumors than did obese ones. Because of its similarity to human breast cancer, researchers have widely used the 7,12-dimethylbenz(a)anthracene (DMBA)-induced rat mammary carcinoma model to investigate breast carcinogenesis. In the present study we observed that DMBA-induced mammary tumors in obese Zucker rats develop faster than they do in lean counterparts, and that obese animals were at more than double the risk for developing DMBA-induced mammary tumors. Therefore, our results suggest that this model parallels the epidemiological data and is an appropriate model in which to investigate the mechanism(s) that underlie the role of obesity in mammary tumor development and possible prevention strategies.
Materials and methods
Experimental design
All animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Arkansas for Medical Sciences. Obese fa/fa (n = 25) and lean (n = 28) Zucker rats were purchased at age 6 weeks (Harlan Industries, Indianapolis, IN, USA) and housed in the animal facilities at the Arkansas Children's Hospital Research Institute. Rats were housed two per cage in polycarbonate cages and allowed free access to water and regular chow (Harlan-Teklad, Madison, WI, USA). At the age of 50 days, all rats received via gavage 65 mg/kg DMBA (Sigma Chemical Co., St. Louis, MO, USA), a chemical procarcinogen used widely to produce mammary adenocarcinoma in rats, in sesame oil [21,22]. Rats were weighed twice per week. Beginning 2 weeks after DMBA treatment, all rats were palpated twice weekly to detect mammary tumors. The detection date and location of each mammary tumor was recorded for each rat. Rats were killed 139 days after DMBA treatment. All mammary tumors were excised, counted, and weighed. Rats with tumor masses exceeding 2.5 cm in diameter were killed early for humane reasons, in accordance with our Institutional Animal Care and Use Committee-approved animal protocol. Sections of all tumors were placed in 10% neutral buffered formalin for histopathologic analysis. Sections (5 μm) of the paraffin-embedded tumors were stained with hematoxylin and eosin for histologic analysis.
Pathology
A board-certified anatomic pathologist (SK) evaluated tumors in a blinded protocol and classified them as benign or intraductal proliferation, as shown by multiple papillomas or with ductal hyperplasia, ductal carcinoma in situ, or invasive ductal and lobular carcinoma.
Statistical analysis
A one-way analysis of variance [23] was used to analyze body weight following DMBA treatment. The tumor-free times were plotted using a Kaplan–Meier curve [24] and the median times were compared using a generalized Wilcoxon test [25]. When there is no censoring, this test is equivalent to a Mann–Whitney rank sum test. Fisher's exact test [26] was used to compare the percentage of rats with tumors and tumor histology in each group. The median numbers of tumors per tumor-bearing rat (multiplicity) for each group were compared using the nonparametric Mann–Whitney test [27]. Statistical significance was set at P < 0.05, and all P values were unadjusted for multiple comparisons. For the few rats that were killed early because of tumor burden, we assumed that the number of tumors remained constant until the end of the study. Data analyses were generated and plots were constructed using SPSS© version 12.0 for Windows (SPSS Inc., Chicago, IL, USA).
Results
Body and organ weights
As expected, all rats gained weight during the course of the experiment. The average body weights (mean ± standard error) are shown in Fig. 1a. Obese rats gained significantly (P < 0.001) more weight than did lean rats. At the beginning of the study (7 days before DMBA treatment) the mean body weights of lean and obese rats were 102.5 ± 1.5 g and 148 ± 3.5 g, respectively. At the end of the study, the final mean body weights for lean and obese rats were 262.3 ± 4.2 g and 517.2 ± 17.9 g, respectively. The obese rats gained approximately twice the weight of the lean rats. Obesity was associated with a significant increase in liver (P < 0.001) and kidney (P < 0.001) weights compared with those in the lean group (Table 1). This increase in liver weight in obese rats also was evident when liver weight was expressed as percentage of body weight (P < 0.001); however, kidney weight as a percentage of body weight was not affected significantly by obesity.
Time course for tumor formation, latency and multiplicity
The time course of palpable mammary tumor detection is shown in Fig. 1b, and data are presented in Table 2. The tumor latency (the number of days after DMBA treatment until detection of the first mammary tumor) was shorter in obese rats than in the lean rats. The first mammary tumor detected in obese rats was 49 days after DMBA treatment versus 86 days in lean rats, which represents a significant 37-day delay for the development of mammary tumors in lean rats (Table 2). In addition, it took only 66 days after DMBA treatment for 25% of the obese rats to develop mammary tumors versus 118 days in lean rats – a delay of 52 days. The median tumor-free time was significantly lower in the obese group (P < 0.001; Fig. 1b). By the end of the study (139 days after DMBA treatment), 68% of obese rats had developed mammary tumors as compared with only 32% of the lean rats (P < 0.001; Fig. 1b). The median number of mammary tumors per tumor-bearing rat (multiplicity) increased from one tumor per rat in the lean group (range: one to four tumors per rat) to two tumors per rat in the obese group (range: one to four tumors per rat; Table 2).
For further comparison of mammary tumor development we evaluated tumor multiplicity at three time points (70, 105 and 139 days after DMBA treatment). As demonstrated in Fig. 1c, the obese rats developed mammary tumors earlier than did lean rats. At 70 days after DMBA treatment, several rats in the obese group had one to three tumors each, whereas no mammary tumors were detected in the lean group. At 105 days after DMBA treatment there were several rats in the obese group with tumors, including five rats with one tumor each, four rats with two tumors, two rats with three tumors, and one rat with four tumors. This is in contrast to the lean group, in which only three rats had one tumor each and one rat had two tumors. At 139 days after DMBA treatment, there were seven rats in obese group with one tumor each, three rats with two tumors, four rats with three tumors, and three rats with four tumors. In contrast, the lean group had five rats with one tumor each, two rats with two tumors, one rat with three tumors, and one rat with four tumors. These findings indicate for the first time that obese Zucker rats are an excellent model for investigating the role of obesity in DMBA-induced mammary tumor development.
Mammary tumor characteristics
Mammary tumor histology data and morphology are presented in Table 2 and Fig. 2, respectively. Nine rats (32%) in the lean group developed mammary tumors compared with 17 rats (68%) in the obese group. Only two rats (7%) in the lean group had at least one tumor graded as intraductal proliferation compared with four rats (16%) in the obese group. Five rats (19%) in the lean group had at least one tumor graded as ductal carcinoma in situ, compared with six rats (24%) in the obese group. Obesity significantly increased (P < 0.05) the number of rats (7 rats, 28%) with at least one tumor graded as invasive ductal and lobular carcinoma compared with the lean group (2 rats, 7%). A total of 53 mammary tumors were detected in the study; 37 tumors (70% of the total tumors) were detected in the obese group versus 16 tumors (30% of the total tumors) in the lean group. Obesity was associated with a nonsignificant increase in tumor weight.
Discussion
Higher body weight is associated with increased incidence of both spontaneous and chemically induced mammary tumors in various strains of mice [9,10,12,28]. For example, the Leprdb and Lepob-TGFα transgenic mouse model, used to investigate the effects of obesity on tumorigenesis, are leptin deficient or have a leptin receptor defect. However, these mice do not develop mammary tumors and therefore are not a suitable model in which to study the role of obesity in breast cancer [29]. In the present study we observed that obese Zucker rats are more susceptible to DMBA-induced tumorigenesis than are lean rats. These data clearly demonstrate the following when DMBA was administered as a carcinogen to female Zucker rats: obese rats developed mammary tumors at a faster rate than did lean rats; obese rats exhibited shorter latency periods, both at the time of appearance of the first tumor and at the day at which tumor incidence reached 25%, than lean did rats; obesity resulted in a greater incidence of mammary tumors; obese rats developed significantly more invasive ductal and lobular carcinoma than did lean rats; and obese rats had a higher tumor multiplicity, albeit not statistically significantly so, compared with lean rats.
Using Sprague–Dawley female rats, Klurfeld and colleagues [30] studied the effects of calorie restriction on mammary tumor induction by DMBA in rats fed a high-fat diet. They found that a 25% calorie restriction resulted in a significant reduction in mammary tumor incidence and tumor weight in both control rats and rats fed a high-fat diet. Further experiments implicated decreased serum insulin and insulin-like growth factor I levels in the inhibition of mammary tumor promotion in calorie restricted rats [31].
In agreement with the present data in female Zucker rats, genetically obese LA/N-cp (corpulent) female rats developed more mammary tumors than did phenotypically lean littermates when they were treated with DMBA. Klurfeld and coworkers [28] reported that calorie restriction (40%) reduced tumor incidence to 27%, compared with the 100% tumor incidence observed in obese animals fed ad libitum. Those investigators suggested that calorie restriction resulted in decreased insulin levels, implicating hyperinsulinemia in obese ad libitum fed rats as a mechanism in increased tumor promotion.
The number of overweight and obese Americans has doubled in the past two decades, which may have an impact on cancer risk and survival [1]. The association between obesity and breast cancer risk has been inconclusive in a number of epidemiologic studies. Several studies have reported that increased body mass in premenopausal women is not associated with increased risk for breast cancer, whereas the same studies have demonstrated a positive association between increased body mass and breast cancer risk in postmenopausal women [32,33]. Recent epidemiologic studies have suggested that the age at which a woman gains weight may be a more relevant factor in determining breast cancer risk [34].
Our results are in contrast to those of a previous study, which used MNU as a carcinogen to induce mammary tumors. In that study [20] the authors reported that lean and obese rats had developed tumors at the same rate at 29 weeks after MNU treatment. However, 50% of the lean rats developed carcinomas of the mammary gland, as compared with only 10% of the obese rats. Also, they reported a higher incidence of colorectal tumors in obese rats than in lean rats. Although MNU and DMBA both produce mammary tumors in female rats, these tumors arise by different mechanisms of action [35]. MNU is a direct acting carcinogen and does not require metabolic activation in order to form adducts that damage DNA [36,37]. DMBA is a procarcinogen that requires metabolic activation by cytochrome P450 enzymes to reactive metabolites (dihydrodiolepoxides) that can form mutagenic DNA adducts [35]. These reactions are catalyzed principally by CYP1A2 in the liver and CYP1A1 and CYP1B1 in peripheral tissues such as the mammary gland.
Our findings are consistent with those reported for induction of colon cancer with azoxymethane in mature, genetically obese male Zucker rats. In that study [38] obese Zucker rats had significantly more colonic aberrant crypt formation than did any of the lean rats. Although the mechanism for an association with colon carcinogenesis is unknown, it is hypothesized that insulin and leptin resistance may play a role because leptin levels are typically threefold higher in the genetically obese Zucker (fa/fa) rats than in their lean counterparts, because of the leptin receptor defect in these animals [39]. However, recent results from epidemiological studies do not support the hypothesis that plasma leptin is a risk factor for breast cancer [40]. Nevertheless, body composition and weight are considered breast cancer risk factors that may influence prognosis [41].
The relevance of DMBA is that humans are exposed to DMBA and other polycyclic aromatic hydrocarbons through environmental or dietary sources, which may function in a synergistic manner with obesity and breast carcinogenesis. Exposure to polycyclic aromatic hydrocarbons in meats cooked at high temperature has been implicated in the development of breast cancer [42,43]. We have demonstrated that obesity increases the susceptibility of female Zucker rats to development of mammary tumors when DMBA is used as the inducing carcinogen. The mechanisms responsible for the observed increased risk for developing mammary tumors with obesity are not clearly understood. Several mechanisms may play roles in explaining the relationship we observed between obesity and increased mammary tumor development, including adipose associated hormones, adipokines, and inflammation [44], which are manifested by the general genetics of the animals; the exact relationships remain undefined and further work is needed. Future experiments with this model will focus on delineating the mechanisms responsible for this increased susceptibility to mammary cancer.
Conclusion
In conclusion, we have demonstrated that obesity increases the susceptibility of female Zucker rats to development of mammary tumors when DMBA is used as the inducing carcinogen. The mechanisms responsible for the observed increased risk for developing mammary tumors with obesity remain to be defined.
Abbreviations
BMI = body mass index; DMBA = 7,12-dimethylbenz(a)anthracene; MNU = N-methyl-N-nitrosourea.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RH applied for and received funding for this project, designed and supervised the study, and participated in collection of tumor data and writing of the manuscript. AH coordinated and carried out the daily activities regarding overseeing the animals, recording data, and processing the samples. SM participated in the study design, interpretation of the results, and drafting of the manuscript. PS and CHJ participated in the study design and performed the statistical analyses. GF participated in the study design. TK-E contributed to the study design, analysis and interpretation of the results, and writing of the manuscript. SK provided expertise in classifying and analyzing all of the mammary tumors and participated in interpretation of study results and writing of the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This work was supported by the University of Arkansas for Medical Sciences Children's University Medical Group and Arkansas Cancer Research Center Breast Cancer Program and the Arkansas Bioscience Institute, the major research component of the Tobacco Settlement Proceeds Act of 2000. The authors wish to thank Phaedra Yount for valuable assistance in preparing the manuscript.
Figures and Tables
Figure 1 Body weights and mammary tumor incidence and multiplicity in DMBA-treated Zucker rats. (a) Body weights of lean and obese female rats. (b) Mammary tumor incidence (percentage of rats with tumors) in female rats. Dashed lines indicate the post-DMBA (7,12-dimethylbenz(a)anthracene) days at which 25% of the obese and lean rats developed at least one mammary tumor. (c) Mammary tumor multiplicity in obese and lean rats.
Figure 2 Mammary tumor histology. (a) Intraductal proliferation (intraductal papilloma). Original magnification: 40×. (b) Higher magnification of panel a. Original magnification 200×. (c) Ductal carcinoma in situ. Original magnification: 40×. (d) Higher magnification shows proliferation of uniform neoplastic cells with high nucleus to cytoplasmic ratio within ductal structures. Original magnification: 200×. (e) Invasive ductal carcinoma; arrow shows tumor necrosis in lower left corner. Original magnification: 40×. (f) Higher magnification shows neoplastic cells diffusely infiltrating the stroma. Original magnification: 200×.
Table 1 Organ weights of lean and obese female zucker rats
Organ Lean (n = 28) Obese (n = 25)
Liver
Absolute 8.10 ± 0.14 22.08 ± 0.72**
Relative 3.05 ± 0.038 4.38 ± 0.12**
Kidney
Absolute 1.62 ± 0.35 2.95 ± 0.10**
Relative 0.61 ± 0.01 0.59 ± 0.02
Organ weights (absolute weights) are given in grams; organ weight to body weight ratios are given as grams organ weight/grams body weight as a percentage (mean ± standard error). **P < 0.001 versus the lean group (Fisher's exact test).
Table 2 Characteristics of DMBA-induced mammary tumors in lean and obese female Zucker rats
Characteristic Lean (n = 28) Obese (n = 25)
Tumor onset
Day of first tumora 86 49**
Day at 25% tumorsb 118 66**
Pc 0.001
Tumor incidence
% of rats with tumorsd 32 68**
Pe 0.001
Multiplicityf 1 (1–4) 2 (1–4)
Pc 0.42
Tumor weight (g)g 0.37 (0.01–6.44) 0.42 (0.05–7.70)
Pc 0.47
Rats with tumorsh
IDP 2 (7%) 4 (16%)
DCIS 5 (19%) 6 (24%)
IDC 2(7%) 7 (28%)*
aPost-DMBA (7,12-dimethylbenz(a)anthracene) day at which the first mammary tumor was detectable by palpation. bPost-DMBA day at which the probability of tumor development was 25%. cP values based on generalized Wilcoxon test. dPercentage of rats with at least one mammary tumor. eP values based on the Fisher's exact test. fMedian number of tumors in tumor bearing rats (minimum-maximum in parentheses). gMedian tumor weight (minimum-maximum in parentheses). hNumber and percentage of rats with at least one tumor graded as intraductal proliferation (IDP), ductal carcinoma in situ (DCIS), or invasive ductal and lobular carcinoma (IDC), as described in the Materials and method section. **P < 0.001, *P < 0.05.
==== Refs
National Center for Health Statistics National Health and Nutrition Examination Survey
Anonymous NHLBI Obesity Education Initiative Expert Panel Obes Res 1998 6 51S 9813653
Mokdad AH Ford ES Bowman BA Dietz WH Vinicor F Bales VS Marks JS Prevalence of obesity, diabetes, and obesity-related health risk factors, 2001 JAMA 2003 289 76 79 12503980 10.1001/jama.289.1.76
Cleary MP Maihle NJ The role of body mass index in the relative risk of developing premenopausal versus postmenopausal breast cancer Proc Soc Exp Biol Med 1997 216 28 43 9316608
Willett WC Fat, energy and breast cancer J Nutr 1997 127 921S 923S 9164264
Galanis DJ Kolonel LN Lee J Le Marchand L Anthropometric predictors of breast cancer incidence and survival in a multi-ethnic cohort of female residents of Hawaii, United States Cancer Causes Control 1998 9 217 224 9578299 10.1023/A:1008842613331
Ballard-Barbash R Swanson CA Body weight: estimation of risk for breast and endometrial cancers Am J Clin Nutr 1996 63 437S 441S 8615337
Calle EE Rodriguez C Walker-Thurmond K Thun MJ Overweight, obesity, and mortality from cancer in a prospectively studied cohort of U.S. adults N Engl J Med 2003 348 1625 1638 12711737 10.1056/NEJMoa021423
Waxler SH Tabar P Melcher LR Obesity and the time of appearance of spontaneous mammary carcinoma in C3H mice Cancer Res 1953 13 276 278 13042818
Waxler S Obesity and cancer susceptibility in mice Am J Clin Nutr 1960 8 760 766
Haseman JK Bourbina J Eustis SL Effect of individual housing and other experimental design factors on tumor incidence in B6C3F1 mice Fundam Appl Toxicol 1994 23 44 52 7958562 10.1006/faat.1994.1077
Wolff GL Kodell RL Cameron AM Medina D Accelerated appearance of chemically induced mammary carcinomas in obese yellow (Avy/A) (BALB/c X VY) F1 hybrid mice J Toxicol Environ Health 1982 10 131 142 6290678
Tartaglia LA Dembski M Weng X Deng N Culpepper J Devos R Richards GJ Campfield LA Clark FT Deeds J Identification and expression cloning of a leptin receptor, OB-R Cell 1995 83 1263 1271 8548812 10.1016/0092-8674(95)90151-5
Chua SC JrChung WK Wu-Peng XS Zhang Y Liu SM Tartaglia L Leibel RL Phenotypes of mouse diabetes and rat fatty due to mutations in the OB (leptin) receptor Science 1996 271 994 996 8584938
Zucker L Zucker TF Fatty, a new mutation in the rat J Heredity 1961 52 275 278
Zucker TF Zucker LM Fat accretion and growth in the rat J Nutr 1963 80 6 19 14004033
Zucker LM Fat mobilization in vitro and in vivo in the genetically obese Zucker rat 'fatty' J Lipid Res 1972 13 234 243 5016304
Bray GA York DA Fisler JS Experimental obesity: a homeostatic failure due to defective nutrient stimulation of the sympathetic nervous system Vitam Horm 1989 45 1 125 2688303
Bray GA The Zucker-fatty rat: a review Fed Proc 1977 36 148 153 320051
Lee WM Lu S Medline A Archer MC Susceptibility of lean and obese Zucker rats to tumorigenesis induced by N-methyl-N-nitrosourea Cancer Lett 2001 162 155 160 11146220 10.1016/S0304-3835(00)00635-2
Hakkak R Korourian S Shelnutt SR Lensing S Ronis MJ Badger TM Diets containing whey proteins or soy protein isolate protect against 7,12-dimethylbenz(a)anthracene-induced mammary tumors in female rats Cancer Epidemiol Biomarkers Prev 2000 9 113 117 10667471
Hakkak R Kourourian S Fletcher T Hale J Fergusonn M Ronis M Badger T The effects of rice protein isolate on DMBA-induced mammary tumors in rats Proc Am Assoc Cancer Res 2002 48 823
Altman DG Practical Statistics for Medical Research 1991 London: Chapman and Hall
Cox DR Oakes D Analysis of Survival Data 1984 London: Chapman and Hall
Gross AJ Clark VA Survival Distribution: Reliability Applications in Biomedical Sciences 1975 New York: Wiley
Bhattacharyya GK Johnson RA Statistical Concepts and Methods 1977 New York: Wiley
Kruskal WH Wallis WA Use of ranks in one-criterion variance analysis J Am Stat Assoc 1952 47 583 621
Klurfeld DM Lloyd LM Welch CB Davis MJ Tulp OL Kritchevsky D Reduction of enhanced mammary carcinogenesis in LA/N-cp (corpulent) rats by energy restriction Proc Soc Exp Biol Med 1991 196 381 384 1901169
Hu X Juneja SC Maihle NJ Cleary MP Leptin: a growth factor in normal and malignant breast cells and for normal mammary gland development J Natl Cancer Inst 2002 94 1704 1711 12441326
Klurfeld DM Welch CB Lloyd LM Kritchevsky D Inhibition of DMBA-induced mammary tumorigenesis by caloric restriction in rats fed high-fat diets Int J Cancer 1989 43 922 925 2497075
Ruggeri BA Klurfeld DM Kritchevsky D Furlanetto RW Caloric restriction and 7,12-dimethylbenz(a)anthracene-induced mammary tumor growth in rats: alterations in circulating insulin, insulin-like growth factors I and II, and epidermal growth factor Cancer Res 1989 49 4130 4134 2501021
Lahmann PH Hoffmann K Allen N van Gils CH Khaw KT Tehard B Berrino F Tjonneland A Bigaard J Olsen A Body size and breast cancer risk: findings from the European Prospective Investigation into Cancer And Nutrition (EPIC) Int J Cancer 2004 111 762 771 15252848 10.1002/ijc.20315
van den Brandt PA Spiegelman D Yaun SS Adami HO Beeson L Folsom AR Fraser G Goldbohm RA Graham S Kushi L Pooled analysis of prospective cohort studies on height, weight, and breast cancer risk Am J Epidemiol 2000 152 514 527 10997541 10.1093/aje/152.6.514
Morimoto LM White E Chen Z Chlebowski RT Hays J Kuller L Lopez AM Manson J Margolis KL Muti PC Obesity, body size, and risk of postmenopausal breast cancer: the Women's Health Initiative (United States) Cancer Causes Control 2002 13 741 751 12420953 10.1023/A:1020239211145
Rowlands JC He L Hakkak R Ronis MJ Badger TM Soy and whey proteins downregulate DMBA-induced liver and mammary gland CYP1 expression in female rats J Nutr 2001 131 3281 3287 11739881
Russo J Russo IH Experimentally induced mammary tumors in rats Breast Cancer Res Treat 1996 39 7 20 8738602
Russo IH Russo J Mammary gland neoplasia in long-term rodent studies Environ Health Perspect 1996 104 938 967 8899375
Weber RV Stein DE Scholes J Kral JG Obesity potentiates AOM-induced colon cancer Dig Dis Sci 2000 45 890 895 10795750 10.1023/A:1005560621722
White BD Martin RJ Evidence for a central mechanism of obesity in the Zucker rat: role of neuropeptide Y and leptin Proc Soc Exp Biol Med 1997 214 222 232 9083255
Stattin P Soderberg S Biessy C Lenner P Hallmans G Kaaks R Olsson T Plasma leptin and breast cancer risk: a prospective study in northern Sweden Breast Cancer Res Treat 2004 86 191 196 15567935 10.1023/B:BREA.0000036782.11945.d7
Enger SM Greif JM Polikoff J Press M Body weight correlates with mortality in early-stage breast cancer Arch Surg 2004 139 954 960 15381612 10.1001/archsurg.139.9.954
Li D Wang M Dhingra K Hittelman WN Aromatic DNA adducts in adjacent tissues of breast cancer patients: clues to breast cancer etiology Cancer Res 1996 56 287 293 8542582
Rundle A Tang D Hibshoosh H Estabrook A Schnabel F Cao W Grumet S Perera FP The relationship between genetic damage from polycyclic aromatic hydrocarbons in breast tissue and breast cancer Carcinogenesis 2000 21 1281 1289 10874004 10.1093/carcin/21.7.1281
Trayhurn P Wood IS Adipokines: Inflammation and the pleiotropic role of white adipose tissue Br J Nutr 2004 92 347 355 15469638 10.1079/BJN20041213
|
16168107
|
PMC1242129
|
CC BY
|
2021-01-04 16:04:33
|
no
|
Breast Cancer Res. 2005 Jun 6; 7(5):R627-R633
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1263
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12641616810610.1186/bcr1264Research ArticlePhase I clinical study of the recombinant antibody toxin scFv(FRP5)-ETA specific for the ErbB2/HER2 receptor in patients with advanced solid malignomas von Minckwitz Gunter 1Harder Sebastian 2Hövelmann Sascha 3Jäger Elke 4Al-Batran Salah-Eddin 4Loibl Sibylle 1Atmaca Akin 4Cimpoiasu Christian 1Neumann Antje 4Abera Aklil 3Knuth Alexander 45Kaufmann Manfred 1Jäger Dirk 45Maurer Alexander B 3Wels Winfried S [email protected] Department of Gynecology and Obstetrics, University Hospital Frankfurt, Germany2 Institute for Clinical Pharmacology, University Hospital Frankfurt, Germany3 G2M Cancer Drugs AG, Frankfurt, Germany4 Medizinische Klinik II, Hämatologie-Onkologie, Krankenhaus Nordwest, Frankfurt, Germany5 Department of Oncology, University Hospital Zürich, Switzerland6 Chemotherapeutisches Forschungsinstitut Georg-Speyer-Haus, Frankfurt, Germany2005 1 6 2005 7 5 R617 R626 20 1 2005 25 2 2005 6 4 2005 4 5 2005 Copyright © 2005 von Minckwitz et al, licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is cited.
Introduction
ScFv(FRP5)-ETA is a recombinant antibody toxin with binding specificity for ErbB2 (HER2). It consists of an N-terminal single-chain antibody fragment (scFv), genetically linked to truncated Pseudomonas exotoxin A (ETA). Potent antitumoral activity of scFv(FRP5)-ETA against ErbB2-overexpressing tumor cells was previously demonstrated in vitro and in animal models. Here we report the first systemic application of scFv(FRP5)-ETA in human cancer patients.
Methods
We have performed a phase I dose-finding study, with the objective to assess the maximum tolerated dose and the dose-limiting toxicity of intravenously injected scFv(FRP5)-ETA. Eighteen patients suffering from ErbB2-expressing metastatic breast cancers, prostate cancers, head and neck cancer, non small cell lung cancer, or transitional cell carcinoma were treated. Dose levels of 2, 4, 10, 12.5, and 20 μg/kg scFv(FRP5)-ETA were administered as five daily infusions each for two consecutive weeks.
Results
No hematologic, renal, and/or cardiovascular toxicities were noted in any of the patients treated. However, transient elevation of liver enzymes was observed, and considered dose limiting, in one of six patients at the maximum tolerated dose of 12.5 μg/kg, and in two of three patients at 20 μg/kg. Fifteen minutes after injection, peak concentrations of more than 100 ng/ml scFv(FRP5)-ETA were obtained at a dose of 10 μg/kg, indicating that predicted therapeutic levels of the recombinant protein can be applied without inducing toxic side effects. Induction of antibodies against scFv(FRP5)-ETA was observed 8 days after initiation of therapy in 13 patients investigated, but only in five of these patients could neutralizing activity be detected. Two patients showed stable disease and in three patients clinical signs of activity in terms of signs and symptoms were observed (all treated at doses ≥ 10 μg/kg). Disease progression occurred in 11 of the patients.
Conclusion
Our results demonstrate that systemic therapy with scFv(FRP5)-ETA can be safely administered up to a maximum tolerated dose of 12.5 μg/kg in patients with ErbB2-expressing tumors, justifying further clinical development.
Please see related commentary by Messmer and Kipps at,
==== Body
Introduction
Aberrant expression of the epidermal growth factor receptor or the closely related ErbB2 (HER2/neu) receptor tyrosine kinase has been implicated in the formation of various human malignancies [1,2], making these receptors interesting targets for directed anticancer therapeutics. Antibodies that block ligand binding or interfere with receptor function can directly inhibit the growth of cancer cells in addition to their potential to direct effector cells of the immune system to the tumor [3]. With the humanized mAb Herceptin™ (trastuzumab), an ErbB2-specific reagent for the treatment of breast carcinomas is in clinical use. Monotherapy with the antibody or combination with chemotherapy protocols resulted in increased clinical benefit for a significant proportion of patients with ErbB2-overexpressing metastatic breast cancers [4,5]. Nevertheless, responses could not be achieved in all patients with tumors expressing high ErbB2 levels, suggesting that in addition to enhanced expression of the target receptor, other factors such as limited recruitment of endogenous immune effector mechanisms or the presence of alternative signaling pathways in tumor cells can also influence treatment outcome.
In contrast to such unmodified antibodies, antibody toxins are not dependent on the inhibition of signaling or on the recruitment of complement or endogenous killer cells for antitumoral activity, but they combine antibody-mediated recognition of tumor cells with specific delivery of a potent cytotoxic effector molecule [6-8]. These tailor-made targeting reagents might therefore represent a valuable alternative to unmodified mAbs, and could complement their use in the clinic. ScFv(FRP5)-ETA is a recombinant single-chain antibody toxin with binding specificity for ErbB2-overexpressing tumor cells [9,10]. The N-terminal portion of the bacterially expressed molecule is contributed by a single-chain antibody fragment (scFv) derived from heavy-chain and light-chain variable domains of murine mAb FRP5, which recognizes the extracellular domain of human ErbB2 [11]. ScFv(FRP5)-ETA harbors a truncated Pseudomonas aeruginosa exotoxin A (ETA, PE) fragment (amino acids 252–613 of mature exotoxin A) at the C-terminus, which is devoid of the toxin's natural cell-binding domain [9]. Upon specific binding of the scFv domain to ErbB2 on the surface of tumor cells, the antibody toxin is internalized by receptor-mediated endocytosis, the enzymatic domain of the molecule is released into the cytoplasm and ADP ribosylates elongation factor 2, a critical component of the target cell's translation machinery [12]. Toxin-mediated inactivation of elongation factor 2 causes the inhibition of protein synthesis and results in subsequent tumor cell death by apoptosis [13,14].
Efficacy of scFv(FRP5)-ETA in the treatment of ErbB2-overexpressing tumors has been established in numerous preclinical in vitro and in vivo studies. ScFv(FRP5)-ETA displayed potent antitumoral activity in vitro against a wide range of established and primary human tumor cells, including breast and ovarian carcinomas [9,14,15], squamous cell carcinomas [10,16] and prostate carcinomas [17]. In experimental animals, locally or systemically applied scFv(FRP5)-ETA effectively inhibited the in vivo growth of human tumor xenografts [9,10,14,16], and murine and rat tumor cells stably transfected with human c-erbB2 constructs [18,19]. Thereby, complete elimination of subcutaneously growing tumors [16] and prevention of metastasis formation by disseminated cancer cells [19] was observed in some models.
Potent antitumoral activity in animal models has also been described for antibody toxins derived from the ErbB2-specific antibody e23 [20,21]. In cancer patients, however, intravenous application of such an e23-based antibody toxin resulted in severe liver toxicity, and effective doses could not be reached [22]. In contrast, for the scFv(FRP5)-ETA molecule utilizing the different FRP5 antibody domain, we could previously show that local treatment of cutaneous lesions of ErbB2-expressing tumors by intratumoral injection of the scFv(FRP5)-ETA molecule was well tolerated, and resulted in shrinkage or complete regression of injected tumor nodules in the majority of patients [23].
Here we now report the first systemic application of scFv(FRP5)-ETA in a phase I dose-finding study in human cancer patients, with the objective to assess the maximum tolerated dose and the dose-limiting toxicity. Furthermore, we have obtained data concerning pharmacokinetic properties of scFv(FRP5)-ETA and its ability to induce a neutralizing antibody response.
Patients and methods
Patients
Patients eligible for treatment with scFv(FRP5)-ETA had to be 18 years of age or older, with ErbB2-overexpressing tumors confirmed by immunohistochemistry (DAKO-Hercep test 3+) or fluorescence in situ hybridization analysis, showing clinical, radiological, or serological evidence for a progression. Other eligibility criteria included at least one previous palliative systemic chemotherapy treatment and an absence of any standard treatment option. Patients with serious illness or medical conditions besides the diagnosis of cancer, a Karnofsky index <60%, and immunoreactivity against scFv(FRP5)-ETA were excluded from the study. See Table 1 for further details on patient characteristics. The study was conducted in compliance with the Helsinki Declaration [24]. The treatment protocol and consent form were approved by the regulatory authorities and institutional ethics committees. Informed consent was obtained from the patients before therapy.
ErbB2-specific antibody toxin
Recombinant scFv(FRP5)-ETA was produced as an experimental drug under Good Manufacturing Practice conditions and was kindly provided by Ciba Geigy AG (Basel, Switzerland). Bacterial expression and isolation of recombinant protein from inclusion bodies was carried out following a protocol adapted for large-scale production from the basic procedures described elsewhere [18]. Homogeneity of the material was analyzed by SDS-PAGE and Coomassie staining, and the identity of the purified protein was confirmed by immunoblot analysis and amino acid sequencing. The content of endotoxins (<10 EU/ml at 0.1 mg/ml protein), content of Escherichia coli proteins (<11 μg/ml) and content of DNA (<20 pg/ml) were determined following standard procedures (data not shown). The antibody toxin was supplied as a sterile solution at 0.3 mg/ml scFv(FRP5)-ETA in a phosphate buffer [23]. Aliquots of scFv(FRP5)-ETA solution were transferred to 2 ml vials under sterile conditions in the hospital pharmacy and were stored at -70°C. Upon thawing and subsequent storage at temperatures between 4°C and 8°C, scFv(FRP5)-ETA retained full bioactivity for a minimum of 6 days (IC50 for ErbB2-expressing Renca-lacZ/ErbB2 cells, 5–6 ng/ml) [19,23]. For application, the required amount of antibody toxin was thawed and kept until use for a maximum of 5 days at temperatures between 4°C and 8°C.
Treatment schedule and dose levels
Experiments in mice and rats demonstrated efficacy of daily intravenous injections of scFv(FRP5)-ETA for 10 days against localized and metastatic tumors [18,19,25]. Based on these data, the treatment schedule for the phase I clinical study was developed: patients received a total of 10 intravenous infusions of scFv(FRP5)-ETA on day 1, day 2, day 3, day 4, day 5, day 8, day 9, day 10, day 11, and day 12. ScFv(FRP5)-ETA was diluted in physiological Ringer solution to achieve a total injection volume of 50 ml. A test dose of 10 μg intravenously was given as a 15-min infusion on day 1. The remaining dose was administered 4 hours later as a 15-min infusion. The dose levels of scFv(FRP5)-ETA were 2, 4, 10, 12.5, or 20 μg/kg per treatment day. Patients received 8 mg dexamethasone, 50 mg ranitidine, and 2 mg clemastine as a supportive treatment 30 min prior to scFv(FRP5)-ETA infusion to avoid anaphylactic reactions.
Tumor assessments
The modified Response Evaluation Criteria in Solid Tumors were used for objective tumor response assessment in this trial. Chest X-ray, abdominal ultrasound or computed tomography scan, specific measurement of an indicator lesion, bone scan, or bone X-ray in the case of hot spots in the bone scan were performed within 3 months before therapy and on day 29 after the onset of scFv(FRP5)-ETA therapy. To ensure comparability, every effort was made to use the same instrumental examination from baseline through follow-up.
Safety evaluation
The overall proportion of patients experiencing any toxicity was determined using the National Cancer Institute Common Toxicity Criteria, and the corresponding grading system was used to grade adverse events for recording in the case report form. For all adverse events not classified by National Cancer Institute Common Toxicity Criteria, the COSTART grading classification [26] was used (severity: 1, mild; 2, moderate; 3, severe; and 4, life-threatening). Cardiac function was determined before treatment and was monitored by electrocardiography, by Multiple Gated Acquisition scan, or by echocardiography on day 22 after initiation of therapy to assess potential cardiotoxicity.
Detection of scFv(FRP5)-ETA plasma levels by sandwich ELISA
Ninety-six-well microtiter plates (Greiner Bio-One, Frickenhausen, Germany) were coated overnight with 100 μl/well of 9 μg/ml goat anti-exotoxin A capture antibody (List Biological Laboratories, Campbell, CA, USA) diluted in 50 mM carbonate buffer, pH 9.5. After washing with PBS, the plates were blocked for 2 hours with PBS containing 1% BSA. For standard, control or patients' samples, 100 μl/well serum diluted 1:10 with PBS containing 6 mM ethylenediamine tetraacetic acid were added in duplicate, and were incubated for 2.5 hours at 37°C. After washing with PBS, 100 μl rabbit anti-exotoxin A antibody [9], diluted 1:250 in PBS, were added to each well for 1 hour at 37°C. After another washing step, bound rabbit antibodies were detected with 100 μl/well horseradish peroxidase-coupled anti-rabbit IgG antibody (Amersham Biosciences, Freiburg, Germany) for 1 hour at 37°C, after a final wash followed by 100 μl/well of 1% 3,3',5,5'-tetramethyl-benzidine (Sigma-Aldrich, Deisenhofen, Germany) substrate solution for 5–15 min at room temperature. Then the reaction was stopped by adding 50 μl/well of 1 M HCl, and the absorbance at 450 nm was measured using a Wallac Victor 2 ELISA reader (PerkinElmer Wallac, Freiburg, Germany).
Pharmacokinetic analysis
The main parameters measured were Cmax and the area under the concentration–time curve (AUC0–5 hours) of scFv(FRP5)-ETA at day 5, obtained from the individual concentration–time profiles (area under the concentration–time curve by trapezoidal rule) and using the software program TOPFIT® 2.0 (Fischer Verlag, Stuttgart, Germany). Furthermore, the elimination half-life and plasma clearance were determined from the concentration–time profiles using a noncompartmental approach from the TOPFIT Library.
Detection of circulating antibodies
Ninety-six-well microtiter plates (Greiner Bio-One) were coated overnight with 100 μl/well of 1 μg/ml scFv(FRP5)-ETA in PBS. After washing with PBS, the plates were blocked for 2 hours with PBS containing 1% BSA. For standard, control or patients' samples, 100 μl/well serum diluted 1:50 with PBS were added in duplicate, and were incubated for 3 hours at 37°C. After washing with Tris-buffered saline, 100 μl alkaline phosphatase-coupled rabbit anti-human IgG/IgM antibody (Sigma-Aldrich), diluted 1:2000 in PBS, were added to each well for 1 hour at 37°C, after a final wash followed by 100 μl/well nitrophenylpyrophosphate substrate solution at room temperature. The absorbance at 450 nm was measured using a Wallac Victor 2 ELISA reader (PerkinElmer Wallac). Neutralizing antibodies were determined in cell viability assays as described previously [23].
Results
The primary objectives of the study were the determination of the maximum tolerated dose and the dose-limiting toxicity (grade 4 hematologic toxicity or grade 3 nonhematologic toxicity) of scFv(FRP5)-ETA after intravenous application. The secondary objectives were the determination of the pharmacokinetic profile of scFv(FRP5)-ETA, the time to progression, the objective response rate (complete and partial), and the immunological response to treatment.
Eighteen patients suffering from ErbB2-expressing metastatic breast cancers (13 patients), prostate cancers (two patients), head and neck cancer (one patient), non small cell lung cancer (one patient), or transitional cell carcinoma (one patient) (Table 1) were given at least five daily infusions of scFv(FRP5)-ETA, with total daily doses ranging from 100 μg to 1.4 mg. A total of 15 patients received the complete treatment cycle of 10 days without showing signs of dose-limiting toxicity, whereas therapy in three patients was stopped on day 8 due to a grade 3–4 elevation of liver enzymes (alanine aminotransferase [ALT], aspartate aminotransferase [AST]) (Table 2). A dose escalation scheme was pursued, which started at 2 μg/kg scFv(FRP5)-ETA per treatment day, followed by 4, 10, and 20 μg/kg scFv(FRP5)-ETA. Due to dose-limiting toxicity in two out of three patients treated at 20 μg/kg scFv(FRP5)-ETA, a protocol amendment was adopted to include a further dose level at 12.5 μg/kg scFv(FRP5)-ETA. Five out of six patients at 12.5 μg/kg scFv(FRP5)-ETA received the complete dose without experiencing severe side effects. In one patient at this dose level, however, therapy had to be stopped on day 8 due to toxicity. No objective response was observed in any of the patients, but two patients remained in stable disease for more than 3 months, and clinical signs of activity were seen in three patients.
Toxicity
An increase in the serum levels of liver enzymes was found in the majority of the patients, with grade 1–2 elevation of ALT or AST seen in seven patients and grade 1–2 elevation of gamma-glutamyl transferase in six patients. A dose-limiting toxicity with grade 3–4 elevation of ALT or AST was found in two of three patients treated at 20 μg/kg scFv(FRP5)-ETA and in one patient treated at 12.5 μg/kg scFv(FRP5)-ETA, resulting in discontinuation of therapy in these patients on day 8. Enzyme levels returned to baseline values within 14 days after cessation of therapy in all patients. One patient with liver metastases, treated at the lowest dose level (2 μg/kg), developed cholestasis, which was due to massive disease progression but was unrelated to treatment, requiring a cessation of therapy on day 10. Another patient developed fever and dyspnoe on day 23 after onset of therapy, which resolved with antibiotics. However, this patient died on day 40, most probably unrelated to therapy but due to massive disease progression.
Anti-tumoral efficacy
Complete or partial remissions were not observed after scFv(FRP5)-ETA treatment, which was not unexpected given the severity of the patients' disease and their tumor load. Stable disease was seen in two patients, however: one prostate cancer patient treated at 10 μg/kg scFv(FRP5)-ETA per day, and one breast cancer patient treated at 12.5 μg/kg scFv(FRP5)-ETA per day. Furthermore, clinical signs of activity were observed in another three patients, with two patients treated at 10 and 12.5 μg/kg scFv(FRP5)-ETA per day experiencing signs of healing of cancer-related cutaneous lesions. In the patient receiving 10 μg/kg scFv(FRP5)-ETA, a reduction in the size of a cervical lymph node metastasis was also observed and the morphine dose could be reduced by 50%. A third patient treated at 20 μg/kg scFv(FRP5)-ETA per day, despite discontinuation of treatment on day 8 due to dose-related side effects, demonstrated signs of an inflammatory response and softening of a large tumor mass in her right breast. Interestingly, the breast cancer patient with stable disease and two of the patients with clinical signs of activity had previously progressed under therapy with trastuzumab (Table 1).
Pharmacokinetic profile
No scFv(FRP5)-ETA could be detected in patient plasma at the 2 μg/kg dose. At dose levels equal to or greater than 4 μg/kg scFv(FRP5)-ETA per day, there was a good correlation between the dose level and the plasma concentration, with peak levels of scFv(FRP5)-ETA reached 15 min after the beginning of infusion, and a fast decline to baseline levels within 4 hours, indicating that the drug was not accumulating (Fig. 1a). Peak concentrations at steady state (day 5 of therapy) ranged from 18 to 49 ng/ml (mean, 39 ng/ml) at 4 μg/kg scFv(FRP5)-ETA, ranged from 128 to 129 ng/ml (mean, 129 ng/ml) at 10 μg/kg scFv(FRP5)-ETA, ranged from 93 to 204 ng/ml (mean, 160 ng/ml) at 12.5 μg/kg scFv(FRP5)-ETA, and ranged from 115 to 307 ng/ml (mean, 236 ng/ml) at 20 μg/kg scFv(FRP5)-ETA (Fig. 2a). The correlation between the dose of scFv(FRP5)-ETA and the area under the concentration–time curve was less pronounced (Figs 1b and 2b). Plasma clearance was calculated as 6.6 l/hour at 4 μg/kg, as 5.3 l/hour at 10 μg/kg, as 4.9 l/hour at 12.5 μg/kg, and as 3.8 l/hour at 20 μg/kg. The calculated half-life of scFv(FRP5)-ETA at the three higher dose levels was approximately 44 min. The pharmacokinetic data are summarized in Table 3.
Immunogenicity
Antibody responses to scFv(FRP5)-ETA were analyzed in detail in 13 patients. None of the patients had pre-existing antibodies reactive with scFv(FRP5)-ETA, but most patients developed antibodies to scFv(FRP5)-ETA beginning on day 8 after onset of therapy, as determined by ELISA (Fig. 3). Only in five patients, however, could antibodies with neutralizing capacity against scFv(FRP5)-ETA be detected in cell viability assays. A strong neutralizing capacity (neutralizing activity at a serum dilution of 1:100) was only seen in two patients, whereas another three patients developed weak or moderate neutralizing activity (neutralizing activity only at a serum dilution of 1:50) (data not shown). Clinical symptoms were not associated with these responses (Table 2).
Discussion
Previous experimental work in various rodent models provided evidence for efficacy of the recombinant antibody toxin scFv(FRP5)-ETA against ErbB2-expressing tumors [9,10,14,16,18,19]. Further support for the use of scFv(FRP5)-ETA in tumor therapy arose from a study investigating intratumoral injection of the antibody toxin in patients with dermal metastases of ErbB2-expressing tumors [23]. Here we report results on the first systemic application of scFv(FRP5)-ETA in cancer patients, demonstrating safety at doses up to 12.5 μg/kg per day.
The predominant dose-limiting toxicity encountered in two of three patients treated at 20 μg/kg was transient liver toxicity, as indicated by a grade 3–4 elevation of serum liver enzymes (patient N10, ALT grade 4, AST grade 3; patient N12, ALT grade 3). An amendment to the protocol was implemented to test a dose level of 12.5 μg/kg. Liver toxicity (elevation of ALT/AST grade 3) was observed at this amended dose level in only one out of six patients. Therefore, 12.5 μg/kg was determined as the maximum tolerated dose. Liver toxicity can be a common complication encountered after treatment with recombinant toxins based on ETA. This was drastically demonstrated in a clinical study using erb-38, an ErbB2-specific toxin similar to scFv(FRP5)-ETA, which consists of a disulfide-bridged Fv fragment of the ErbB2-specific monoclonal antibody e23 linked to truncated Pseudomonas toxin [21]. Intravenous injection of 1 or 2 μg/kg recombinant protein every other day caused liver toxicity in all patients after 3 days of treatment [22]. This approximately 10-fold difference in the daily doses causing liver toxicity when comparing erb-38 with scFv(FRP5)-ETA might be explained by direct adverse effects of erb-38 against hepatocytes due to low ErbB2 expression in the liver, as postulated by Pai-Scherf and colleagues [22].
In tissue culture, scFv(FRP5)-ETA displayed selective cytotoxicity towards ErbB2-overexpressing tumor cells with IC50 values in the nanograms per milliliter range [9,17]. While scFv(FRP5)-ETA and other ErbB2-specific antibody toxins have not been compared directly, the same cell lines and similar assays were used in some studies for in vitro characterization. For example, when tested in protein synthesis inhibition assays, scFv(FRP5)-ETA displayed an IC50 value towards SKBR3 breast carcinoma cells of 29 ng/ml [9], compared with a value of 32 ng/ml for e23(Fv)PE40, an ErbB2-specific molecule employing a scFv antibody fragment of e23 for cell targeting [20]. Improved cytotoxic activity was reported for erb-38 [21], which might indeed explain the higher degree of toxicity of this antibody toxin in cancer patients. It is noteworthy, however, that significantly lower toxicity of scFv(FRP5)-ETA was also seen in animal experiments, where specific binding to endogenous ErbB2 on liver cells can be excluded as the cause for toxicity. ScFv(FRP5)-ETA could be applied intravenously in mice at doses up to 1 mg/kg daily for 10 days without causing any measurable side effects [19]. This contrasts with erb-38, for which an LD50 value in mice of 450 μg/kg was reported after three doses given every other day, and which caused death of animals by hepatic failure [22].
Various studies link the hepatotoxic effects of Pseudomonas exotoxin to the increased production of tumor necrosis factor alpha (TNF-α) by Kupffer cells in the liver and the resulting liver damage from activated T cells [27]. Mice depleted of T cells prior to ETA challenge failed to develop acute hepatic failure, whereas mice not immunologically compromised demonstrated hepatocyte apoptosis and increased plasma transaminase activity. Furthermore, in mice treated with the ETA-containing antibody toxin LMB-2, inhibition of TNF-α production in Kupffer cells by a specific TNF-binding protein or indomethacin prevented LMB-2-induced liver damage [28]. Agents such as infliximab, which neutralize the effects of TNF-α, are currently in clinical use for the treatment of rheumatoid arthritis and Crohn's disease [29]. These substances as well as nonsteroidal anti-inflammatory drugs may also be of use in preventing some of the unspecific toxic effects of ETA-based antibody toxins in humans.
Generally, cross-reactivity with normal tissues and severity of adverse reactions might at least in part depend also on the nature and the position of the epitope recognized by the antibody domain [30]. The mAb FRP5 and its scFv derivatives recognize a peptide epitope located in the N-terminal region of the receptor [14,31]. In contrast, the humanized ErbB2-specific antibody Herceptin™ (trastuzumab), which can induce cardiotoxicity in some patients [32], recognizes the juxtamembrane region of ErbB2 [33]. As in a previous study with locally applied scFv(FRP5)-ETA [23], here we have not observed cardiovascular complications in any of the patients treated with scFv(FRP5)-ETA, nor were such toxicities reported for the erb-38 molecule based on mAb e23 [22], for which information on the binding epitope is not available.
Our results demonstrate that intravenous administration of scFv(FRP5)-ETA leads to serum concentrations of the recombinant protein over several hours, predicted to be therapeutically relevant. IC50 values ranging from 10 to 100 ng/ml were determined in in vitro experiments with human tumor cells [16]. In the present study a peak serum concentration of 129 ng/ml was found in cancer patients at a dose of 10 μg/kg scFv(FRP5)-ETA, and serum concentrations between 50 and 100 ng/ml could be maintained for 2 hours after administration of 12.5 μg/kg scFv(FRP5)-ETA. The calculated half-life of 44 min indicates that the recombinant toxin is rapidly cleared from the body and is not accumulating. Nevertheless, systemic treatment with scFv(FRP5)-ETA in mice was successful despite a relatively short half-life of 30 min in the circulation [9], suggesting that sufficient amounts of the molecule could reach the tumor site.
Although objective responses could not be achieved in this study, clinical signs of activity such as healing of cutaneous lesions and stable disease were observed in some of the patients treated with scFv(FRP5)-ETA. The lack of major responses to treatment may be due to the advanced stage of the patients' disease, the limited treatment time of 2 × 5 days, or the limited number of patients treated at higher dose levels. Furthermore, problems with tumor cell accessibility can occur. Solid tumors of epithelial origin are often poorly vascularized, possibly limiting the use of large therapeutic molecules such as antibodies and antibody toxins when administered as a single agent. This might explain why most progress has been made so far in the clinical development of antibody toxins that target cell surface molecules such as CD22 and CD25 expressed on certain malignancies of hematologic origin [34,35], where tumor cells are usually more accessible. When scFv(FRP5)-ETA was directly administered into tumor lesions by intratumoral injection in a previous study, complete regression or partial reduction in the size of injected tumor nodules was found in five out of seven patients with tumors expressing high ErbB2 levels [23], indicating that the antibody toxin can very well selectively eliminate ErbB2-overexpressing target cells if they are accessible.
Antibody responses to scFv(FRP5)-ETA were analyzed in 13 patients. While none of these patients had pre-existing antibodies to ETA, they all developed antibody responses of varying intensity to scFv(FRP5)-ETA measurable after 8 days of therapy. Importantly, only in five of the 13 patients analyzed did these antibodies have scFv(FRP5)-ETA neutralizing activity, and two of these patients had stable disease for longer than 3 months. The pharmacokinetic parameters shown were determined on treatment day 5. Thereby a tendency towards a lower circulation half-life and lower area under the concentration–time curve values was found when compared with treatment day 1 (data not shown). This could at least partially be due to onset of anti-toxin antibody responses. During the second week of treatment, anti-toxin antibodies might have affected the half-life of scFv(FRP5)-ETA further, which was not formally investigated. Nevertheless, in a previous study, complete remission of tumor nodules locally injected with scFv(FRP5)-ETA could be achieved after a second treatment cycle despite pre-existing neutralizing antibodies induced during the first round of treatment [23]. Continued treatment of patients after development of anti-toxin antibodies might therefore still be efficacious.
Conclusion
Taken together, our results demonstrate that the ErbB2-specific antibody toxin scFv(FRP5)-ETA can be safely administered intravenously at doses up to 12.5 μg/kg per day to treat cancer patients with ErbB2-overexpressing tumors. Thereby, serum concentrations of the drug were reached that could be of therapeutic value. The major dose-limiting side effect of scFv(FRP5)-ETA was hepatotoxicity. This may become controllable in future studies by using TNF neutralizing reagents or by temporary suppression of T-cell activation. Whether the development of neutralizing antibodies will limit the therapeutic utility of the antibody toxin or whether alternative treatment schedules or immunosupressant co-medication can overcome this problem remains unclear at present. To further investigate the efficacy of systemic scFv(FRP5)-ETA therapy in cancer patients, we recommend the dose of 12.5 μg/kg for subsequent, carefully planned phase II studies.
Abbreviations
ALT = alanine aminotransferase; AST = aspartate aminotransferase; BSA = bovine serum albumin; ELISA = enzyme-linked immunosorbent assay; ETA = Pseudomonas exotoxin A; mAb = monoclonal antibody; PBS = phosphate-buffered saline; TNF-α = tumor necrosis factor alpha.
Competing interests
G2M Cancer Drugs AG holds rights for commercial development of the study drug scFv(FRP5)-ETA. ABM, SHö and AAb have contributed to the study as employees of G2M AG. WSW is a shareholder of G2M AG. GvM, SHa, EJ, SEAB, SL, AAt, CC, AN, AK, MK and DJ declare that they have no competing interests.
Authors' contributions
GvM, ABM, DJ, AK, MK, EJ, SHa and WSW participated in the design and coordination of the study. GvM, DJ, EJ, SEAB, SL, AAt, CC, AK and MK provided the clinical data. ABM, SHö, SHa, AAb and AN performed and evaluated the assays to determine pharmacokinetic parameters and antibody responses to scFv(FRP5)-ETA. WSW, ABM, SHö, SHa and GvM drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Dr Bernd Groner, Dr Marc Azemar, and Dr Bernd Hentsch for helpful discussions and organizational support, and thank Barbara Uherek for scFv(FRP5)-ETA activity measurements. This work was supported in part by a grant from the National Genome Research Network program of the German 'Bundesministerium für Bildung und Forschung'.
Figures and Tables
Figure 1 (a) Plasma levels of scFv(FRP5)-ETA at steady state. Plasma samples were taken from individual patients at each dose level at the indicated time points before and after infusion of scFv(FRP5)-ETA on treatment day 5. Plasma concentrations were determined by capture ELISA with 1:10 diluted plasma including standard scFv(FRP5)-ETA concentrations for quantification. No scFv(FRP5)-ETA was detected in the predose levels, indicating that no accumulation occurred with a once-daily dose interval. (b) Area under the concentration–time curve (AUC0–5 hours) at steady state on treatment day 5 for the different dose levels. The area under the concentration–time curve was calculated according to the trapezoidal rule from 0 to 5 hours. In accordance with the concentration-time profiles there was a dose-dependent increase of the area under the concentration–time curve, indicating linear pharmacokinetic behavior in the investigated dose range.
Figure 2 Pharmacokinetic parameters for individual patients on treatment day 5. (a) Dose versus peak plasma concentration, Cmax. (b) Dose versus area under the concentration–time curve (AUC).
Figure 3 Development of scFv(FRP5)-ETA-specific antibodies after treatment. Relative levels of scFv(FRP5)-ETA-specific antibodies induced in patients by the treatment were determined by ELISA with plates coated with the antibody toxin. Sera taken on day 0 before treatment, and sera taken at the indicated days after onset of therapy were diluted 1:50 for analysis. The baseline was determined using several human control sera negative for scFv(FRP5)-ETA-specific antibodies (not shown). Dose levels and patients are indicated.
Table 1 Patient characteristics
Patient Age (years), sex Cancer type Stage at diagnosis/surgery Sites of metastasis Prior therapy
N01 61, female Head and neck T2N2bMX Local, intrapulmonal, mediastinal Surgery, radiation, chemotherapy
U01 61, female Breast T2N2M1 Liver, bone Surgery, chemotherapy, herceptin
U02 56, female Breast T4N1bM0 Lymph node, liver, central nervous system, skin, bone Surgery, radiation, chemotherapy, herceptin
U03 68, female Breast T3N1M0 Lymph node, bone Surgery, radiation, chemotherapy, herceptin, hormonal therapy
U04 64, female Breast T3N0M0 Local, skin, bone, intrapulmonal Surgery, chemotherapy, herceptin
U05 71, female Breast T1aN3cM1 Lymph node, skin Surgery, chemotherapy, herceptin
N03 63, female Breast T3N1bM0 Liver Surgery, chemotherapy, herceptin
N04 55, male Transitional cell carcinoma TXNXMX Lymph node Surgery, chemotherapy, herceptin
N05 72, male Prostate T4N3M1 Lymph node, bone, other Surgery, hormonal therapy
N06 63, male Prostate T2bN0M0 Lymph node Surgery, chemotherapy, hormonal therapy
N07 74, female Breast T1N0M0 Liver Surgery, radiation, hormonal therapy
N09 50, female Breast T1N0M0 Lymph node, liver, bone, intrapulmonal, pleural Surgery, chemotherapy
N10 45, female Breast T3bN1bMX Bone Surgery, radiation, chemotherapy, herceptin
N12 69, female Breast T4N3M1 Local, lymph node, skin Chemotherapy, hormonal therapy, immunotherapy
N13 46, female Breast TXNXMX Lymph node, liver, central nervous system, skin, bone Surgery, radiation, chemotherapy, herceptin
N14 82, female Breast TXNXMX Liver, pleural Surgery, hormonal therapy
N15 70, female Breast T2N1MX Lymph node, skin, bone Surgery, radiation, chemotherapy, herceptin
N17 62, male Non small cell lung carcinoma T2cN2cM1 Lymph node, bone, intrapulmonal Patient refused standard therapy
Table 2 Study summary
Patient Dose level (μg/kg) Course of therapy Toxicities ≥ grade 1 Dose-limiting toxicity Neutralizing antibodies Clinical response
N01 2 According to plan GGT grade 2 No No Progression
U01 2 According to plan None No n.d. Progression
U02 2 Stopped on day 10 Cholestasis due to liver metastasisa No n.d. Progression
N03 4 According to plan GGT grade 2 No No Progression
N04 4 According to plan ALT grade 1 No No Progression
N05 4 According to plan Hemoglobin grade 3a No No Progression
N06 10 According to plan ALT grade 2, AST grade 1 No + Stable disease
N07 10 According to plan ALT/AST grade 1, GGT grade 2 No No Progression
U03 10 According to plan Fever and dyspnoeb No ++ n.d.c
N13 12.5 According to plan ALT grade 1, GGT grade 2, AP grade 1 No No Progression
N14 12.5 Stopped on day 8 ALT/AST grade 3, GGT grade 2, LDH grade 1 Yes n.d. n.d.
N15 12.5 According to plan ALT grade 2, AST grade 1, AP grade 2 No + Stable disease
N17 12.5 According to plan ALT/AST grade 2 No No Progression
U04 12.5 According to plan Dyspnoe No No n.d.c
U05 12.5 According to plan None No ++ Progression
N09 20 According to plan ALT/AST grade 2 No +++ Progression
N10 20 Stopped on day 8 ALT grade 4, AST grade 3, GGT grade 2 Yes n.d. n.d.
N12 20 Stopped on day 8 ALT grade 3, AST grade 2 Yes n.d. n.d.c
ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma-glutamyl transferase; n.d., not determined.
aCausal relationship with study drug unlikely.
bPatient U03 developed fever and dyspnoe after therapy on day 23, which was resolved with antibiotics; the patient died on day 40, causal relationship with study drug unlikely.
cClinical signs of activity while on therapy including healing of cutaneous lesion (U03, U04), size reduction of lymph node metastasis (U03), and inflammatory response and softening of large tumor mass (N12).
Table 3 Pharmacokinetics of scFv(FRP5)-ETA
Dose level (μg/kg) Peak concentration range (ng/ml) Peak concentration mean (ng/ml) AUC (ng* hour/ml) Plasma clearance (l/hour) Half-life (hours)
2 n.d. n.d. n.d. n.d. n.d.
4 18–49 39 ± 18 39 ± 1 6.6 ± 0.6 0.55 ± 0.02
10 128–129 129 ± 1 138 ± 64 5.3 ± 2.1 0.74 ± 0.27
12.5 93–204 160 ± 35 178 ± 100 4.9 ± 3.1 0.73 ± 0.30
20 115–307 236 ± 105 326 ± 146 3.8 ± 2.2 0.73 ± 0.22
AUC, area under the concentration–time curve; n.d., not determined.
==== Refs
Olayioye MA Neve RM Lane HA Hynes NE The ErbB signaling network: receptor heterodimerization in development and cancer EMBO J 2000 19 3159 3167 10880430 10.1093/emboj/19.13.3159
Yarden Y Sliwkowski MX Untangling the ErbB signalling network Nat Rev Mol Cell Biol 2001 2 127 137 11252954 10.1038/35052073
Mendelsohn J Baselga J The EGF receptor family as targets for cancer therapy Oncogene 2000 19 6550 6565 11426640 10.1038/sj.onc.1204082
Slamon DJ Leyland-Jones B Shak S Fuchs H Paton V Bajamonde A Fleming T Eiermann W Wolter J Pegram M Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2 N Engl J Med 2001 344 783 792 11248153 10.1056/NEJM200103153441101
Vogel CL Cobleigh MA Tripathy D Gutheil JC Harris LN Fehrenbacher L Slamon DJ Murphy M Novotny WF Burchmore M Efficacy and safety of trastuzumab as a single agent in first-line treatment of HER2-overexpressing metastatic breast cancer J Clin Oncol 2002 20 719 726 11821453 10.1200/JCO.20.3.719
Pastan I FitzGerald D Recombinant toxins for cancer treatment Science 1991 254 1173 1177 1683495
Wels W Groner B Hynes NE Intervention in receptor tyrosine kinase-mediated pathways: recombinant antibody fusion proteins targeted to ErbB2 Curr Top Microbiol Immunol 1996 213 113 128 8815001
Kreitman RJ Recombinant toxins for the treatment of cancer Curr Opin Mol Ther 2003 5 44 51 12669470
Wels W Harwerth IM Mueller M Groner B Hynes NE Selective inhibition of tumor cell growth by a recombinant single-chain antibody-toxin specific for the erbB-2 receptor Cancer Res 1992 52 6310 6317 1358432
Wels W Beerli R Hellmann P Schmidt M Marte BM Kornilova ES Hekele A Mendelsohn J Groner B Hynes NE EGF receptor and p185erbB-2-specific single-chain antibody toxins differ in their cell-killing activity on tumor cells expressing both receptor proteins Int J Cancer 1995 60 137 144 7814146
Harwerth IM Wels W Marte BM Hynes NE Monoclonal antibodies against the extracellular domain of the erbB-2 receptor function as partial ligand agonists J Biol Chem 1992 267 15160 15167 1353079
Iglewski BH Liu PV Kabat D Mechanism of action of Pseudomonas aeruginosa exotoxin A: adenosine diphosphate-ribosylation of mammalian elongation factor 2 in vitro and in vivo Infect Immun 1977 15 138 144 188760
Keppler-Hafkemeyer A Kreitman RJ Pastan I Apoptosis induced by immunotoxins used in the treatment of hematologic malignancies Int J Cancer 2000 87 86 94 10861457 10.1002/1097-0215(20000701)87:1<86::AID-IJC13>3.3.CO;2-9
Schmidt M McWatters A White RA Groner B Wels W Fan Z Bast RC Synergistic interaction between an anti-p185HER-2 Pseudomonas exotoxin fusion protein [scFv(FRP5)-ETA] and ionizing radiation for inhibiting growth of ovarian cancer cells that overexpress HER-2 Gynecol Oncol 2001 80 145 155 11161852 10.1006/gyno.2000.6040
Spyridonidis A Schmidt M Bernhardt W Papadimitriou A Azemar M Wels W Groner B Henschler R Purging of mammary carcinoma cells during ex vivo culture of CD34+ hematopoietic progenitor cells with recombinant immunotoxins Blood 1998 91 1820 1827 9473251
Azemar M Schmidt M Arlt F Kennel P Brandt B Papadimitriou A Groner B Wels W Recombinant antibody toxins specific for ErbB2 and EGF receptor inhibit the in vitro growth of human head and neck cancer cells and cause rapid tumor regression in vivo Int J Cancer 2000 86 269 275 10738256 10.1002/(SICI)1097-0215(20000415)86:2<269::AID-IJC18>3.0.CO;2-8
Wang L Liu B Schmidt M Lu Y Wels W Fan Z Antitumor effect of an HER2-specific antibody-toxin fusion protein on human prostate cancer cells Prostate 2001 47 21 28 11304726 10.1002/pros.1043
Altenschmidt U Schmidt M Groner B Wels W Targeted therapy of schwannoma cells in immunocompetent rats with an erbB2-specific antibody-toxin Int J Cancer 1997 73 117 124 9334818 10.1002/(SICI)1097-0215(19970926)73:1<117::AID-IJC18>3.0.CO;2-D
Maurer-Gebhard M Schmidt M Azemar M Altenschmidt U Stöcklin E Wels W Groner B Systemic treatment with a recombinant erbB-2 receptor-specific tumor toxin efficiently reduces pulmonary metastases in mice injected with genetically modified carcinoma cells Cancer Res 1998 58 2661 2666 9635594
Batra JK Kasprzyk PG Bird RE Pastan I King CR Recombinant anti-erbB2 immunotoxins containing Pseudomonas exotoxin Proc Natl Acad Sci USA 1992 89 5867 5871 1352878
Reiter Y Brinkmann U Jung SH Lee B Kasprzyk PG King CR Pastan I Improved binding and antitumor activity of a recombinant anti-erbB2 immunotoxin by disulfide stabilization of the Fv fragment J Biol Chem 1994 269 18327 18331 7913461
Pai-Scherf LH Villa J Pearson D Watson T Liu E Willingham MC Pastan I Hepatotoxicity in cancer patients receiving erb-38, a recombinant immunotoxin that targets the erbB2 receptor Clin Cancer Res 1999 5 2311 2315 10499598
Azemar M Djahansouzi S Jäger E Solbach C Schmidt M Maurer AB Mross K Unger C von Minckwitz G Dall P Regression of cutaneous tumor lesions in patients intratumorally injected with a recombinant single-chain antibody-toxin targeted to ErbB2/HER2 Breast Cancer Res Treat 2003 82 155 164 14703062 10.1023/B:BREA.0000004371.48757.19
WMA World Medical Association Declaration of Helsinki – Ethical Principles for Medical Research Involving Human Subjects 1964 Helsinki: WMA
Maurer-Gebhard M Schmidt M Azemar M Stöcklin E Wels W Groner B A novel animal model for the evaluation of the efficacy of drugs directed against the ErbB2 receptor on metastasis formation Hybridoma 1999 18 69 75 10211791
FDA 'COSTART': Coding Symbols for Thesaurus of Adverse Reactions Terms 1989 3 Rockville, MD: FDA
Schumann J Angermuller S Bang R Lohoff M Tiegs G Acute hepatotoxicity of Pseudomonas aeruginosa exotoxin A in mice depends on T cells and TNF J Immunol 1998 161 5745 5754 9820556
Onda M Willingham M Wang QC Kreitman RJ Tsutsumi Y Nagata S Pastan I Inhibition of TNF-alpha produced by Kupffer cells protects against the nonspecific liver toxicity of immunotoxin anti-Tac(Fv)-PE38, LMB-2 J Immunol 2000 165 7150 7156 11120846
Olsen NJ Stein CM New drugs for rheumatoid arthritis N Engl J Med 2004 350 2167 2179 15152062 10.1056/NEJMra032906
Yip YL Ward RL Anti-ErbB-2 monoclonal antibodies and ErbB-2-directed vaccines Cancer Immunol Immunother 2002 50 569 587 11807621 10.1007/s002620100226
Gerstmayer B Altenschmidt U Hoffmann M Wels W Costimulation of T cell proliferation by a chimeric B7-2 antibody fusion protein specifically targeted to cells expressing the erbB2 proto-oncogene J Immunol 1997 158 4584 4590 9144470
Keefe DL Trastuzumab-associated cardiotoxicity Cancer 2002 95 1592 1600 12237930 10.1002/cncr.10854
Cho HS Mason K Ramyar KX Stanley AM Gabelli SB Denney DW JrLeahy DJ Structure of the extracellular region of HER2 alone and in complex with the Herceptin Fab Nature 2003 421 756 760 12610629 10.1038/nature01392
Kreitman RJ Wilson WH White JD Stetler-Stevenson M Jaffe ES Giardina S Waldmann TA Pastan I Phase I trial of recombinant immunotoxin anti-Tac(Fv)-PE38 (LMB-2) in patients with hematologic malignancies J Clin Oncol 2000 18 1622 1636 10764422
Kreitman RJ Wilson WH Bergeron K Raggio M Stetler-Stevenson M FitzGerald DJ Pastan I Efficacy of the anti-CD22 recombinant immunotoxin BL22 in chemotherapy-resistant hairy-cell leukemia N Engl J Med 2001 345 241 247 11474661 10.1056/NEJM200107263450402
|
16168106
|
PMC1242130
|
CC BY
|
2021-01-04 16:54:50
|
no
|
Breast Cancer Res. 2005 Jun 1; 7(5):R617-R626
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1264
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12661616811010.1186/bcr1266Research ArticleImpact of intercensal population projections and error of closure on breast cancer surveillance: examples from 10 California counties Phipps Amanda I [email protected] Christina A [email protected] Rochelle R [email protected] Marin County Department of Health and Human Services, Epidemiology Program, San Rafael, California, USA2 Currently working at the Northern California Cancer Center, Fremont, California, USA3 Northern California Cancer Center, Fremont, California, USA2005 7 6 2005 7 5 R655 R660 9 9 2004 23 11 2004 25 4 2005 11 5 2005 Copyright © 2005 Phipps et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
In 2001, data from the California Cancer Registry suggested that breast cancer incidence rates among non-Hispanic white (nHW) women in Marin County, California, had increased almost 60% between 1991 and 1999. This analysis examines the extent to which these and other breast cancer incidence trends could have been impacted by bias in intercensal population projections.
Method
We obtained population projections for the year 2000 projected from the 1990 census from the California Department of Finance (DOF) and population counts from the 2000 US Census for nHW women living in 10 California counties and quantified age-specific differences in counts. We also computed age-adjusted incidence rates of invasive breast cancer in order to examine and quantify the impact of differences between the population data sources.
Results
Differences between year 2000 DOF projections and year 2000 census counts varied by county and age and ranged from underestimates of 60% to overestimates of 64%. For Marin County, the DOF underestimated the number of nHW women aged 45 to 64 years by 32% compared to the 2000 US census. This difference produced a significant 22% discrepancy between breast cancer incidence rates calculated using the two population data sources. In Los Angeles and Santa Clara counties, DOF-based incidence rates were significantly lower than rates based on census data. Rates did not differ significantly by population data source in the remaining seven counties examined.
Conclusion
Although year 2000 population estimates from the DOF did not differ markedly from census counts at the state or county levels, greater discrepancies were observed for race-stratified, age-specific groups within counties. Because breast cancer incidence rates must be calculated with age-specific data, differences between population data sources at the age-race level may lead to mis-estimation of breast cancer incidence rates in county populations affected by these differences, as was observed in Marin County. Although intercensal rates based on population projections are important for timely breast cancer surveillance, these rates are prone to bias due to the error of closure between population projections and decennial census population counts. Intercensal rates should be interpreted with this potential bias in mind.
==== Body
Introduction
From the inception of the Surveillance, Epidemiology, and End Results (SEER) national cancer registry network in 1973, Marin County, California, a small county near San Francisco, has consistently reported higher than average annual incidence rates of breast cancer. Averaged from 1973 to 1999, Marin County reported the highest overall breast cancer incidence rate of the 199 counties included in the SEER database (based on the SEER 9 November 2001 submission released April 2004) [1]. In recent years, reports of rapidly increasing breast cancer rates in Marin County attracted public and media attention. These reports suggested that overall age-adjusted incidence rates of invasive breast cancer in non-Hispanic white (nHW) women living in Marin County had increased approximately 60% between 1990 and 1999, as compared to 5% in surrounding regions (Fig. 1) [2]. These trends have resulted in Marin County having one of the highest incidence rates reported in the world and have prompted public and scientific concern.
Several possible explanations have been suggested for these breast cancer incidence patterns. Overall, the socio-demographic profile of most women living in Marin County would suggest a higher prevalence of women with known risk factors for breast cancer: relatively high proportions of the county population are of nHW ethnicity and college-educated, and the county has a median household income almost double the national average (derived from data from Population estimates, 2000 Census of Population and Housing, 1990 Census of Population and Housing, Small Area Income and Poverty Estimates, County Business Patterns, 1997 Economic Census, Minority- and Women-Owned Business, Building Permits, Consolidated Federal Funds Report, 1997 Census of Governments [3]). In addition, women living in Marin County have fewer children, report a later age at first childbirth, and have higher rates of alcohol consumption than most areas of California, all of which correspond to an increased risk of developing breast cancer [4-6].
Another possible explanation for the observed increase in breast cancer incidence in Marin County during the 1990s involves error in population estimates used in the calculation of cancer rates. Intercensal population estimates, as are used to calculate breast cancer incidence rates for Marin County, are used by a variety of health surveillance organizations nationwide. In order to track changes in the occurrence of health outcomes in a timely manner, disease registries, vital statistics agencies, and local health departments must rely on timely estimates of annual population size; however, for most locales in the United States, the population is counted only once every 10 years as part of the national census. Population estimates for intercensal years are projected from census counts from the most recent decennial census along with other governmental data (e.g., vital statistics and immigration records), and are subject to adjustment after the release of data from the subsequent census. The discrepancy between year 2000 population data from the 2000 census and population projections for the year 2000 based on the 1990 census is known as the 'error of closure'.
In California, intercensal population projections are available from two sources, the US Census Bureau and the California Department of Finance (DOF). Although it is uncertain which agency produces more accurate population projections, most California health agencies rely on data from the DOF, perhaps because the methodology used by the census includes little county-specific information, because significant flaws in Census Bureau-produced estimates have been cited in the past, and because the DOF incorporates additional county- and state-specific information into population projections [7-9]. In order to project the size of the population of California by county, gender, race/ethnicity, and 1 year age increments, for the next 50 years, the DOF not only uses data from the most recent national census, but also enhanced state data resources, such as state records of drivers license change of address transactions, migration patterns based on previous censuses, ethnic group-specific fertility rates, information from the Department of Corrections regarding the capacity and flow of prisoners through facilities, and information from the Pentagon to predict military base closures and reassignments. The DOF also makes adjustments to all intercensal population projections dating back to the previous national census with the release of new national census counts [9,10].
Despite the detailed algorithm used by the DOF to project the distribution and size of the California population, these projections are subject to the same limitations as intercensal population figures generated by the Census Bureau; the risk of inaccuracy increases as annual estimates become more temporally removed from the most recent census. Furthermore, estimates for small areas, or certain age, gender, and racial/ethnic groups are prone to even larger biases due to algorithm inaccuracies: areas with a high growth rate, a large population of retirees, or a large population of foreign-born individuals are likely to be underestimated, while areas with high poverty, and areas with a negative growth rate are likely to be overestimated [7].
The following analysis was conducted to assess the impact on breast cancer incidence rates of the error of closure in stratified DOF projections, 10 years removed from the most recent national census. The goals of this analysis were: to examine how closely population estimates for the year 2000 from the California DOF correlated overall and for selected population strata with counts from the 2000 US census; and to assess how breast cancer incidence rates in selected California counties could be affected by the error of closure between DOF estimates and census counts.
Materials and methods
Data sources and study population
At the time of this analysis, 10 counties in California participated in the National Cancer Institute's SEER program (Alameda, Contra Costa, Marin, San Francisco, San Mateo, Monterey, San Benito, Santa Clara, Santa Cruz, and Los Angeles). Data on incident invasive breast cancers diagnosed between 1999 and 2001 in these counties were accessed with public-use SEER data files (based on the SEER 11 Sub for Expanded Races November 2003 submission released April 2004) [1].
Age, sex, and race/ethnicity-specific population data for the year 2000 for counties under analysis were obtained from the DOF and the US Census Bureau [9,11]. Year 2000 DOF estimates used in these analyses were projected based on the 1990 census and were not adjusted to the 2000 census, although DOF data adjusted to the 2000 census are now available. Year 2000 data from the US Census represent actual year 2000 counts.
Analyses were limited to nHW women. We limited the population to this group to avoid confounding by race/ethnicity; because breast cancer incidence rates are higher among nHW women than among women of any other race, to include breast cancer incidence estimates for Los Angeles County (where only 31% of the population is nHW) and estimates for Marin County (where 79% of the population is nHW) in the same analysis, without accounting for race, could be misleading [5]. There are, however, important differences in the way the DOF and the census categorize race/ethnicity. The 2000 US Census allowed individuals to report up to six distinct ethnicities concurrently to categorize themselves, whereas the DOF stratified the population into five mutually exclusive race categories (white, Hispanic, African-American, Asian Pacific Islander, and American Indian) [10,11]. To control for these differences in race categorization, a US census dataset with bridged race categories was used [12].
Comparison of population data
Year 2000 population data from the DOF and the census were compared overall and stratified by county, gender, and age group. We examined these stratified groups in order to identify and describe those most likely to be impacted by discrepancies in population estimates.
For all comparisons, 2000 census data were chosen as the standard. Percent differences between data sources can thus be interpreted as the percent by which DOF estimates overestimate or underestimate corresponding census counts. Very small percent differences are to be expected due to the fact that census estimates are based on the population as of 1 April 2000, whereas DOF estimates are based on the population as of the middle of the year [10,11].
Analysis of incidence rates
For the comparison of breast cancer incidence rates, we included cases of invasive breast cancer (classified as 50.0–50.9 by the International Classification of Diseases; Oncology, 2nd edition) diagnosed between the years 1999 and 2001 [13]. County-specific incidence rates were age-adjusted using direct-standardization methods, adjusting to the 2000 US standard population [14]. Incidence rates based on year 2000 DOF population estimates, and their corresponding 95% confidence intervals, were compared to incidence rates based on year 2000 census counts for each county under analysis.
Standardized rate ratios were calculated to compare DOF-based and census-based incidence rates for all counties under analysis. Census-based rates were used as the reference in all regression models, such that rate ratios derived from each of the 10 county-specific models describe the influence of discrepancies between rate denominators independent of the rate numerator.
Results
Overall, the DOF estimated the size of the year 2000 California population to be 2.3% larger than was counted by the census. When restricted to nHW women in the 10 counties under analysis, DOF population estimates exceeded census counts by approximately 2.9%, ranging from 5.7% below census counts in San Francisco County to 10.4% above census counts in San Benito County. Table 1 summarizes discrepancies between population data sources by age strata and county.
When further stratified by age group, DOF and census county population data for nHW women differed more significantly, although patterns of overestimation and underestimation across age strata differed by county. Percent differences between the two population data sources ranged from <0.1% to 64.1%. Discrepancies by age group were largest in San Francisco County, where the percent difference between DOF and census data was more than 30% in 4 of 10 age groups for nHW women (ranging from -60.3% to 64.1%), and in Marin County, where percent differences also exceeded 30% in 4 of 10 age groups (ranging from -32.9% to 41.4%). More importantly, however, were discrepancies between population estimates for age groups with the highest incidence of breast cancer; among nHW women aged 45 years and older, the most substantial population data discrepancies were in Marin County, where DOF population projections for nHW women aged 45 to 64 years fell below census estimates by approximately 31.7% and in San Benito County, where DOF estimates for the 55 years and older population exceeded census estimates by 30.6%. Age-specific discrepancies for Marin County are plotted in Fig. 2.
Direct comparison of year 2000 DOF- and census-based age-adjusted incidence rates by county revealed significant differences in estimated incidence rates by population source in three of the ten counties under analysis (Table 2). Breast cancer incidence rates in Santa Clara and Los Angeles counties were significantly lower when based on DOF county population estimates compared to census county population data, adjusting for age: the DOF-based rate was 143.4, 95% CI = (137.5–149.5) versus the census-based rate of 158.6, 95% CI = (152.0–165.4) in Santa Clara; and the DOF-based rate was 153.8, 95% CI = (150.7–156.9) versus the census-based rate of 161.0, 95% CI = (157.8–164.3) in Los Angeles. Marin County was the only county where the DOF-based rate was significantly higher than the census-based rate: the DOF-based rate was 213.6, 95% CI = (198.4–229.9) versus the census-based rate of 175.8, 95% CI = (163.2–189.5)). The DOF-based rates for Marin County were approximately 22% higher than census-based rates based on the same numerators.
Discussion
These analyses have explored the extent to which use of intercensal population projections, extrapolated and 10 years removed from the 1990 census, may have biased breast cancer incidence rates reported in California in the 1990s. DOF-based incidence rates for Marin, Santa Clara, and Los Angeles counties were found to differ significantly from census-based incidence rates: county-specific DOF-based rates were lower than census-based rates in Santa Clara and Los Angeles counties, but higher than census-based rates in Marin County.
Direct comparison of year 2000 DOF and census population data revealed accuracy of DOF projections at the state level, although joint stratification of population estimates by county, gender, race/ethnicity, and age introduced greater discrepancy between population data sources. These discrepancies between stratified population estimates were significant enough to lead to notable differences in breast cancer incidence rates, and can be expected to have a notable effect on other statistics based on DOF intercensal estimates not adjusted to the 2000 census. For example, in the case of San Francisco County, substantial overestimation of the 5–14 year old nHW female population (64.1%) by the DOF compared to the 2000 US census, while having a negligible effect on county breast cancer incidence rates due to the negligible rate of breast cancer among this age group, may be anticipated to have a notable impact on childhood cancer rates.
The fact that differences between population data sources had a significant effect on breast cancer incidence rates in the two largest counties analyzed (Los Angeles, population 9,519,338, and Santa Clara, population 1,682,585) and the second smallest county analyzed (Marin, population 247,289) suggests that population size is not responsible for variation between census and DOF data. Indeed, no pattern of deviation between the two population data sources is discernable by county size, age distribution, or county urban/rural status. It is possible that methods used by the census and methods employed by the DOF are differentially effective among different populations, or that differing levels of domestic migration explains the discrepancies between these population data sources. The source of these discrepancies, however, remains unknown and was beyond the scope of this analysis.
One limitation to the applicability of this analysis is that DOF intercensal population projections are not used as widely as projections provided by the US census, as DOF projections are only available for counties in the state of California; however, problems similar to those noted in this analysis have been noted with the application of census projections [7]. Errors in intercensal population projections provided by the US census for the years 1991 to 1999 were recently implicated as a source of significant overestimation of racial disparities in cancer incidence rates [7,15]. In Marin County, US census intercensal population projections were subject to error of closure problems similar to those identified in DOF projections; census projections, unadjusted to the 2000 US census, substantially underestimated the high risk group of Marin County nHW women aged 45 to 74 years (data not shown), resulting in an overestimation of the overall incidence rate of breast cancer in the latter years of the 1990s (Fig. 3). Thus, it is likely that similar conclusions would have been reached had census intercensal projections been used rather than DOF intercensal projections. Although methodology used to generate intercensal population projections by both the DOF and the Census Bureau is intricate, complex, and complete, both agencies have produced inaccurate projections. These error of closure problems mean that population data, and incidence rates based on these data, become less reliable as they become further removed from the most recent census.
Conclusion
The California Cancer Registry, as well as county and local governmental agencies and a broad range of community organizations, must rely on intercensal population projections to estimate health trends, allocate resources, and establish priorities with respect to the populations they serve. Timely surveillance requires that intercensal population projections be used to generate population-based rates and trends as soon as reliable incidence counts become available. This analysis, however, demonstrates that intercensal population projections can differ substantially from later decennial census counts. Although it is unrealistic to recommend that disease surveillance be paused in intercensal years, these data remind us that population denominator quality can have a major impact on the interpretation of health statistics. Health agencies must judge whether aberrant health trends should be acted upon prior to the release of population information that could inform the accuracy of population projections, a process that could take five years or more. The gravity of this problem is magnified in the case of diseases like breast cancer that are the focus of public concern and activism, which intensifies demand for information and public health action.
The results of this analysis support the need for a restructuring of population estimation procedures; perhaps more frequent collection of population counts, particularly in regions experiencing high levels of migration. A 10-year period between population censuses is problematic for accurate projection of the age/gender/race-specific yearly population counts needed for health tracking. Alternatively, government agencies producing population projections would benefit from improvements in ways to make more accurate assumptions regarding the growth and distribution of the population. At the least, more health agencies should develop better ways to describe and quantify the uncertainties in population projections and related bias to consumers of health statistics.
Abbreviations
DOF = Department of Finance; nHW = non-Hispanic White; SEER = Surveillance, Epidemiology, and End Results;
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AIP completed all analyses and led the writing. R.R. Ereman was the supervising researcher on this project, helped conceive of the study idea and supervised the writing process. C.A. Clarke assisted with the conception of the study analysis and advised the writing. All authors helped review drafts of the manuscript and interpret study findings.
Acknowledgements
The authors would like to thank Lee Ann Prebil for her comments and helpful assistance, Dr. Steve Selvin for his assistance, and Jennifer Welle and Michael Musante for their assistance in the analysis.
Figures and Tables
Figure 1 Breast cancer incidence in Marin County, California, and surrounding areas based on Department of Finance population projections for 1990 to 1999. Rates are age-adjusted to the 2000 US standard population, based on population data unadjusted to the 2000 US census. **Case data from the California Cancer Registry.
Figure 2 Marin County, California, population estimates for non-Hispanic white women by age and population data source (2000). DOF, Department of Finance.
Figure 3 Breast cancer incidence in Marin County, California, by population data source, and adjustment 2000 US census, for 1992–1999.
Table 1 Percent difference between DOF and census population estimates for non-Hispanic white women by age (2000)a
Age (years) Alameda Contra Costa Marin San Francisco San Mateo Monterey San Benito Santa Clara Santa Cruz Los Angeles
All ages 3.4 3.8 -1.4 -5.7 3.8 4.0 10.4 8.7 4.7 2.1
Under 5 8.5 -1.1 -5.1 23.2 1.5 19.2 6.0 5.2 11.5 0.8
5–14 13.2 <0.1 -1.5 64.1 9.2 27.7 -6.7 10.4 10.8 4.1
15–24 -3.7 26.8 40.0 -54.9 16.2 5.3 41.7 6.0 -3.8 -12.8
25–34 -19.5 -0.9 41.4 -60.3 -21.9 -11.4 15.4 -4.4 1.7 -11.1
35–44 9.2 -4.4 17.3 40.7 2.3 14.4 -8.0 14.8 4.3 13.6
45–54 10.9 4.7 -31.1 27.0 8.0 -2.5 -0.6 12.7 5.5 9.1
55–64 13.9 13.1 -32.9 11.4 14.5 -3.8 24.0 20.1 7.6 9.1
65–74 3.9 9.7 5.9 7.7 10.6 -2.3 47.0 13.0 8.8 -0.3
75–84 -4.4 -4.8 -5.8 9.1 2.6 -3.8 24.6 -2.1 5.2 -3.0
85+ -4.9 -10.3 -24.7 20.8 -2.8 -1.5 25.0 -14.5 3.3 2.8
aPercent difference = ([DOF - census]/census) × 100. DOF, Department of Finance.
Table 2 DOF-based versus census-based breast cancer incidence rates among non-Hispanic white women (1999–2001)
County Average cases per year DOF-based incidence rate (95% CI)a Census-based incidence rate (95% CI)a Standardized rate ratio
Alameda 575 143.3 (136.6–150.4) 151.8 (144.6–159.2) 0.94
Contra Costa 592 158.0 (150.7–165.7) 164.9 (157.2–172.9) 0.96
Marin 245 213.6 (198.4–229.9) 175.8 (163.2–189.5) 1.22b
San Francisco 296 140.3 (131.0–150.7) 159.8 (149.2–171.0) 0.88
San Mateo 406 152.3 (143.7–161.4) 163.4 (154.2–173.2) 0.93
Monterey 172 152.7 (139.7–166.9) 150.7 (137.7–164.8) 1.01
San Benito 19 114.0 (85.9–149.0) 138.7 (104.6–182.0) 0.82
Santa Clara 743 143.4 (137.5–149.5) 158.6 (152.0–165.4) 0.90
Santa Cruz 161 148.2 (135.1–162.4) 157.8 (143.8–173.0) 0.94b
Los Angeles 3587 153.8 (150.7–156.9) 161.0 (157.8–164.3) 0.96b
aIncidence rates are expressed per 100,000 women based on year 2000 population denominators. Data are adjusted to the 2000 US standard. bp < 0.05. DOF, Department of Finance.
==== Refs
Surveillance, Epidemiology, and End Results (SEER) Program Public-Use Data
Northern California Cancer Center Data summary of Marin County breast cancer incidence rates
US Census Bureau State and County Quickfacts
Marin County Department of Health and Human Services Marin Community Health Survey 2001
Kelsey JL Bernstein L Epidemiology and prevention of breast cancer Annu Rev Public Health 1996 17 47 67 8724215 10.1146/annurev.pu.17.050196.000403
Clarke CA Glaser SL West DW Ereman RR Erdmann CA Barlow JM Wrensch MR Breast cancer incidence and mortality trends in an affluent population: Marin County, California, USA, 1990–1999 Breast Cancer Res 2002 4 R13 12473174 10.1186/bcr458
Boscoe FP Miller BA Population estimation error and its impact on 1991–1999 cancer rates Prof Geographer 2004 56 516 529
Rosenwaike I Yaffe N Sagi PC The recent decline in mortality of the extreme aged: An analysis of statistical data Am J Public Health 1980 70 1074 1080 6998306
State of California, Department of Finance Projected Total Population of California Counties 1990 to 2040 Report 93 P-3 1993 Sacramento, California
State of California, Department of Finance County Population Projections with Age, Sex and Race/Ethnic Detail 1998 Sacramento, California
US Census Bureau Racial and Ethnic Classifications Used in Census 2000 and Beyond
National Center for Health Statistics Bridged-race population estimates for April 1, by county, single-year of age, bridged-race, Hispanic origin, and sex 2000
Young JL JrRoffers SD Ries LAG Fritz AG Hurlbut AA (eds) SEER Summary Staging Manual – 2000 Codes and Coding Instructions National Cancer Institute, NIH Pub No 01-4969, Bethesda, MD 2001
Breslow NE Day NE Statistical Methods in Cancer Research The Design and Analysis of Cohort Studies 1996 2 New York: Oxford
Faulty estimates led NCI to overstate Black-White cancer disparity in Atlanta The Cancer Letter 2002 28 1 5
|
16168110
|
PMC1242131
|
CC BY
|
2021-01-04 16:54:50
|
no
|
Breast Cancer Res. 2005 Jun 7; 7(5):R655-R660
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1266
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12731616811210.1186/bcr1273Research ArticleEpigenetic silencing of DSC3 is a common event in human breast cancer Oshiro Marc M [email protected] Christina J [email protected] Ryan J [email protected] Damian J [email protected]ñoz-Rodríguez José L [email protected] Jeanne A [email protected] Matthew [email protected] Sangita C [email protected] Anne E [email protected] Frederick E [email protected] Bernard W [email protected] Departments of Pharmacology and Toxicology, Arizona Cancer Center, University of Arizona, Tucson, AZ, USA2 Department of Surgery, Arizona Cancer Center, University of Arizona, Tucson, AZ, USA3 Department of Cell Biology and Anatomy, Arizona Cancer Center, University of Arizona, Tucson, AZ, USA4 Department of Radiation Oncology, Free Radical and Radiation Biology Program, Holden Comprehensive Cancer Center, University of Iowa, Iowa City, IA, USA2005 16 6 2005 7 5 R669 R680 16 3 2005 4 5 2005 10 5 2005 23 5 2005 Copyright © 2005 Oshiro et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Desmocollin 3 (DSC3) is a member of the cadherin superfamily of calcium-dependent cell adhesion molecules and a principle component of desmosomes. Desmosomal proteins such as DSC3 are integral to the maintenance of tissue architecture and the loss of these components leads to a lack of adhesion and a gain of cellular mobility. DSC3 expression is down-regulated in breast cancer cell lines and primary breast tumors; however, the loss of DSC3 is not due to gene deletion or gross rearrangement of the gene. In this study, we examined the prevalence of epigenetic silencing of DSC3 gene expression in primary breast tumor specimens.
Methods
We used bisulfite genomic sequencing to analyze the methylation state of the DSC3 promoter region from 32 primary breast tumor specimens. We also used a quantitative real-time RT-PCR approach, and analyzed all breast tumor specimens for DSC3 expression. Finally, in addition to bisulfite sequencing and RT-PCR, we used an in vivo nuclease accessibility assay to determine the chromatin architecture of the CpG island region from DSC3-negative breast cancer cells lines.
Results
DSC3 expression was downregulated in 23 of 32 (72%) breast cancer specimens comprising: 22 invasive ductal carcinomas, 7 invasive lobular breast carcinomas, 2 invasive ductal carcinomas that metastasized to the lymph node, and a mucoid ductal carcinoma. Of the 23 specimens showing a loss of DSC3 expression, 13 (56%) were associated with cytosine hypermethylation of the promoter region. Furthermore, DSC3 expression is limited to cells of epithelial origin and its expression of mRNA and protein is lost in a high proportion of breast tumor cell lines (79%). Lastly, DNA hypermethylation of the DSC3 promoter is highly correlated with a closed chromatin structure.
Conclusion
These results indicate that the loss of DSC3 expression is a common event in primary breast tumor specimens, and that DSC3 gene silencing in breast tumors is frequently linked to aberrant cytosine methylation and concomitant changes in chromatin structure.
==== Body
Introduction
Aberrant cytosine methylation of CpG dinucleotides in the promoter region of genes is often associated with changes in their chromatin structure and transcriptional silencing of the gene during carcinogenesis and tumor progression [1-7]. Cytosine methylation has been shown to play a fundamental role in breast tumor progression, as silenced genes have been identified that fall into each of the six 'acquired capabilities of cancer' as described by Hanahan and Weinberg [8]. Targeted genes include regulators of cell cycle, maintainers of genomic integrity, tumor suppressors, as well as adhesion molecules [9]. Examples of hypermethylated genes in breast cancer include: maspin, E-cadherin, BRCA1, ras association domain family 1A, tissue inhibitor of metalloproteinase-3, and A Disintegrin And Metalloprotease domain 23 gene [1,10-17].
Desmosomes, together with adherens junctions, represent the major adhesive cell-junctions of epithelial cells [18-20]. E-cadherin is one example of an integral component of adherens junctions whose role in breast tumor progression has been clearly established [10,13,21]. The participation of desmosomal components in cancer, however, is enigmatic. Desmosomes are multifaceted intracellular junctions that participate in cell adhesion and maintenance of normal tissue structure in the epidermis [20,22]. Desmocollins (DSCs) are members of the cadherin superfamily, and fundamental members of the desmosome. DSC family members are uniquely expressed in epidermal tissue, with DSC2 being expressed in all desmosome-bearing tissues, while DSC1 and DSC3 expression is restricted to certain specialized epithelia, mainly stratified squamous epithelia [23,24]. Furthermore, DSC1 is expressed in the higher terminally differentiated cell layers, while DSC3 is mainly expressed in the basal layers [23,24]. In addition, the DSCs are present in both 'a' and 'b' isoforms, resulting from the alternate splicing of exon 16 [25]. They differ with respect to their carboxy-terminal end, with the 'b' form having a shortened carboxy-terminal domain that removes the major binding site for plakoglobin [26].
One of the more intriguing functions of desmosomal proteins as they relate to cancer is their ability to inhibit cell motility. Notably, Tselepis et al. [27] showed that the expression of multiple desmosomal components (DSC, desmoglein, and plakoglobin) were sufficient to induce adherence of the normally non-adherent invasive L929 fibroblast. This induced adhesion could be blocked by the addition of short peptides corresponding to the putative cell adhesion recognition sites of DSC and desmoglein. In addition, the introduction of these desmosomal proteins was also sufficient to inhibit L929 invasion into collagen gels. Recently, functional studies that targeted the inhibition of DSC3 with dominant negative constructs showed that DSC3 expression is required for the formation of desmosomes and adherens junctions [28]. In total, these studies support the idea that intact desmosomes can inhibit cellular motility.
Down-regulation of DSC3 in breast cancer was first reported by Klus [29]. They showed that DSC3 was expressed in normal breast while its expression was down-regulated in both primary breast tumors and breast tumor cell lines. We recently performed two-color fluorescence cDNA microarray experiments to identify p53 response genes in human breast tumor cell lines [6]. Our results identified DSC3 as a p53 response gene whose expression was downregulated in 80% of breast tumor cell lines tested. In addition, analysis of breast cancer cell lines showed that DSC3 is silenced in association with cytosine hypermethylation and histone deacetylation [6]. Therefore, the loss of DSC3 expression in the cell lines appears to be due to both epigenetic and genetic changes.
In this study, we extend our in vitro analysis of DSC3 to the investigation of the frequency of epigenetic silencing of DSC3 expression in primary breast tumor specimens. DNA from freshly isolated tumor specimens was analyzed for cytosine methylation by sodium bisulfite sequence analysis while RNA from the same tumors was analyzed by quantitative real-time RT-PCR for DSC3 expression. Our results show that epigenetic silencing of DSC3 is a common event in primary breast tumor specimens, as 72% of breast carcinomas analyzed showed a loss of DSC3 expression and that the loss of expression strongly correlated with cytosine methylation of its promoter region in 56% of DSC3-negative breast carcinomas analyzed, and 41% of all specimens analyzed.
Concurrently, we analyzed a panel of breast tumor cell lines for DSC3 expression and concomitant cytosine methylation of its promoter region. As aberrant cytosine methylation of promoter regions is associated with alterations to chromatin structure, we also compared the in vivo nuclease accessibility of the DSC3 promoter region in normal and tumor breast cell lines. Our results indicate that the loss of DSC3 is a common event in breast tumor cell lines at both the mRNA and protein levels and that the loss of expression is frequently correlated with cytosine methylation of its promoter region and an inaccessible chromatin structure. These results indicate that epigenetic silencing of DSC3 is an underlying event in breast tumorigenesis.
Materials and methods
Cell culture and manipulations
Normal human mammary epithelial cells (HMECs) and human prostate epithelial cells were obtained from Clonetics (San Diego, CA, USA), fetal skin keratinocytes from Cell Applications (San Diego, CA, USA); these were grown according to manufacturers' instructions. Human foreskin fibroblasts were maintained and cultured in the Arizona Cancer Center Cell Culture Shared Service (Tucson, AZ, USA). Peripheral blood lymphocytes were obtained from the whole blood of healthy donors in accordance with the health insurance portability and accountability act of 1996 (HIPAA) guidelines. Briefly, whole blood was collected into BD Vacutainer CPT cell preparation tubes containing sodium heparin (Becton Dickinson, Franklin Lakes, NJ, USA) and processed according to the manufacturer's protocol. Primary cultures of normal human oral keratinocytes were established and maintained in short-term culture as described [30-32]. Primary cultures of human airway epithelial cells were obtained by enzymatic digestion of bronchial samples from lung transplants and maintained in short-term culture as described [33,34].
The MCF10A, MDA-MB-453, MDA-MB-435, MDA-MB-231, MDA-MB-157, MDA-MB-468, BT549, ZR-75-1, and HS578T breast cancer cells were obtained from the American Type Culture Collection (Rockville, MD, USA). The HaCaT cells, a normal immortalized keratinocyte cell line [35], were obtained from Norbert E Fusenig (German Cancer Research Center, University of Heidelberg, Heidelberg, Germany). The early passage sporadic breast cancer cell lines UACC1179, UACC2087, UACC893, UACC3133, UACC3199, and UACC2648 were developed and maintained at the Arizona Cancer Center Cell Culture Shared Service.
Breast tumor specimens
Thirty flash frozen breast cancer tissue specimens were obtained from patients who underwent surgery for breast cancer, either lumpectomy or mastectomy, at the University Medical Center in Tucson, AZ, from 2003 to 2004. All patients signed surgical and clinical research consents for tissue collection in accordance with the University of Arizona Institutional Review Board and HIPAA regulations. At the time of surgery, a 1–3 cm section of the tumor was immediately snap frozen in liquid nitrogen and stored in our prospective breast tissue bank at -80°C. From each tissue block, a series of 5 micron sections were cut and stained with hematoxylin and eosin (H&E) for pathological evaluation. All of the H&E slides were reviewed by one breast pathologist to determine the integrity of the tumor specimen and this was correlated with the clinical pathologic review performed by an independent pathologist.
Nucleic acid isolation
Total RNA was isolated from cells using an RNeasy® Mini or Midi Kit (Qiagen, Valencia, CA, USA), and genomic DNA was isolated using the QIAamp DNA Mini Kit (Qiagen, Valencia, CA). Isolation of RNA from frozen breast tumor specimens was done as follows: 30–50 μg of tissue was disrupted in a 1.5 ml RNAase free tube with an RNAase free Pellet Pestle (Kimble-Kontes, Vineland, New Jersey, USA) then passed through a 21 gauge needle to homogenize the sample. Following homogenization, RNA was isolated using the RNeasy® Mini kit. Isolation of DNA from frozen breast tumor specimens was done using a Medimachine (BD Biosciences, San Jose, CA, USA). Briefly, a 50 μm Medicon (BD Biosciences) was washed twice with 1 ml of TKM1-NP buffer (10 mM Tris-HCl, pH 7.6, 10 mM KCl, 10 mM MgCl2, 2 mM EDTA, and 2.5 μl/ml NP40) then a 30–50 μg piece of frozen breast tumor tissue was further cut into 3–6 mm3 pieces then placed into the Medicon filled with 1 ml of TKM1-NP buffer. The tissue was disaggregated for 30 s then allowed to rest for 20 s and then disaggregated for another 30 s. The cell suspension was then passed through 100 μm Filcon (BD Biosciences) into a 15 ml conical tube. Disaggregation in the Medicon was repeated four to six more times to completely disaggregate the tissue. Once done, the cell suspension was spun down for 10 minutes at 250 × g at 4°C. Completion of DNA isolation was done using the Qiagen DNA mini kit, Tissue Protocol. RNA and DNA samples were quantified by UV absorbance measurements at 260 nm. Furthermore, all breast tumor specimen RNAs were run out on an Agilent RNA Labchip (Agilent Technologies, Waldbronn, Germany) for quantitative and qualitative assessment of the RNA.
Sodium bisulfite genomic sequencing of the DSC3 promoter
Genomic DNA (5 μg) was modified with sodium bisulfite under conditions previously described [1]. The DSC3 promoter was amplified from the bisulfite-modified DNA by two rounds of PCR using nested primers specific to the bisulfite-modified sequence of the DSC3 CpG Island. First round primers were: UTDSC3_F1, GATTGGGGTTTTGTATTGAGA; UTDSC3_R1, TTAACCTCTCTCAAACTTACC. Second round primers were: UTDSC3_F2, ATTTGGGTTGTTAGGGTTTTTTT; UTDSC3_R2, AAAACAACTTCACTTCTAAAACC. Both rounds of PCR were performed under the same parameters, with 1% of the first round PCR product serving as the template in the second round of PCR. PCR amplification was performed under the following conditions: 94°C for 4 minutes followed by 5 cycles of 94°C for 1 min, 56°C for 2 min, 72°C for 3 min, then 35 cycles of 94°C for 30 s, 56°C for 2 min, 72°C for 1.5 min, and ending with a final extension of 72°C for 6 min.
The resultant PCR product was cloned into a TA vector according to the manufacturer's instructions (pGEM-T-Easy cloning kit; Promega, Madison, WI, USA)). Ten positive recombinants were isolated using a Qiaprep Spin Plasmid Miniprep kit (Qiagen) according to the manufacturer's instructions and sequenced on an ABI automated DNA sequencer (Applied Biosystems, Foster City, CA, USA). The methylation status of individual CpG sites was determined by comparison of the sequence obtained with the known DSC3 sequence. The number of methylated CpGs at a specific site was divided by the number of clones analyzed (minimum of 10 in all cases) to yield a percent methylation for each site.
Western blot
Cells were lysed by incubating on ice for 2 minutes in RIPA buffer (1 × PBS containing 1% NP40, 0.5% deoxycholate, and 0.1% SDS) with 1 mM phenylmethylsulfonyl fluoride (Boehringer Mannheim Corp, Indianapolis, IN, USA) added directly before use. Samples were sonicated using five pulses of 1 s each. Protein concentration was determined by BCA (bicinchoninic acid) assay (Pierce, Rockford, IL, USA). Whole cell lysates of 20 μg of total protein were diluted in 2 × non-reducing sample buffer and then boiled for 3 minutes before loading onto a 7.5% polyacrylamide gel for analysis. Proteins resolved in the gel were electrotransferred to Millipore Immobilon-P PVDF membrane (Millipore, Bedford, MA, USA). The membranes were blocked in 5% non-fat milk in TBST (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% Tween 20). Primary mouse monoclonal antibody anti-Desmocollin Clone Dsc3-U114 (Research Diagnostics Inc., Flanders, NJ, USA) was diluted 1:10 in 5% non-fat milk/TBST and incubated with the membrane overnight at 4°C. Membranes were washed three times with TBST, incubated with a donkey anti-mouse horseradish peroxidase-conjugated secondary antibody (Chemicon International, Temecula, CA, USA), washed six more times with TBST, visualized with an ECL Western Blotting Detection Kit (Amersham Biosciences, Piscataway, NJ USA), and detected with BioMax MR film (Kodak, Rochester, NY, USA).
Quantitative real time RT-PCR
For real time quantitative RT-PCR analysis of DSC3 and GAPDH gene expression, a reverse transcription step was performed using TaqMan® Reverse Transcription Reagents (Roche Molecular Systems, Branchburg, NJ, USA) and 250 ng of total RNA in a 50 μl reaction. The reverse transcription reaction was primed with random hexamers and incubated at 25°C for 10 minutes followed by 48°C for 30 minutes, 95°C for 5 minutes and a chill at 4°C. For the PCR reaction, 10 ng of cDNA was used in accordance with the protocol outlined in the ABI user manual (Applied Biosystems). DSC3 and GAPDH primer probes were purchased from ABI Assays-on-demand (Fwd Primer, CCAATCCGGTTTCAGAAGTGA; Rev Primer, CTCGCCGCTGCTTGTTTT; FAM Probe, CTCTCTCAGGCTTGCC) were used and data collected using the ABI Prism 7000 real-time sequence detection system (Applied Biosystems). Differences in expression were determined using the comparative Ct method described in the ABI user manual (Applied Biosystems).
Chromatin accessibility assays
Chromatin accessibility assays were performed as previously described (Oshiro et al. [6]) with minor modifications. Ten million cells were washed twice with ice cold 1 × PBS, gently scraped and collected by centrifugation. Nuclei were extracted by resuspension of cells in ice cold 1 × RSB (10 mM Tris HCl, pH 8, 3 mM MgCl2, 10 mM NaCl, 0.05% NP40). The nuclei were collected by centrifugation, resuspended in appropriate 1 × restriction endonuclease buffer, and divided into two aliquots of 200 μl/aliquot. MspI (0 or 75 units; Gibco BRL, Bethesda, MD, USA) was added to the nuclei and incubated at 37°C for 15 minutes. Genomic DNA was isolated using the QIAamp DNA Mini Kit (Qiagen) and ligated to linkers specific for the MspI ends. The linker 'marks' accessible sites of chromatin, and acts as the primer sequence for PCR along with the DSC3 promoter specific primer. Primers were: linker specific primer, GGATTTGCTGGTGCAGTACT; first round gene specific primer, CCTAAATCCCTTTTCAAGTCT; second round gene specific primer, CTCAAAACAAAAAGCTCAGTCCAGA. To increase specific amplification of our band of interest, a second round of PCR was performed using a 1:1000 dilution of the first round PCR product, adding a second, nested primer that was specific for the genomic region being analyzed and internal to the first region-specific primer.
First round PCR was performed using RTG PCR beads (Pharmacia, Piscataway, NJ, USA) to amplify 100 ng of linkered DNA. The initial step in the first round PCR reaction was a 15 minute incubation at 72°C followed by a denaturation at 95°C for 2 minutes then 25 cycles of 95°C for 30 s, 55°C for 1 min, 72°C for 2 s and a final extension at 72°C for 5 minutes. The second round of PCR was performed using the ABI Prism 7000 real-time sequence detection system (Applied Biosystems). For the nested PCR step, 25 pmol (1 μl) of internal DSC3 specific primer was added to 5 μl of diluted first-round product (1:1000), 19 μl of PCR water and 25 μl of 2 × SYBR® Green PCR Master Mix (Applied Biosystems). The PCR conditions for this second round of PCR were as follows: a 10 minute denaturation at 95°C and 40 cycles of 94°C for 1 min, 56°C for 40 s and 72°C for 30 s. Relative levels of chromatin accessibility were determined using the comparative Ct method. Real-time PCR products were also separated on a 3% TBE agarose gel to verify the presence of a single PCR product of the appropriate size (241 bp).
Results
Relative levels of DSC3 mRNA expression from a panel of normal tissue RNA were determined by quantitative real time RT-PCR analysis. DSC3 mRNA levels were normalized to the ubiquitously expressed GAPDH gene, then expression values reported relative to the primary HMEC line expression (Fig. 1). DSC3 expression was limited to certain epithelial cell types, including those of the airway, breast, skin, prostate, and mouth. DSC3 was undetectable in the following non-epithelial cell types: skin fibroblasts, lymphocytes, bone marrow, heart, and kidney. Therefore, expression analysis of DSC3 shows a significant cell type specific pattern of expression, which is limited to cells of epithelial origin, including breast epithelium.
To confirm and extend previous studies [6,29], we analyzed 32 frozen breast cancer specimens from patients who underwent lumpectomy or mastectomy randomly obtained from patients who underwent surgery at the University Medical Center in Tucson, AZ. Our data set comprised 32 specimens: 24 invasive ductal carcinomas (IDC), two of which are metastatic IDCs isolated from patients' lymph nodes, seven invasive lobular carcinomas (ILCs), and one mucoid ductal carcinoma. Incidentally, we received two independent tumors from one diseased breast both of which were IDC specimens. From these specimens, total RNA was collected and DSC3 expression was analyzed by quantitative real time RT-PCR with expression levels of the tumor samples being normalized to HMECs. DSCs are expressed in both 'a' and 'b' isoforms as a result of alternate splicing of exon 16. The ABI probe used in these studies spans exon1 and exon2, which are both conserved in the DSC3a and DSC3b isoforms, allowing us to analyze both isoforms in the specimens tested. DSC3 expression is reduced to less than 10% of the expression seen in HMEC in 18 of 24 (75%) of the IDCs, 5 of 7 (71%) ILCs, and in the mucinous carcinoma (Table 1). The 10% cutoff was chosen to address the potential of reduced expression of DSC3 due to contaminating stromal, non-epithelial elements. The majority of specimens analyzed consisted of 50% tumor based on pathology examination. Thus, the greatly reduced expression of DSC3 is a common event in primary breast tumor specimens.
We next examined the cytosine methylation profiles of the breast tumor specimens to see if loss of expression correlates with cytosine methylation of the promoter region (Table 1, Fig. 2b). The DSC3 promoter region meets the criteria of a CpG island based on size, GC content, CpG dinucleotide frequency, as well as its location with respect to the transcriptional unit (Fig. 2a). We used sodium bisulfite genomic sequencing to assess the cytosine methylation status of 24 CpG dinucleotides within the DSC3 promoter region upstream of the DSC3 transcriptional start site. The region analyzed consists of the p53 binding site, the minimal promoter region, and 75 to 100 bases immediately 5' of the minimal promoter region [6,36]. Ten to twelve cloned PCR products were sequenced to determine the percent methylation of the 24 CpG sites in the 5' promoter region. Of the 18 IDC samples that showed a loss of DSC3 expression, 10 (56%) of these specimens contained methylated cytosines within the CpG island. In the eight remaining IDC specimens that lack DSC3 expression we predict that other mechanisms of silencing such as mutation to p53 or loss of other transcription factors are participating in DSC3 gene silencing. Of the five ILC specimens that lacked DSC3, two (40%) were shown to contain methylated CpG islands. The one mucinous carcinoma specimen analyzed showed a loss of DSC3 expression with a concomitant increase in cytosine methylation. In addition, we analyzed two benign fibrocystic disease specimens and in both cases we saw no methylation of the CpG island and DSC3 gene expression in one of two specimens analyzed. At the very 5' region we saw CpG sites that show methylation variable positions in many of the DSC-positive specimens; we interpret these CpGs to likely be demarcating the edge of the CpG island. Indeed, the first four 5' sites are outside of the minimal promoter region and are likely to be at the edge of the functional CpG island where methylation is more variable [37-39]. Nonetheless, methylation of the DSC3 promoter correlates with a lack of expression of DSC3 in a significant proportion of the primary tumor specimens examined.
To further characterize in vitro models for studying the epigenetic state of the DSC3 promoter we extended prior studies [6,29] and analyzed 14 human breast tumor cell lines for DSC3 expression by quantitative real time RT-PCR. Tumor expression levels were normalized to GAPDH, and expression was then compared to HMECs. Normalized expression levels are shown in Fig. 3. In the breast tumor cell lines tested, 11 of 14 (79%) showed a loss of DSC3 expression, whereas HS578T, MDA-MB-468, and UACC3199 showed moderate expression levels. Of note, three of the breast tumor cell lines tested, BT549, MDA-MB-231, and MDA-MB-157, are in agreement with earlier findings [29].
To determine if loss of mRNA expression correlated with a decrease in protein levels, we conducted western blot analysis of DSC3 in a select group of cell lines. Chosen for analysis were the MDA-MB-157, MDA-MB-231, UACC1179, HS578T, and BT549 breast tumor cell lines, as well as the immortalized but non-tumorigenic breast epithelial cell line MCF10A. HaCaT cells, which are a spontaneously immortalized human keratinocyte cell line, served as a positive control for DSC3 expression [40]. The lack of mRNA expression resulted in a marked reduction of DSC3 protein expression in the cell lines tested (Fig. 4). The HS578T cell line, which showed a 7% expression of DSC3 mRNA, did not produce any detectable protein expression, which is likely below the limit of detection for the western blot conducted. MCF10A cells, which express DSC3 mRNA, showed protein expression comparable to the HaCaT cells; however, no protein bands were present in any of the tumor cell lines examined. Therefore, the lack of DSC3 mRNA expression results in a significant loss of DSC3 protein expression in breast tumor cell lines.
To determine if DSC3 expression is lost in association with aberrant methylation of the DSC3 promoter we used sodium bisulfite genomic sequencing to assess the cytosine methylation status of the DSC3 promoter region. Again, 10 to 12 cloned PCR products were sequenced to determine the percent methylation of the 24 CpG sites in the 5' promoter region. The DSC3 promoter region was relatively unmethylated in the DSC3-positive, HMECs, and MCF10A cells (Fig. 5). In the DSC3-negative cell lines MB231, UACC1179, and BT549, there is a strong correlation between cytosine methylation of the promoter region and lack of expression. Interestingly, in the two remaining DSC3-negative cell lines, UACC2087 and MB-453, loss of expression does not correlate with cytosine methylation of its promoter region, which suggests that other mechanisms of gene silencing are present in these cell lines. These results are similar to the conditions found in the clinical specimens where DSC3 is silenced due to cytosine methylation of its promoter region in 41% of specimens analyzed. Notably, as each of these cell lines contain mutant p53, the loss of this transcription factor is likely participating in the silencing of DSC3 [6,41]. Finally, in the two remaining tumor cell lines that express DSC3 we saw little or no methylation of the promoter region. Therefore, the lack of DSC3 expression in these breast tumor specimens is due in part to both epigenetic and genetic mechanisms of gene silencing.
Another facet of epigenetic regulation causally linked to aberrant cytosine methylation is localized changes to chromatin architecture. Generally, methylated and silenced regions are associated with a 'closed' chromatin structure whereas unmethylated and transcriptionally competent regions are associated with an 'open' chromatin structure. We therefore analyzed the chromatin structure of the DSC3 CpG island region by measuring the accessibility of MspI to its cognate binding site (CCGG) (Fig. 2a) using a quantitative real-time, linker-mediated PCR approach [6]. The PCR for this assay involved a hemi-nested amplification approach, with one primer being specific to the ligated linker and two gene specific primers. The first round of PCR used the linker specific primer and a downstream gene specific primer, while the second round used a portion of the first round product and a gene specific primer 3' to that of the first round gene specific primer to increase specificity of the reaction. Using this technique, we showed a 6.5 to 8.5 cycle difference, which translates to a 90 to 362-fold decrease in chromatin accessibility between the two tumor cell lines tested in comparison to MCF10A cells (Fig. 6). Therefore, DSC3 gene silencing is linked to aberrant cytosine methylation and a closed chromatin structure.
Discussion
The purpose of this study was to determine the frequency of DSC3 gene silencing in primary breast tumor specimens and to determine if the loss of expression was due to the aberrant methylation of the DSC3 CpG island promoter. Analysis of a panel of normal tissue revealed DSC3 mRNA expression to be limited to cell types of epithelial origin. The loss of DSC3 expression in primary breast carcinomas and tumor cell lines has been previously reported [29]. We extended these studies and show that downregulation of DSC3 is a common event in both primary breast tumors as well as breast tumor cell lines, in which we saw a 72% and 79% loss of expression, respectively. DSC3 gene silencing is linked to aberrant cytosine methylation in 41% of the primary breast carcinomas tested. Furthermore, using in vitro models we show that the epigenetic silencing of DSC3 is due in part to cytosine methylation of its promoter region and to concomitant changes in chromatin structure that lead to it forming a closed, inaccessible conformation in the breast cancer cell lines.
We analyzed the cytosine methylation status of 24 CpG sites just 5' of transcriptional start. Within the first seven CpG sites analyzed, we saw methylation variable positions in nearly all primary tumors and cell lines analyzed. This finding signifies to us that the first seven sites are at the very edge of the functional CpG island, where methylation tends to be more variable and coincidentally resides outside of the defined minimal promoter [36-39]. The last 17 CpGs analyzed are within the minimal human promoter region previously identified and are almost completely unmethylated in DSC3-positive cells. Therefore, cytosine methylation within the DSC3 minimal promoter region results in the silencing of DSC3 gene expression.
Of the primary tumor specimens exhibiting a loss of DSC3 expression, several were not associated with cytosine methylation of its promoter region. In these particular cases we predict that the loss of critical transcription factors may be contributing to the silencing of gene expression. Notably, we have shown [6] that DSC3 is a p53 response gene and that the addition of wild-type p53 is sufficient to induce acetylation of the DSC3 promoter region and induce re-expression of DSC3 in breast tumors. Thus, we hypothesize that loss of transcription factors is an early event in tumor progression. We further hypothesize that subsequent to the loss of critical transcription factors, the promoter regions become 'unprotected' and aberrant cytosine methylation that occurs in the region induces long term gene silencing, similar to epigenetically regulated cell type-specific genes [32].
While functional studies have identified DSC3 as a potential tumor suppressor gene [27], future studies are necessary to determine the role of DSC3 in breast tumor initiation and progression. Compelling evidence in the literature indicates an important role for the loss of cellular adhesion in breast tumor progression. In an elegant study, Sternlicht et al. [42] showed that transgenic mice that express an auto-activating form of MMP-3/stromelysin-1, under the control of the whey acidic protein gene promoter, undergo spontaneous development of premalignant and malignant lesions in the mammary glands when compared to their non-transgenic littermates. This study, conducted over a two year period, shows that the single addition of MMP-3, a gene that encodes an enzyme that degrades extracellular components such as fibronectin, laminin, collagens III, IV, IX, and X, and cartilage proteoglycans, is sufficient to induce moderate to severe mammary hyperplasia, lymphocytic infiltrates, ductal carcinoma in situ, and mammary carcinomas. The loss of cell adhesion molecules is thus sufficient to induce neoplastic mammary diseases and warrants the further investigation of the role of desmosomal protein loss in breast tumor progression. Furthermore, desmosomal proteins such as DSC3 have been shown to be critical for desmosome formation, cell position, and inhibition of cell motility [27,43]. As such, the identification of DSC3 as a gene that is commonly downregulated in breast cancer necessitates the need for further examination of its role in breast tumor progression.
Conclusion
The finding that the DSC3 gene is frequently silenced by epigenetic mechanisms in breast cancer opens new avenues to understanding the underlying causes of malignant progression in breast cancer and helps to identify new targets for therapeutic intervention.
Abbreviations
bp = base pair; DSC = desmocollin; H&E = hematoxylin and eosin; HIPAA = health insurance portability and accountability act of 1996; HMEC = human mammary epithelial cell; IDC = invasive ductal carcinoma; ILC = invasive lobular carcinoma; PBS = phosphate-buffered saline.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MMO helped develop the methods to isolate RNA and DNA from patient specimens, participated in the sodium bisulfite sequencing analysis, helped in the design of the study, and drafted the manuscript. CJK is the collaborating surgical oncologist responsible for obtaining patient specimens in accordance with HIPAA guidelines and also participated in the pathology review of all specimens. RJW helped develop and implement the chromatin accessibility assay for this study. DJJ conducted the microarray study that identified DSC3. JLMR developed the primers for use in the sodium bisulfite sequence analysis and also participated in the sequencing of the patient specimens. JAB was responsible for RNA and DNA isolations from all patient specimens, conducted the real-time RT-PCR analysis of patient specimens, and also participated in sodium bisulfite sequence analysis. MF participated in sodium bisulfite sequence analysis. SP conducted the protein isolations and performed the western blot analysis. AEC, an expert in cell adhesion, helped to critically review the manuscript and provided the resources for the western blot analysis. FED, an expert in the field of epigenetics, contributed to the design of the study, participated in and provided the resources for sodium bisulfite sequencing of DSC3, and critically reviewed the manuscript. BWF conceived of the study, participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript
Acknowledgements
We would like to thank Greg Loeffelholz for his technical assistance. This work was supported, in part, by NIH grants CA65662 to BWF, CA73612 to FED, and CA56666 and CA75152 to AEC, as well as P30 CA23074 to the Arizona Cancer Center. The Graduate Training Program in Toxicology grant ES07091 supported MMO. DJJ was supported by T32 CA09213.
Figures and Tables
Figure 1 DSC3 expression is restricted to a subset of normal human epithelial cell types. DSC3 expression relative to human mammary epithelium cells (HMECs) was assessed by real-time quantitative RT-PCR; GAPDH expression was used to normalize the data.
Figure 2 The DSC3 promoter is aberrantly methylated in primary breast tumor samples. (a) Diagram of the DSC3 promoter region analyzed (with the minimal promoter region demarcated as described in [36]). (b) Summary of 5-methylcytosine levels obtained by sodium bisulfite genomic sequencing of the DSC3 promoter. Ten to twelve cloned PCR products were sequenced to determine the percent methylation of the 24 CpG sites in the region analyzed. Cytosine methylation frequency histograms are shown for normal HMECs and eight primary tumor specimens. The y-axis is percent cytosine methylation and the x-axis is the nucleotide position relative to the transcription start site.
Figure 3 DSC3 gene expression is silenced or greatly reduced in a high percentage of breast tumor cell lines. DSC3 expression relative to human mammary epithelium cells (HMECs) was assessed by real-time quantitative RT-PCR; GAPDH expression was used to normalize the data.
Figure 4 DSC3 protein is not expressed in breast tumor cells with undetectable DSC3 mRNA levels. Protein expression was analyzed by western blot analysis. MCF10A and HaCaT cells were used as positive controls for DSC3 expression.
Figure 5 The DSC3 promoter is aberrantly methylated in breast tumor cell lines. Ten to twelve cloned PCR products were sequenced to determine the percent methylation of the 24 CpG sites in the region analyzed. Cytosine methylation frequency histograms are shown for human mammary epithelium cells (HMECs) and the immortalized non-tumorigenic MCF10A cells, and seven human breast cancer cell lines examined. The y-axis is percent cytosine methylation and the x-axis is the nucleotide position relative to the transcription start site.
Figure 6 Hypermethylated DSC3 promoter regions are inaccessible to in vivo MspI endonuclease digestion. Intact nuclei were isolated from MDA-MB-231 and UACC1179 cells and digested in vivo with MspI. Isolated DNA was ligated with a linker specific to the MspI ends, and hemi-nested, linker mediated PCR was conducted with two rounds of PCR with two gene specific primers. Increased amounts of PCR product reveal the presence of accessible chromatin. Inset within the graph is the average calculated fold decrease and standard deviation when MDA-MB-231 and UACC1179 cells are compared to MCF10A. The graph shown is representative of three independent replicates.
Table 1 Summary of DSC3 expression and methylation state in primary breast tumors
Cell/tumor Expressiona Methylationb %Mec Age (years) Histology
HMEC 100.0% - 15 N/A N/A
MCF10A 150.0% - 7 N/A N/A
120T 81.4% - 3 59 IDC
7732T 6.9% - 10 44 IDC
6385T 0.0% ++ 69 70 IDC
173T 0.1% + 30 55 IDC
8900T 77.2% - 2 56 IDC
4658T 52.4% - 8 83 IDC
7768T 2.3% - 7 40 IDC
2504T 67.8% - 16 53 IDC
9613T 8.0% - 9 42 IDC
6010T 0.3% + 21 55 IDC
2909T 26.3% - 1 38 IDC
2845T 0.0% ++ 49 77 IDC
d5974-1T 10.0% + 23 43 IDC
d5974-2T 10.0% ++ 48 43 IDC
7093T 10.0% + 30 63 IDC
9068T 0.0% - 8 73 IDC
4392T 0.0% + 33 47 IDC
5799T 0.0% + 22 53 IDC
6245T 0.0% - 1 30 IDC
2405T 30.0% + 20 58 IDC
2420T 0.0% - 8 54 IDC
5256T 3.9% - 1 60 IDC
1139T 6.4% - 11 40 IDC lymph node Met.
9663T 0.1% + 60 41 IDC lymph node Met.
9985T 5.8% - 0 61 ILC
7788T 1.3% + 23 76 ILC
6861T 20.2% + 29 71 ILC
5358T 73.9% ++ 58 42 ILC
6608T 5.6% - 8 74 ILC
7491T 30.0% - 2 64 ILC
6809T 0.0% ++ 49 57 ILC
4099T 0.6% + 37 43 Mucoid ductal CA
aRNA expression levels determined by quantitative real-time PCR and relative to human mammary epithelial cells (HMECs). bCpG island methylation levels determined by bisulfite sequencing: ++, >40% methylation of total CpG sites analyzed; +, >20% methylation of total CpG sites; -, <20% methylation of total CpG sites analyzed. cPercent methylation was calculated based on the number of methylated CpG sites compared to the total number of sites analyzed. dTwo independent tumors isolated from the same breast. IDC, invasive ductal carcinoma; ILC, invasive lobular carcinoma; N/A, not applicable, IDC lymph node Met, invasive ductal carcinoma that metastasized to the lymph node, Mucoiod Ductal CA, mucinous ductal carcinoma.
==== Refs
Domann FE Rice JC Hendrix MJ Futscher BW Epigenetic silencing of maspin gene expression in human breast cancers Int J Cancer 2000 85 805 810 10709100 10.1002/(SICI)1097-0215(20000315)85:6<805::AID-IJC12>3.0.CO;2-5
Esteller M Fraga MF Paz MF Campo E Colomer D Novo FJ Calasanz MJ Galm O Guo M Benitez J Herman JG Cancer epigenetics and methylation Science 2002 297 1807 1808 12229925 10.1126/science.297.5588.1807d
Baylin SB Herman JG Graff JR Vertino PM Issa JP Alterations in DNA methylation: a fundamental aspect of neoplasia Adv Cancer Res 1998 72 141 196 9338076
Jones PA DNA methylation and cancer Oncogene 2002 21 5358 5360 12154398 10.1038/sj.onc.1205597
Rice JC Futscher BW Transcriptional repression of BRCA1 by aberrant cytosine methylation, histone hypoacetylation and chromatin condensation of the BRCA1 promoter Nucleic Acids Res 2000 28 3233 3239 10954590 10.1093/nar/28.17.3233
Oshiro MM Watts GS Wozniak RJ Junk DJ Munoz-Rodriguez JL Domann FE Futscher BW Mutant p53 and aberrant cytosine methylation cooperate to silence gene expression Oncogene 2003 22 3624 3634 12789271 10.1038/sj.onc.1206545
Rice JC Allis CD Code of silence Nature 2001 414 258 261 11713509 10.1038/35104721
Hanahan D Weinberg RA The hallmarks of cancer Cell 2000 100 57 70 10647931 10.1016/S0092-8674(00)81683-9
Widschwendter M Jones PA DNA methylation and breast carcinogenesis Oncogene 2002 21 5462 5482 12154408 10.1038/sj.onc.1205606
Graff JR Herman JG Myohanen S Baylin SB Vertino PM Mapping patterns of CpG island methylation in normal and neoplastic cells implicates both upstream and downstream regions in de novo methylation J Biol Chem 1997 272 22322 22329 9268383 10.1074/jbc.272.35.22322
Bianco T Chenevix-Trench G Walsh DC Cooper JE Dobrovic A Tumour-specific distribution of BRCA1 promoter region methylation supports a pathogenetic role in breast and ovarian cancer Carcinogenesis 2000 21 147 151 10657950 10.1093/carcin/21.2.147
Dobrovic A Simpfendorfer D Methylation of the BRCA1 gene in sporadic breast cancer Cancer Res 1997 57 3347 3350 9269993
Graff JR Herman JG Lapidus RG Chopra H Xu R Jarrard DF Isaacs WB Pitha PM Davidson NE Baylin SB E-cadherin expression is silenced by DNA hypermethylation in human breast and prostate carcinomas Cancer Res 1995 55 5195 5199 7585573
Rice JC Ozcelik H Maxeiner P Andrulis I Futscher BW Methylation of the BRCA1 promoter is associated with decreased BRCA1 mRNA levels in clinical breast cancer specimens Carcinogenesis 2000 21 1761 1765 10964110 10.1093/carcin/21.9.1761
Dammann R Yang G Pfeifer GP Hypermethylation of the cpG island of Ras association domain family 1A (RASSF1A), a putative tumor suppressor gene from the 3p21.3 locus, occurs in a large percentage of human breast cancers Cancer Res 2001 61 3105 3109 11306494
Bachman KE Herman JG Corn PG Merlo A Costello JF Cavenee WK Baylin SB Graff JR Methylation-associated silencing of the tissue inhibitor of metalloproteinase-3 gene suggest a suppressor role in kidney, brain, and other human cancers Cancer Res 1999 59 798 802 10029065
Costa FF Verbisck NV Salim AC Ierardi DF Pires LC Sasahara RM Sogayar MC Zanata SM Mackay A O'Hare M Epigenetic silencing of the adhesion molecule ADAM23 is highly frequent in breast tumors Oncogene 2004 23 1481 1488 14661055 10.1038/sj.onc.1207263
Wheelock MJ Soler AP Knudsen KA Cadherin junctions in mammary tumors J Mammary Gland Biol Neoplasia 2001 6 275 285 11547897 10.1023/A:1011319507155
Getsios S Huen AC Green KJ Working out the strength and flexibility of desmosomes Nat Rev Mol Cell Biol 2004 5 271 281 15071552 10.1038/nrm1356
Garrod DR Merritt AJ Nie Z Desmosomal cadherins Curr Opin Cell Biol 2002 14 537 545 12231347 10.1016/S0955-0674(02)00366-6
Kowalski PJ Rubin MA Kleer CG E-cadherin expression in primary carcinomas of the breast and its distant metastases Breast Cancer Res 2003 5 R217 222 14580257 10.1186/bcr651
Chidgey MA Yue KK Gould S Byrne C Garrod DR Changing pattern of desmocollin 3 expression accompanies epidermal organisation during skin development Dev Dyn 1997 210 315 327 9389456 10.1002/(SICI)1097-0177(199711)210:3<315::AID-AJA11>3.0.CO;2-9
Koch PJ Franke WW Desmosomal cadherins: another growing multigene family of adhesion molecules Curr Opin Cell Biol 1994 6 682 687 7833048 10.1016/0955-0674(94)90094-9
Arnemann J Sullivan KH Magee AI King IA Buxton RS Stratification-related expression of isoforms of the desmosomal cadherins in human epidermis J Cell Sci 1993 104 741 750 8314871
Collins JE Legan PK Kenny TP MacGarvie J Holton JL Garrod DR Cloning and sequence analysis of desmosomal glycoproteins 2 and 3 (desmocollins): cadherin-like desmosomal adhesion molecules with heterogeneous cytoplasmic domains J Cell Biol 1991 113 381 391 2010468 10.1083/jcb.113.2.381
Troyanovsky SM Troyanovsky RB Eshkind LG Leube RE Franke WW Identification of amino acid sequence motifs in desmocollin, a desmosomal glycoprotein, that are required for plakoglobin binding and plaque formation Proc Natl Acad Sci USA 1994 91 10790 10794 7971964
Tselepis C Chidgey M North A Garrod D Desmosomal adhesion inhibits invasive behavior Proc Natl Acad Sci USA 1998 95 8064 8069 9653140 10.1073/pnas.95.14.8064
Hanakawa Y Amagai M Shirakata Y Sayama K Hashimoto K Different effects of dominant negative mutants of desmocollin and desmoglein on the cell-cell adhesion of keratinocytes J Cell Sci 2000 113 1803 1811 10769211
Klus GT Rokaeus N Bittner ML Chen Y Korz DM Sukumar S Schick A Szallasi Z Down-regulation of the desmosomal cadherin desmocollin 3 in human breast cancer Int J Oncol 2001 19 169 174 11408939
Huang Y Domann FE Transcription factor AP-2 mRNA and DNA binding activity are constitutively expressed in SV40-immortalized but not normal human lung fibroblasts Arch Biochem Biophys 1999 364 241 246 10190980 10.1006/abbi.1999.1142
Johnson GK Organ CC Prostaglandin E2 and interleukin-1 concentrations in nicotine-exposed oral keratinocyte cultures J Periodontal Res 1997 32 447 454 9266496
Futscher BW Oshiro MM Wozniak RJ Holtan N Hanigan CL Duan H Domann FE Role for DNA methylation in the control of cell type specific maspin expression Nat Genet 2002 31 175 179 12021783 10.1038/ng886
Kondo M Finkbeiner WE Widdicombe JH Simple technique for culture of highly differentiated cells from dog tracheal epithelium Am J Physiol 1991 261 L106 L117 1651662
Zabner J Zeiher BG Friedman E Welsh MJ Adenovirus-mediated gene transfer to ciliated airway epithelia requires prolonged incubation time J Virol 1996 70 6994 7003 8794344
Boukamp P Petrussevska RT Breitkreutz D Hornung J Markham A Fusenig NE Normal keratinization in a spontaneously immortalized aneuploid human keratinocyte cell line J Cell Biol 1988 106 761 771 2450098 10.1083/jcb.106.3.761
Gadhavi PL Greenwood MD Strom M King IA Buxton RS The regulatory region of the human desmocollin 3 promoter forms a DNA four-way junction Biochem Biophys Res Commun 2001 281 520 528 11181078 10.1006/bbrc.2001.4375
Millar DS Krawczak M Cooper DN Variation of site-specific methylation patterns in the factor VIII (F8C) gene in human sperm DNA Hum Genet 1998 103 228 233 9760209 10.1007/s004390050810
Millar DS Ow KK Paul CL Russell PJ Molloy PL Clark SJ Detailed methylation analysis of the glutathione S-transferase pi (GSTP1) gene in prostate cancer Oncogene 1999 18 1313 1324 10022813 10.1038/sj.onc.1202415
Tost J Schatz P Schuster M Berlin K Gut IG Analysis and accurate quantification of CpG methylation by MALDI mass spectrometry Nucleic Acids Res 2003 31 e50 12711695 10.1093/nar/gng050
Nuber UA Schafer S Stehr S Rackwitz HR Franke WW Patterns of desmocollin synthesis in human epithelia: immunolocalization of desmocollins 1 and 3 in special epithelia and in cultured cells Eur J Cell Biol 1996 71 1 13 8884173
Beroud C Soussi T The UMD-p53 database: new mutations and analysis tools Hum Mutat 2003 21 176 181 12619103 10.1002/humu.10187
Sternlicht MD Lochter A Sympson CJ Huey B Rougier JP Gray JW Pinkel D Bissell MJ Werb Z The stromal proteinase MMP3/stromelysin-1 promotes mammary carcinogenesis Cell 1999 98 137 146 10428026 10.1016/S0092-8674(00)81009-0
Runswick SK O'Hare MJ Jones L Streuli CH Garrod DR Desmosomal adhesion regulates epithelial morphogenesis and cell positioning Nat Cell Biol 2001 3 823 830 11533662 10.1038/ncb0901-823
|
16168112
|
PMC1242132
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jun 16; 7(5):R669-R680
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1273
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12741616811310.1186/bcr1274Research ArticleAddition of 5-fluorouracil to doxorubicin-paclitaxel sequence increases caspase-dependent apoptosis in breast cancer cell lines Zoli Wainer [email protected] Paola 1Tesei Anna [email protected] Francesco 1Rosetti Marco 2Maltoni Roberta 1Giunchi Donata Casadei 1Ricotti Luca 1Brigliadori Giovanni 2Vannini Ivan 2Amadori Dino [email protected] Division of Oncology and Diagnostics, Morgagni Pierantoni Hospital, Forlì, Italy2 Istituto Oncologico Romagnolo, Forlì, Italy2005 22 6 2005 7 5 R681 R689 9 3 2005 26 5 2005 Copyright © 2005 Zoli et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
The aim of the study was to evaluate the activity of a combination of doxorubicin (Dox), paclitaxel (Pacl) and 5-fluorouracil (5-FU), to define the most effective schedule, and to investigate the mechanisms of action in human breast cancer cells.
Methods
The study was performed on MCF-7 and BRC-230 cell lines. The cytotoxic activity was evaluated by sulphorhodamine B assay and the type of drug interaction was assessed by the median effect principle. Cell cycle perturbation and apoptosis were evaluated by flow cytometry, and apoptosis-related marker (p53, bcl-2, bax, p21), caspase and thymidylate synthase (TS) expression were assessed by western blot.
Results
5-FU, used as a single agent, exerted a low cytotoxic activity in both cell lines. The Dox→Pacl sequence produced a synergistic cytocidal effect and enhanced the efficacy of subsequent exposure to 5-FU in both cell lines. Specifically, the Dox→Pacl sequence blocked cells in the G2-M phase, and the addition of 5-FU forced the cells to progress through the cell cycle or killed them. Furthermore, Dox→Pacl pretreatment produced a significant reduction in basal TS expression in both cell lines, probably favoring the increase in 5-FU activity. The sequence Dox→Pacl→48-h washout→5-FU produced a synergistic and highly schedule-dependent interaction (combination index < 1), resulting in an induction of apoptosis in both experimental models regardless of hormonal, p53, bcl-2 or bax status. Apoptosis in MCF-7 cells was induced through caspase-9 activation and anti-apoptosis-inducing factor hyperexpression. In the BRC-230 cell line, the apoptotic process was triggered only by a caspase-dependent mechanism. In particular, at the end of the three-drug treatment, caspase-8 activation triggered downstream executioner caspase-3 and, to a lesser degree, caspase-7.
Conclusion
In our experimental models, characterized by different biomolecular profiles representing the different biology of human breast cancers, the schedule Dox→Pacl→48-h washout→5-FU was highly active and schedule-dependent and has recently been used to plan a phase I/II clinical protocol.
==== Body
Introduction
Breast cancer is still a leading cause of cancer death in women in the United States and Europe [1]. Adjuvant chemotherapy has been shown to provide disease-free and overall survival benefits for patients with node-positive breast cancer in large meta-analyses conducted by the Early Breast Cancer Trialists' Collaborative Group [2]. A study by Weiss et al. [3], however, showed that the overwhelming majority (80%) of node-positive patients relapse and die within 26 years of diagnosis despite the use of cyclophasphamide-methotrexate-5-fluorouracil (CMF)-based adjuvant chemotherapy. The development of drugs and strategies to improve relapse-free and overall survival therefore remains a high priority.
Anthracyclines are among the most active chemotherapeutic drugs for the treatment of breast cancer and anthracycline-containing regimens have an impact, albeit modest, on patient survival [2,4].
Paclitaxel (Pacl), belonging to the chemical class of taxanes, is capable of inducing in vitro apoptosis, independently of p53 status, through its microtubule-stabilizing activity and, as recently published, by inducing the release of cathepsin B from lysosomes [5]. From a clinical point of view, taxanes appear to be promising, although their real impact on the natural history of breast cancer has yet to be defined. A similar impact of Pacl on clinical response has been reported, however, for patients with wild-type or p53-mutated cancers [6-9], the latter representing 15% of in situ breast carcinomas and 50% of invasive disease [10]. Moreover, the loss of normal p53 function sensitizes in vitro cells to drug activity [11-13], hence the interest to use the taxane combination with other drugs as clinical therapy. The association of a taxane with an anthracycline is based on evidence that the two drugs have different mechanisms of action and toxicities and are not cross-resistant. Moreover, in vitro studies [14-17] on human cell lines and primary breast cancer cultures have shown a schedule-dependent interaction of the two drugs. Phase I and II studies with the doxorubicin (Dox)-Pacl sequence defined by preclinical studies have reported objective responses ranging from 40% to more than 90% [18,19], and a recent prospective phase III study on advanced breast cancer reported an impressive response rate (up to 94%) for the sequence, which was shown to prolong overall survival compared to the fluorouracil, epirubicin, cyclophosphamide (FEC) combination [20].
5-Fluorouracil (5-FU), one of the most important agents for the treatment of colorectal, head and neck, pancreatic and breast carcinomas, is a pro-drug that requires conversion to 5-fluoro-deoxyuridine-monophosphate (5FdUMP) and 5-fluoro-deoxyuridine-triphosphate (5FUTP) in cancer cells. Several in vitro studies on human solid tumor cell lines have demonstrated the positive and schedule-dependent interaction of Pacl and 5-FU [21-23]. A synergistic effect was obtained only when tumor cells were exposed to Pacl followed by antimetabolites. Conversely, simultaneous exposure to the two drugs or pretreatment with 5-FU reduced overall cell killing compared to Pacl alone. Specifically, it has been demonstrated that a short pretreatment with Dox increases the activity of Pacl and of the Pacl→gemcitabine sequence [24]. In the light of these preclinical and clinical observations, the present study aimed to investigate the cytotoxic effect produced by the combination of Dox, Pacl and 5-FU in human breast cancer cell lines and to define, at the preclinical level, the most effective treatment scheme.
Materials and methods
Cell lines
The study was performed on two human breast cancer cell lines: a commercial line (MCF-7; 40-h doubling time, obtained from the American Type Culture Collection (Rockville, MD), and a cell line established in our laboratory (BRC-230; 30-h doubling time) [25]. MCF-7 cells express estrogen receptors and bcl-2, harbor a wild-type p53 gene and lack pro-caspase-3. BRC-230 cells have no estrogen receptors, do not express bcl-2, contain a p53 gene mutation, and express pro-caspase-3. Culture medium was composed of DMEM/HAM F12 (1:1) supplemented with FCS (10% v/v), glutamine (2 mM) (Mascia Brunelli SpA, Milan, Italy) and insulin (10 μg/ml) (Sigma Aldrich, Milan, Italy). Cells were used in the exponential growth phase for all experiments.
In vitro chemosensitivity assay
The sulforhodamine B (SRB) assay was used according to the method of Skehan et al. [26]. Briefly, 104 cells in 100 μl of medium/well were plated in 96-well flat-bottomed microtiter plates and 18 to 24 h after plating (an adequate time for exponential growth recovery) culture medium was replaced with fresh medium either containing or not containing the drugs. At the end of drug exposure, cells were fixed for 1 h and stained with 0.4% SRB (Sigma Aldrich) dissolved in 1% acetic acid for 30 minutes. The plates were then washed four times with 1% acetic acid to remove unbound stain, air-dried and solubilized in 100 μl of 10 mM unbuffered Tris base (tris(hydroxymethyl)aminomethane) solution. The optical density of treated cells was detected at 490 nm. Each sample was run in octuplet, and each experiment was repeated three times. Therefore, each experimental value in the graphs represents the median of 24 samples.
Drugs and chemicals
Dox (Pharmacia Italia SpA, Milan, Italy) and 5-FU (Roche, Milan, Italy) were diluted in sterile saline solution, and Pacl (Bristol Meyers Squibb, Rome, Italy) was diluted in 95% ethanol. The drugs were then divided into aliquots and stored at -70°C. Drug stocks were freshly diluted in culture medium before each experiment. Z-LEHD-FMK (caspase-9 inhibitor) and Z-DEVD-FMK (caspase-3 inhibitor) (BD Biosciences Pharmingen, Milan, Italy) were solubilized in dimethylsulfoxide (DMSO) (Sigma Aldrich) and freshly diluted in culture medium at a concentration of 100 μM before each experiment. The final DMSO concentration never exceeded 1% and this condition was used as control in each experiment.
Drug exposure
The drugs were tested singly at scalar concentrations of 0.001, 0.01, 0.1 and 1 μg/ml for Pacl and Dox, and 0.01, 0.1, 1 and 10 μg/ml for 5-FU. The exposure time to each single agent (4 h for Dox and 24 h for Pacl or 5-FU) was chosen from the dose inhibition rate curves and represented the time that produced the maximum effect. Control samples were processed as treated samples but in drug-free medium, and evaluation of the cytotoxic effect was performed immediately after the end of drug exposure. Each experiment testing the activity of the drugs singly and in combination was run in octuplet and repeated three times.
Drug combinations
Dox/Pacl
Based on our previous results, in sequence experiments, cells were exposed to Dox for 4 h, after which the drug-containing medium was removed and cells were incubated for 24 h with fresh medium containing Pacl. In all the combination experiments, each drug was tested at the four different concentrations used for single drug exposure (0.001, 0.01, 0.1 and 1 μg/ml) at a 1:1 ratio.
Dox/Pacl/5-FU
Dox and Pacl were used as described above. After exposure, the medium was removed and cells were cultured in drug-free medium for 48 h, after which they were treated with 5-FU for 24 h. This scheme, derived from previous experimental studies, was considered as the most effective. In these combination experiments, the three drugs were tested at the four different concentrations used for single drug exposure (0.001, 0.01 and 0.1 and 1 μg/ml for Dox and Pacl and 0.01, 0.1, 1 and 10 μg/ml for 5-FU) at a 1:1:10 ratio.
The cytotoxic effect was evaluated immediately after the end of the three-drug exposure. Controls were processed as treated samples but in drug-free medium.
Data analysis
The type of drug interaction was determined by the median effect principle according to the method of Chou and Talalay [27] and was applied for the three-drug treatment [28]. On the basis of this approach, the interaction between the three drugs was quantified by determining a combination index (CI) at increasing levels of cell kill. CI values lower than, equal to, or higher than 1 indicated synergy, additivity, or antagonism, respectively.
Cell cycle perturbations
Cells (2 × 105) were cultured in medium either containing or not containing (control) the cytotoxic drugs in sequence at a concentration of 0.1 μg/ml for Dox and Pacl and 1 μg/ml for 5-FU. After drug exposure, cells were harvested and stained in a solution containing RNAase (10 Ku/ml; Sigma-Aldrich), Nonidet P40 (0.01%; Sigma-Aldrich) and propidium iodide (1 μg/ml; Sigma-Aldrich). Samples were analyzed 30 to 60 minutes later by flow cytometry (Becton Dickinson Italia SpA, Milan, Italy). Data acquisition (10,000 events for each sample) was performed using CELLQuest software (Becton Dickinson Italia). Data were elaborated using Modfit (DNA Modelling System) software (Verity Software House Inc., Topsham, ME, USA) and expressed as fractions of cells in the different cell cycle phases. Samples were run in triplicate and each experiment was repeated three times.
Apoptosis
Apoptosis was evaluated by flow cytometric analysis according to the previously described TUNEL assay procedure [29]. Briefly, the cells, treated as outlined above, were trypsinized, fixed, exposed to TUNEL reaction mixture and counterstained with propidium iodide before FACS analysis.
Trypsinized cells were treated with anti-Fas antibody (DakoCytomation Denmark A/S, Glostrup, Denmark) diluted 1:500 for 1 h at 4°C to evaluate Fas receptor expression by flow cytometry. Cells were then incubated with a goat-mouse FITC-conjugated secondary antibody (DakoCytomation). Flow cytometric data acquisition and analysis were performed using CELLQuest software (Becton Dickinson Italia). For each sample, 15,000 events were recorded.
Western blot analysis
After the three-drug exposure, cells were treated according to the previously described western blot procedure [29]. The monoclonal antibodies used were anti-p53 (PAb 1801, Bioptica, Milan, Italy) 1:400, anti-bcl-2 (clone 124, Dako Corporation, Santa Barbara, CA, USA) 1:50, anti-p21 (clone DCS-60.2, Bioptica) 1:100, antithymidylate synthase (clone TS 106, Bioptica) 1:100 and anti-caspase-8 (clone 12F5, Alexis Biochemicals Corporation, Lausanne, Switzerland) 1:500. We used polyclonal antibodies for anti-caspases -3, -6, -7 and -9 (Cell Signaling Technology, Inc., Beverly, MA, USA), all diluted at 1:500, anti-apoptosis-inducing factor (AIF) (Chemicon International, Inc., Temecula, CA, USA) 1:500 and anti-bax (Pharmingen, San Diego, CA, USA) 1:1000. Actin polyclonal antibody (Sigma-Aldrich) was used as control to demonstrate equal loading. The samples were analyzed using QuantiScan software (Biosoft, Cambridge, UK).
Results
Cytotoxic activity
The cytotoxic activity of 5-FU, of the 4-h Dox→24-h Pacl sequence, and of the 4-h Dox→24-h Pacl→48-h wash-out→24-h 5-FU exposure is reported in Fig. 1. 5-FU showed a moderate cytotoxic activity in both cell lines. The Dox→Pacl sequence produced a strong cytocidal effect and an important synergism in MCF-7 and BRC-230 cell lines, as shown by Chou-Talalay analysis (Table 1). The synergistic effect dramatically increased, starting from the lowest doses, when cells were treated with 5-FU after the sequence anthracycline→taxane→48-h washout (Fig. 1).
Cell cycle perturbations and apoptosis
A 4-h treatment with Dox or a 24-h exposure to 5-FU did not induce biologically relevant cell cycle perturbations (full recovery was observed after a 24-h culture in drug-free medium) or significant apoptosis (data not shown). Conversely, a 24-h treatment with Pacl produced a considerable increase of cells in G2-M phases, with a decrease in G1 phase, which further increased 24 h after drug removal and began to recover after 48 h. Moreover, about 10% and 15% of apoptotic cells were detected in BRC 230 and MCF-7 cell lines, respectively (data not shown). The Dox→Pacl sequence caused a dramatic block of cells in G2-M and a decrease in G1-S phases, which persisted up to 48 h after drug removal. In parallel, the fraction of apoptotic cells (Fig. 2) rose from 15% to 22% as washout time increased. This finding was similar in both cell lines (Table 2).
Cell cycle perturbations consisting of higher and lower fractions of cells in G2 and G1/S phases, respectively, persisted after exposure to the Dox→Pacl→5-FU sequence (Table 2), albeit to a lesser extent than that observed after the Dox→Pacl sequence. A statistically significant increase of up to 40% was observed in apoptotic cells.
The presence of caspase-9 inhibitor during exposure to the Dox→Pacl→5-FU sequence induced a reduction from 40% to 21% in the apoptotic cell fraction in the MCF-7 line (Fig. 3a). Similarly, the presence of caspase-3 inhibitor during the three-drug treatment caused a reduction in the apoptotic fraction from 37% to about 6% in BRC-230 cells (Fig. 3b).
Apoptotic-related markers
In MCF-7 cells basally expressing bcl-2 and harboring wild-type p53, Dox or 5-FU did not have any effect on apoptotic-related markers, whereas Pacl caused bcl-2 phosphorylation as well as p53 overexpression (data not shown). Bcl-2 phosphorylation further increased after the Dox→Pacl sequence. An increase in pro-apoptotic bax, p21 and p53 expression was observed at the end of the two-drug sequence and was even more evident after the three-drug treatment (Fig. 4).
The BRC-230 cell line, lacking bcl-2 and with mutated p53, showed no induction of bax or p53 overexpression after any treatment. Conversely, Dox→Pacl treatment induced an increase in p21 expression that persisted 48 h after drug removal and was still evident at the end of the three-drug exposure (Fig. 4).
With regard to the caspase cascade, caspase-9 was activated in the MCF-7 cell line after the Dox→Pacl sequence and further increased at the end of the three-drug treatment. Similar behavior was observed for caspase-8, but its active 18 kDa fragment was not detected. Caspase-7 activation was induced after the three-drug sequence only (Fig. 5), whereas caspase-6 cleavage (data not shown) was not induced by any of the treatments. In parallel, a hyperexpression of AIF was observed at the end of both two- and three-drug treatments (Fig. 5).
In BRC-230 cells, an induction of caspase-8, with its 18 kDa fragment, was observed at the end of the Dox→Pacl exposure and was still present after the three-drug treatment. Caspases -3 and -7 were activated 48 h after the end of Dox→Pacl treatment and persisted after the three-drug treatment. Conversely, none of the treatments induced caspase-9 (Fig. 5) or -6 expression (data not shown) or altered Fas expression (data not shown).
Thymidylate synthase
Thymidylate synthase (TS) was basally expressed in both cell lines, but at higher levels in MCF-7 than in BRC-230, as seen by a clear western blotting band at 36 kDa (Fig. 6). 5-FU exposure induced a more evident increase in TS expression in the BRC-230 than in the MCF-7 cell line, as demonstrated by the appearance of an additional band of approximately 38 kDa.
Dox-Pacl treatment induced a significant decrease in TS expression in MCF-7 cells characterized by higher basal TS levels, and a slight reduction in low basal TS levels in the BRC-230 cell line. Exposure to 5-FU after Dox→Pacl treatment induced an increase in TS protein expression that was, however, lower than that induced by treatment with 5-FU alone, in both cell lines. Moreover, in BRC-230 cells, the three-drug treatment did not induce a detectable TS ternary complex (Fig. 6).
Discussion
To date the clinical design of polychemotherapeutic protocols has mainly taken into account information derived from experimental studies on mechanisms of action of different agents and has favored combinations of drugs with complementary toxicities, but it has been clearly demonstrated that sequencing and timing of administration are important to optimize drug activity.
Our results confirmed the synergistic interaction of the Dox→Pacl treatment [14-16] and showed that this sequence enhances the efficacy of subsequent exposure to 5-FU, as observed in our previous studies using the antimetabolite gemcitabine [18,24]. In particular, we observed that cells, as expected, were trapped in the G2-M phase after the Dox→Pacl sequence, and the subsequent addition of 5-FU forced cells to progress through the cell cycle or killed them. In fact, the three-drug treatment doubled the percentage of apoptotic cells produced by the Dox→Pacl sequence in both experimental cell lines.
In MCF-7 cells, phosphorylation of the anti-apoptotic gene and persistent upregulation of pro-apoptotic markers was induced by the Dox→Pacl sequence and further enhanced by exposure to 5-FU. It can be hypothesized that the ensuing mitochondrial instability led to apoptosis through both the cytochrome-c-mediated activation of caspases -9 and -7 and the release of AIF. The important role played by AIF in the apoptosis process in MCF-7, which is lacking in caspase-3 [30], was demonstrated by the partial cell death inhibition observed in the presence of caspase-9 inhibitor during the three-drug treatment.
In the BRC-230 cell line, apoptosis was triggered by the activation of caspases -8 and -3 and, albeit to a lesser degree, caspase-7. In particular, the pivotal role of caspase-3 was highlighted by the almost complete suppression of apoptosis in the presence of caspase-3 inhibitor during the three-drug exposure.
TS protein was also modulated after the different treatments in both cell lines. TS, which converts dUMP to dTMP, is the primary target of fluoropyrimidine activity [31]. It represents the rate-limiting nucleotide in DNA synthesis and its overexpression has been shown to be associated with resistance to 5-FU-based treatments. Our results showed, as already reported in colon cancer cell lines [32], an increase in TS free protein and the formation of TS ternary complex following 5-FU exposure. This increase was more evident in the p53-mutated BRC-230 line than in wild-type p53 MCF-7 cells. Following the Dox→Pacl treatment, basal TS expression was significantly reduced in MCF-7 and also, albeit to a lesser degree, in BRC-230 cells. Subsequent exposure to 5-FU increased free TS expression without, however, reaching the levels observed after treatment with 5-FU alone. In particular, the increase involved only free TS in BRC-230 and was associated with the formation of the ternary complex, albeit fourfold lower than that induced by 5-FU alone, in MCF-7 cells.
In conclusion, apoptosis was observed after the Dox→Pacl sequence and was even more evident when followed by 5-FU, which induced an important apoptosis in cell lines characterized by different estrogen receptor status and apoptosis-related markers, both representative of the heterogeneous biology of clinical breast cancers. The efficacy of the antimetabolite was favored by an increase in TS levels following exposure to anthracycline and taxane.
Conclusion
The present preclinical study permitted us to define the most effective three-drug sequence, providing the basis for the rationale of an ongoing phase I/II breast cancer clinical protocol in which the oral formulations of Dox and 5-FU are used to reduce toxicity and increase safety of treatment schemes.
Abbreviations
5FdUMP = 5-fluoro-deoxyuridine-monophosphate; 5-FU = 5-fluorouracil; 5FUTP = 5-fluoro-deoxyuridine-triphosphate; AIF = apoptosis-inducing factor; CI = combination index; CMF = cyclophasphamide-methotrexate-5-fluorouracil; DMEM = Dulbecco's modified Eagle's medium; DMSO = dimethylsulfoxide; Dox = doxorubicin; FCS = fetal calf serum; FEC = fluorouracil, epirubicin, cyclophosphamide; Pacl = paclitaxel; SRB = sulforhodamine B; TS = thymidylate synthase.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
All the authors contributed equally to this study.
Acknowledgements
The authors would like to thank Prof. Rosella Silvestrini for her invaluable scientific contribution, Dr Ivan Vannini for his technical assistance and Gráinne Tierney for editing the manuscript. Supported by Istituto Oncologico Romagnolo, Forlì, Italy.
Figures and Tables
Figure 1 Effect of different drug treatments on the survival of MCF-7 and BRC-230 breast cancer cell lines. (a) 5-FU, (b) Dox→Pacl sequence and (c) Dox→Pacl→5-FU treatment. D1, D2, D3 and D4 represent the doses of the three drugs used in the sequence (see Materials and methods). Each data point is the average of at least three independent experiments performed in octuplet. The standard deviation never exceeded 5%.
Figure 2 BRC-230 cells after Dox (0.1 μg/ml)→Pacl (0.1 μg/ml)→48-h washout→5-FU (1 μg/ml) treatment. Apoptotic nuclei stained with DAPI show intense fluorescence corresponding to chromatin condensation (arrow heads) and fragmentation (arrows).
Figure 3 Inhibition of apoptosis. Results from TUNEL assay showing inhibition induced by Dox (0.1 μg/ml)→Pacl (0.1 μg/ml)→48-h washout→5-FU (1 μg/ml) treatment in (a) BRC-230 cells in the presence of caspase-3 (casp-3) inhibitor and (b) MCF-7cells in the presence of caspase-9 (casp-9) inhibitor.
Figure 4 Protein levels of bcl-2, bax, p53 and p21 following different drug exposures. Protein (50 μg) was loaded for the controls and treated samples: (a) untreated cells; (b) Dox (0.1 μg/ml)→Pacl (0.1 μg/ml); (c) Dox (0.1 μg/ml)→Pacl (0.1 μg/ml)→48-h washout; (d) Dox (0.1 μg/ml)→Pacl (0.1 μg/ml)→48-h washout→5-FU (1 μg/ml).
Figure 5 Western blot analysis of caspases and apoptosis-inducing factor (AIF) proteins following different treatments. Protein (50 μg) was loaded for the controls and treated samples: (a) untreated cells; (b) Dox (0.1 μg/ml)→Pacl (0.1 μg/ml); (c) Dox (0.1 μg/ml)→Pacl (0.1 μg/ml)→48-h washout; (d) Dox (0.1 μg/ml)→Pacl (0.1 μg/ml)→48-h washout→5-FU (1 μg/ml).
Figure 6 Western blot analysis of thymidylate synthase and ternary complex. (a) Untreated cells; (b) 5-FU (1 μg/ml); (c) Dox (0.1 μg/ml)→Pacl (0.1 μg/ml); (d) Dox (0.1 μg/ml)→Pacl (0.1 μg/ml)→48-h washout→5-FU (1 μg/ml). 50 μg of protein were loaded for the controls and treated samples.
Table 1 Combination index values induced by sequential treatments
MCF-7 BRC-230
CI value at inhibition of CI value at inhibition of
Drug combination Combination ratio 75% 90% 95% 75% 90% 95%
Dox→Pacl 1:1 0.3 0.3 0.4 0.1 0.2 0.2
Dox→Pacl→48-h washout→5-FU 1:1:10 0.0005 0.002 0.005 0.002 0.007 0.02
5-FU, 5-fluorouracil; CI, combination index; Dox, doxorubicin; Pacl, paclitaxel.
Table 2 Distribution of cells in the different cell cycle phases (%) and apoptotic cells (%) after different treatments
MCF-7 BRC-230
Treatment G1 S G2-M Apoptosis G1 S G2-M Apoptosis
Untreated cells 48 40 12 1 55 33 12 2
Dox→Pacl 6a 14a 80a 22a 10a 27a 63a 15a
Dox→Pacl→48-h washout 3a 16a 81a 17a 5a 11a 84a 15a
Dox→Pacl→48-h washout→5-FU 15a 49 36a 40a 20a 42a 38a 37a
aP < 0.05 by t-test. 5-FU, 5-fluorouracil; Dox, doxorubicin; Pacl, paclitaxel.
==== Refs
O'Shaughnessy J Clinical experience of capecitabine in metastatic breast cancer Eur J Cancer 2002 38 Suppl 2 10 14 11841930 10.1016/S0959-8049(01)00416-6
Early Breast Cancer Trialists' Collaborative Group Polychemotherapy for early breast cancer: an overview of the randomized trials Lancet 1998 352 930 942 9752815 10.1016/S0140-6736(98)03301-7
Weiss RB Woolf SH Demakos E Holland JF Berry DA Falkson G Cirrincione CT Robbins A Bothun S Henderson IC Natural history of more than 20 years of node-positive primary breast carcinoma treated with cyclophosphamide, methotrexate, and fluorouracil-based adjuvant chemotherapy: a study by the Cancer and Leukemia Group B J Clin Oncol 2003 21 1825 1835 12721260 10.1200/JCO.2003.09.006
Piccart M The role of taxanes in adjuvant treatment of early stage breast cancer Breast Cancer Res Treat 2003 79 Suppl 1 S25 S34 12868803 10.1023/A:1024393926965
Broker LE Huisman C Span SW Rodriguez JA Kruyt FA Giaccone G Cathepsin B mediates caspase-independent cell death induced by microtubule stabilizing agents in non-small cell lung cancer cells Cancer Res 2004 64 27 30 14729603
Safran H King T Choy H Gollerkery A Kwakwa H Lopez F Cole B Myers J Tarpey J Rosmarin A p53 mutations do not predict response to paclitaxel/radiation for nonsmall cell lung carcinoma Cancer 1996 78 1203 1210 8826941 10.1002/(SICI)1097-0142(19960915)78:6<1203::AID-CNCR6>3.0.CO;2-A
Di Leo A The European experience with docetaxel in the treatment of early-stage breast cancer Clin Breast Cancer 2002 3 Suppl 2 S59 S62 12435289
Stearns V Singh B Tsangaris T Crawford JG Novielli A Ellis MJ Isaacs C Pennanen M Tibery C Farhad A A prospective randomized pilot study to evaluate predictors of response in serial core biopsies to single agent neoadjuvant doxorubicin or paclitaxel for patients with locally advanced breast cancer Clin Cancer Res 2003 9 124 133 12538460
Hamilton A Larsimont D Paridaens R Drijkoningen M van de Vijver M Bruning P Hanby A Houston S Treilleux I Guastalla JP A study of the value of p53, HER2, and Bcl-2 in the prediction of response to doxorubicin and paclitaxel as single agents in metastatic breast cancer: a companion study to EORTC 10923 Clin Breast Cancer 2000 1 233 240 discussion 241–242 11899648
Zellars RC Hilsenbeck SG Clark GM Allred DC Herman TS Chamness GC Elledge RM Prognostic value of p53 for local failure in mastectomy-treated breast cancer patients J Clin Oncol 2000 18 1906 1913 10784631
Debernardis D Sire EG De Feudis P Vikhanskaya F Valenti M Russo P Parodi S D'Incalci M Broggini M p53 status does not affect sensitivity of human ovarian cancer cell lines to paclitaxel Cancer Res 1997 57 870 874 9041188
Perego P Romanelli S Carenini N Magnani I Leone R Bonetti A Paolicchi A Zunino F Ovarian cancer cisplatin-resistant cell lines: multiple changes including collateral sensitivity to Taxol Ann Oncol 1998 9 423 430 9636834 10.1023/A:1008265012435
Zaffaroni N Silvestrini R Orlandi L Bearzatto A Gornati D Villa R Induction of apoptosis by taxol and cisplatin and effect on cell cycle-related proteins in cisplatin-sensitive and -resistant human ovarian cells Br J Cancer 1998 77 1378 1385 9652752
Amadori D Frassineti GL Zoli W Milandri C Tienghi A Ravaioli A Gentile A Salzano E A phase I/II study of sequential doxorubicin and paclitaxel in the treatment of advanced breast cancer Semin Oncol 1996 23 5 Suppl 11 16 22 8893894
Amadori D Frassineti GL Zoli W Milandri C Serra P Tienghi A Ravaioli A Gentile A Salzano E Doxorubicin and paclitaxel (sequential combination) in the treatment of advanced breast cancer Oncology (Huntingt) 1997 11 4 Suppl 3 30 33 9144688
Frassineti GL Zoli W Silvestro L Serra P Milandri C Tienghi A Gianni L Gentile A Salzano E Amadori D Paclitaxel plus doxorubicin in breast cancer: an Italian experience Semin Oncol 1997 24 Suppl 17 S17 S19 S17-S25 9374087
Danesi R Conte PF Del Tacca M Pharmacokinetic optimisation of treatment schedules for anthracyclines and paclitaxel in patients with cancer Clin Pharmacokinet 1999 37 195 211 10511918
Ibrahim T Fabbri M Frassineti GL Zoli W Monti M Ricotti L Amadori D Doxorubicin, paclitaxel and gemcitabine: a Phase I study of a new sequential treatment in stage III B – IV breast cancer J Chemother 2003 15 488 494 14598942
Holmes FA Madden T Newman RA Valero V Theriault RL Fraschini G Walters RS Booser DJ Buzdar AU Willey J Hortobagyi GN Sequence-dependent alteration of doxorubicin pharmacokinetics by paclitaxel in a phase I study of paclitaxel and doxorubicin in patients with metastatic breast cancer J Clin Oncol 1996 14 2713 2721 8874332
Zielinski CC Gemcitabine, anthracycline, and taxane combinations for advanced breast cancer Oncology (Huntingt) 2003 17 12 Suppl 14 36 40 14768404
Kano Y Akutsu M Tsunoda S Ando J Matsui J Suzuki K Tkeda T Inoue Y Adaki K Schedule-dependent interaction between paclitaxel and 5-fluorouracil in human carcinoma cell lines in vitro Br J Cancer 1996 74 704 710 8795571
Smorenburg CH Sparreboom A Bontelbal M Verweij J Combination chemotherapy of taxanes and antimetabolites: its use and limitations Eur J Cancer 2001 37 2310 2323 11720823 10.1016/S0959-8049(01)00309-4
Johnson KR Fan W Reduced expression of p53 and p21 WAF1/CIP1 sensitizes human breast cancer cells to paclitaxel and its combination with 5-fluorouracil Anticancer Res 2002 22 3197 3204 12530065
Zoli W Ricotti L Barzanti F Dal Susino M Frassineti GL Milandri C Casadei Giunchi D Amadori D Schedule-dependent interaction of doxorubicin, paclitaxel and gemcitabine in human breast cancer cell lines Int J Cancer 1999 80 413 416 9935183 10.1002/(SICI)1097-0215(19990129)80:3<413::AID-IJC13>3.0.CO;2-I
Amadori D Bertoni L Flamigni A Savini S De Giovanni C Casanova S De Paola F Amadori A Giulotto E Zoli W Establishment and characterization of a new cell line from primary human breast carcinoma Breast Cancer Res Treat 1993 28 251 260 8018954 10.1007/BF00666586
Skehan P Storeng R Scudiero D Monks A McMahon J Vistica D Warren JT Bokesch H Kenney S Boyd MR New colorimetric cytotoxic assay for anticancer drug screening J Natl Cancer Inst 1990 82 1107 1112 2359136
Chou TC Talalay P Quantitative analysis of dose-effect relationship: the combined effects of multiple drugs or enzyme inhibitors Adv Enzyme Regul 1984 22 27 55 6382953 10.1016/0065-2571(84)90007-4
Chou TC Motzer RJ Tong Y Bosl GJ Computerized quantitation of synergism and antagonism of taxol, topotecan, and cisplatin against human teratocarcinoma cell growth: a rational approach to clinical protocol design J Natl Cancer Inst 1994 86 1517 1524 7932806
Zoli W Ricotti L Tesei A Ulivi P Gasperi Campani A Fabbri F Gunelli R Frassineti GL Amadori D Schedule-dependent cytotoxic interaction between epidoxorubicin and gemcitabine in human bladder cancer cells in vitro Clin Cancer Res 2004 10 1500 1507 14977854
Lee ATC Azimahtol HLP Tan AN Styrylpyrone derivative (SPD) induces apoptosis in a caspase-7-dependent manner in the human breast cancer cell line MCF-7 Cancer Cell Int 2003 3 16 23 14580263 10.1186/1475-2867-3-16
Backus HHJ Wouters D Ferreira CG van Houten VMN Brakenhoff RH Pinedo HM Peters GJ Thymidylate synthase inhibition triggers apoptosis via caspase-8 and -9 in both wild-type and mutant p53 colon cancer cell lines Eur J Cancer 2003 39 1310 1317 12763222 10.1016/S0959-8049(03)00204-1
Peters GJ van Triest B Backus HHJ Kuiper CM van der Wilt CL Pinedo HM Molecular downstream events and induction of thymidylate synthase in mutant and wild-type p53 colon cancer cell line after treatment with 5-fluorouracil and the thymidylate synthase inhibitor raltitrexed Eur J Cancer 2000 36 916 924 10785598 10.1016/S0959-8049(00)00026-5
|
16168113
|
PMC1242133
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jun 22; 7(5):R681-R689
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1274
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12761616811510.1186/bcr1276Research ArticleSe-methylselenocysteine inhibits phosphatidylinositol 3-kinase activity of mouse mammary epithelial tumor cells in vitro Unni Emmanual [email protected] Dimpy [email protected] Wai-Kwan Alfred [email protected] Raghu [email protected] Medicine Endocrinology, Baylor College of Medicine, Houston, Texas, USA2 Department of Neuro-Oncology, University of Texas MD Anderson Cancer Center, Houston, Texas, USA3 Department of Biochemistry and Molecular Biology, Penn State College of Medicine, Hershey, Pennsylvania, USA2005 6 7 2005 7 5 R699 R707 25 8 2004 24 11 2004 19 5 2005 27 5 2005 Copyright © 2005 Unni et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Se-methylselenocysteine (MSC), a naturally occurring selenium compound, is a promising chemopreventive agent against in vivo and in vitro models of carcinogen-induced mouse and rat mammary tumorigenesis. We have demonstrated previously that MSC induces apoptosis after a cell growth arrest in S phase in a mouse mammary epithelial tumor cell model (TM6 cells) in vitro. The present study was designed to examine the involvement of the phosphatidylinositol 3-kinase (PI3-K) pathway in TM6 tumor model in vitro after treatment with MSC.
Methods
Synchronized TM6 cells treated with MSC and collected at different time points were examined for PI3-K activity and Akt phosphorylation along with phosphorylations of Raf, MAP kinase/ERK kinase (MEK), extracellular signal-related kinase (ERK) and p38 mitogen-activated protein kinase (MAPK). The growth inhibition was determined with a [3H]thymidine incorporation assay. Immunoblotting and a kinase assay were used to examine the molecules of the survival pathway.
Results
PI3-K activity was inhibited by MSC followed by dephosphorylation of Akt. The phosphorylation of p38 MAPK was also downregulated after these cells were treated with MSC. In parallel experiments MSC inhibited the Raf–MEK–ERK signaling pathway.
Conclusion
These studies suggest that MSC blocks multiple signaling pathways in mouse mammary tumor cells. MSC inhibits cell growth by inhibiting the activity of PI3-K and its downstream effector molecules in mouse mammary tumor cells in vitro.
==== Body
Introduction
Several organic and inorganic selenium compounds have been reported to be effective chemopreventive agents against multiple models of mammary tumorigenesis in both the mouse and the rat [1-5]. Selenium compounds have been shown to exert marked stage specificity, especially in preneoplastic mammary lesions, but neither normal mammary gland development nor existing mammary tumor growth was affected by selenium supplemental status [1,6,7]. Although the precise mechanisms by which selenium compounds inhibit mammary tumorigenesis are not well understood, there is evidence that the inorganic and organic selenium compounds act through different pathways [8-10]. Selenium compounds have been reported to affect numerous cellular events and molecular pathways leading to apoptosis. Molecular targets for various natural and synthetic organoselenium compounds have been reviewed [11-15].
Selenite, a widely used inorganic selenium compound, is considered cytotoxic and causes single-stranded DNA breaks and also other non-specific effects [16]. In contrast, Se-methylselenocysteine (MSC) is a less toxic organic selenium compound occurring naturally. It is the major form of selenium compound in selenium-enriched garlic, onions and broccoli [17]. In the mammary tumor model, MSC is more efficacious than the most extensively studied selenoamino acids in animal models [15,18]. Furthermore, MSC inhibits cell growth in several mouse mammary tumor cell lines [19,20] and human breast cancer cell lines [21]. We and other investigators have shown that this inhibition of cell growth is mediated through the induction of apoptosis in vitro [20-22] and in vivo [23-25]. Using a synchronized mouse mammary cell line TM6, we have shown previously that MSC inhibits DNA synthesis, followed by the arrest of cells in S phase [19]. This block is associated with decreased cdk2 kinase activity [19] and altered cdk2 phosphorylation [26]. In addition, treatment of cells with MSC decreases PKC activity and increases gadd (34, 45 and 153) gene expression in a time-dependent manner [26]. Furthermore, using the same model system, we also reported increased caspase-3, caspase-6 and caspase-8 activities, leading to apoptosis in the MSC-treated TM6 cells in a synchronized model [22].
The effect of MSC on mammary survival pathways is not well understood. One of the earliest responses of starved cells that are exposed to extracellular stimulation with growth factors including serum is the simultaneous activation of both the Raf–MAP kinase/ERK kinase–extracellular signal-related kinase (Raf–MEK–ERK) and phosphatidylinositol 3-kinase (PI3-K)–Akt pathways [27,28]. Activation of Raf can lead to opposing cellular responses such as proliferation, growth arrest, apoptosis or differentiation, depending on the duration and strength of the external stimulation and on the cell type [29]. There is a lack of published data on the effect of selenium on Raf in mammary tumors. PI3-K regulates diverse cellular functions such as growth, survival and malignant transformation through its multiple enzymatic functions, namely lipid kinase and protein kinase activities [30,31], and acts either synergistically with the Raf pathway [32] or in opposition to it [33]. There are few reports demonstrating effects of selenium on PI3-K, but the effect of MSC on PI3-K activity has not been reported previously. One of the possible anti-apoptotic effects of PI3-K is brought about by the phosphorylation of Akt, which in turn can cross-talk with Raf by phosphorylating it at a highly conserved serine residue (Ser259) in its regulatory domain and inhibiting the activation of the Raf–MEK–ERK pathway. The effects of selenium on Akt are limited and the results vary depending on the form (whether inorganic or organic) and on cell type. For the present investigation we examined the effects of MSC on the components of the PI3-K–Akt and Raf–MEK–ERK pathways to improve our understanding of the mechanisms of growth inhibition in the synchronized TM6 mouse mammary tumor cell line.
Materials and methods
Cell culture and treatment with MSC
The TM6 tumor cell line was originally derived from the non-tumorigenic COMMA-D mouse mammary epithelial cell line [34]. TM6 tumor cells generate alveolar mammary tumors in Balb/c mice when injected into the fat pads. These tumors are p53 mutant and are predicted to be estrogen independent. TM6 cells were cultured routinely in DMEM/F-12 medium containing growth factors (5 ng/ml epidermal growth factor, 10 μg/ml insulin), serum (2% adult bovine serum) and 1 × antibiotic–antimycotic solution (Invitrogen Corporation, Carlsbad, CA, USA) in the presence of 5% CO2 in air at 37°C [19]. In brief, the cells were plated at a density of 6.6 × 103 cells/cm2 in either 100 mm dishes or six-well plates. After growth for 48 hours (Fig. 1) the cells were starved in DMEM/F12 medium without growth factors and serum (minimal medium) for a further 48 hours. The cells were then released from starvation with DMEM/F12 medium containing growth factors and serum. After a further 6 hours, MSC (Sigma, St Louis, MO, USA) was added at a final concentration of 100 μM (unless otherwise mentioned) to one set of cells. Cells were collected after starvation (0 hours), then at 6 (before the addition of MSC), 9, 12, 16 and 24 hours. These times reflect the points at which cells were stimulated with growth factors and serum after starvation, minus 6 hours of treatment time with MSC as described previously [19].
MSC pretreatment
To study the effect of MSC on the native and phosphorylated Akt, Raf and MEK signals that arise immediately after the addition of medium containing growth factors and serum to starved cells, the cells were synchronized in minimal medium for at least 24 hours. MSC was then added (in minimal medium) for the stipulated time points. The cells were stimulated with fresh DMEM/F12 medium containing growth factors and serum in the continued presence of MSC and were harvested 1 hour later. In these experiments, the time refers to the point at which the cells were pretreated with MSC before the stimulation.
Incorporation of [3H]thymidine
Synchronized TM6 cells grown in 12-well plates (2.5 × 104 cells per well) were treated with 50 μM MSC for various durations and pulsed for 1 hour with 1 μCi of [3H]thymidine (MP Biomedicals, Irvine, CA, USA) per well. After three washings with Tris-buffered saline, the cells were treated with 10% trichloroacetic acid for 5 min followed by two washes with trichloroacetic acid. The incorporation of [3H]thymidine was determined by counting the vials in a liquid-scintillation counter. The assay was performed in triplicate for all time points [19].
Antibodies
Polyclonal anti-(phospho-Akt (Ser473)), anti-Akt, anti-(phospho-Raf), anti-(phospho-MEK), anti-(phospho-ERK (p44/p42)), anti-(phospho-p38 MAPK) and horseradish peroxidase (HRP)-conjugated anti-rabbit antibody were purchased from New England Biolabs (Beverly, MA, USA). Monoclonal anti-PTEN, anti-actin and HRP-conjugated anti-goat antibody were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti-(PI3-K (p85)) antibody was purchased from Upstate (Lake Placid, NY, USA).
Isolation of protein and immunoblotting
Cell pellets collected after being washed with cold PBS were lysed for 30 min in a buffer containing 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na3VO4, 1 μg/ml leupeptin and 1 mM phenylmethylsulphonyl fluoride on ice. The post-mitochondrial supernatants were collected after centrifugation at 8,000 g for 10 min and were measured for total protein content with a BCA™ Protein Assay Kit (Pierce, Rockford, IL, USA). Equal amounts of protein were loaded for a given western blot analysis. A range of 20 to 50 μg of protein was loaded in each lane as indicated in the respective figure legends. Immunoblot analysis was performed as described previously [19]. The signals were detected by enhanced chemiluminescence (Amersham Biosciences Corp, Piscataway, NJ, USA) and quantified with the ImageQuant software (Molecular Dynamics, Sunnyvale, CA, USA). The protein loading on gels was normalized to that of actin.
PI3-K activity
PI3-K activity was measured with the method described by Truitt and colleagues [35]. The cell pellets were lysed in solubilization buffer containing 50 mM HEPES (pH 7.0), 150 mM NaCl, 1 mM EGTA, 10 mM NaF, 10 mM sodium pyrophosphate, 10% glycerol, 1% Triton X-100, 1 mM Na3VO4, 1 μM pepstatin, 10 μg/ml aprotinin, 5 mM iodoacetic acid and 2 μg/ml leupeptin. Cell extracts (500 μg) were then incubated for 2 hours with 4 μl of anti-PI3-K at 4°C and for a further 2 hours with 50 μl of Protein A–Sepharose beads (Amersham Biosciences Corp). After centrifugation, the immunoprecipitates were washed sequentially as follows: first, three times with PBS containing 1% Triton X-100 and 100 μM Na3VO4; second, twice with 100 mM Tris-HCl (pH 7.6), 0.5 M LiCl and 100 μM Na3VO4; third, twice with 100 mM Tris-HCl (pH 7.6), 100 mM NaCl, 1 mM EDTA and 100 μM Na3VO4; and fourth, twice with 20 mM HEPES (pH 7.5), 50 mM NaCl, 1 mM EDTA, 30 mM sodium pyrophosphate, 200 μM Na3VO4, 0.03% Triton X-100 and 1 mM phenylmethylsulphonyl fluoride.
The washed immunoprecipitates were resuspended in 30 μl of kinase buffer containing 33.3 mM Tris-HCl (pH 7.6), 125 mM NaCl, 16.6 mM MgCl2, 164.3 mM adenosine and 16.6 μM ATP. To this mix, 30 μCi of [γ-32P]ATP (1 mCi/100 μl), 7 μl of water and 20 μg of phosphatidylinositol 4-monophosphate prepared in 10 μl of 20 mM HEPES (pH 7.5) was added. The reaction was performed at room temperature on a rotary mixer for 30 min. After the addition of 100 μl of 1 M HCl to stop the reaction, the phosphorylated substrate was extracted with 600 μl of chloroform : methanol (1:1). The organic phase was then separated by centrifugation at 3,000 r.p.m. for 5 min, re-extracted with 200 μl of deionized water and dried by centrifugation under vacuum. The lipid was redissolved in 20 μl of chloroform : methanol (1:1) mixture. The radiolabeled phosphatidylinositol phosphate was resolved on silica gel G-60 thin-layer chromatography plates by chromatography for 3 hours in a solvent system of chloroform : methanol : ammonium hydroxide : water (60:47:2:11.3) and was revealed by autoradiography.
Results
Treatment with MSC inhibited DNA synthesis in both asynchronous (Fig. 2a) and synchronized (Fig. 2b) TM6 mouse mammary epithelial tumor cells, as measured by [3H]thymidine incorporation. The untreated control cells incorporated maximum [3H]thymidine at 16 hours when most of the cells are in S phase, as reported previously [19], whereas DNA synthesis in cells treated with 50 μM MSC was inhibited by 33% at this time point. The same dose of MSC suppressed [3H]thymidine incorporation to a greater degree in asynchronous cells; this was mainly due to the longer treatment period, 48 hours.
MSC induces apoptosis in mammary epithelial tumor cells [19,20] and we have documented that caspase-3 activity is enhanced in MSC-treated cells at 24 hours [22]. Because the activation of caspase-3 is a late event in the progression of apoptosis, we examined the phosphorylation of Akt, which is one of the early key signals controlling proliferation and/or apoptosis. The expression of Akt protein remained unchanged in MSC-treated and untreated control cells until 24 hours (Fig. 3). However, at 24 hours there was an increase in Akt phosphorylation in the control cells, and a 68% decrease in MSC-treated cells. This decrease in phospho-Akt was not due to a decline in the native Akt levels.
Since PI3-K is an upstream target of Akt, we wished to determine whether this decrease in phospho-Akt levels in MSC-treated cells was in fact due to a lower PI3-K activity. For measuring the activity, PI3-K from control and MSC-treated cells (16 hours and 24 hours) was immunoprecipitated with anti-p85 antibody and assayed for its ability to phosphorylate phosphatidylinositol 4-monophosphate. In the TM6 synchronized model, PI3-K activity increased within 1 hour of stimulation with serum (Fig. 4); this was blocked by 1 μM wortmannin (PI3-K inhibitor). There was a 73% and 84% decrease in PI3-K activity in MSC-treated cells at 16 and 24 hours, respectively, in comparison with the control cells.
Because PI3-K is inactivated by the lipid phosphatase PTEN (MMAC1), we further examined whether the decrease in PI3-K activity was due to an increase in PTEN levels. The levels of PTEN were determined at different time points by immunoblotting (Fig. 5); no appreciable differences were observed between MSC-treated and control cells up to 24 hours.
Treatment with MSC of TM6 cells at 24 hours inhibited both Akt phosphorylation (Fig. 3) and PI3-K activity (Fig. 4). The lowered PI3-K activity could be due either to an effect of MSC on the enzyme activity or to the inhibition of an upstream event, such as Ras activation. To dissect the two possibilities we examined the two independent downstream parallel pathways that were activated by Ras: first, the activation of Raf by Ras and its downstream targets MEK and ERK, and second, the activation of PI3-K and its downstream targets Akt and p38 mitogen-activated protein kinase (MAPK). We speculated that if MSC inhibits Ras along with the decrease in phospho-Akt levels, which we had observed at 24 hours, the phosphorylation of p38 MAPK or ERK should also decline. Fig. 6 shows the phosphorylated state of Raf in MSC-treated and untreated cells at different time points. The levels remained unchanged in both the samples at 9, 12 and 16 hours. At 24 hours the phospho-Raf levels were 58% lower in MSC-treated cells. A similar pattern of decreased phosphorylation was observed for phospho-Erk (p44/42) when MSC-treated and control cells were compared at different time points. The phosphorylation pattern of phospho-p38 MAPK, a downstream target of Akt, mimicked the pattern of phospho-Akt levels in MSC-treated versus control cells. There was no difference in the phospho-p38 MAPK levels in MSC-treated and control cells until 24 hours. However, the levels of phospho-p38 MAPK increased at 24 hours in control cells and were inhibited more than threefold in MSC-treated cells. The levels of native Akt, ERK, p38 MAPK and Raf proteins did not change with treatment with MSC (data not shown).
To distinguish between the tolerance of MSC concentrations and their effects in signaling, components of both the Raf and Akt pathways, namely phosphoprotein levels of Akt, Raf and MEK, were analyzed in TM6 cells synchronized in minimal medium for 24 hours and then treated with different doses of MSC in minimal medium for 16 and 24 hours before stimulation with growth factors and serum. As expected, all three proteins were phosphorylated within 1 hour of stimulation (Fig. 7). At 16 hours, even at 400 μM MSC, the phosphorylated protein levels of Akt and Raf were comparable to that of the control. However, at 24 hours their levels decreased with increasing concentrations of MSC. The native Akt and MEK levels did not show an appreciable change at all time points (data not shown); the native Raf protein expression did not change either during this experiment. The immunoblot in Fig. 6 also demonstrates that at 24 hours the levels of these phosphoproteins started to increase in the control cells, indicating the start of a second wave of stimulation.
To examine whether MSC needs to be metabolized to have an effect on the phosphorylation of Akt, cells were synchronized with minimal medium for 24 hours and were subsequently treated with 100 μM MSC for various periods (0 to 24 hours), stimulated with growth factors and serum for 1 hour and examined for Akt phosphorylation (Fig. 8a). Pretreatment of the cells with MSC for 10 hours, equivalent to the cells collected at 16 hours in the previous scheme of experiments (Fig. 1), Akt phosphorylation was inhibited by only 26% (Fig. 8b). After 18 and 24 hours' pretreatment of TM6 cells with MSC, the inhibition in phospho-Akt levels was 49% and 65%, respectively, and was significant (P<0.05) when compared with untreated cells.
Discussion
The results presented here demonstrate that MSC inhibits PI3-K activity and subsequently inactivates Akt in vitro. This is a significant observation in establishing one of the mechanisms by which MSC inhibits mouse mammary epithelial cell growth in vitro.
Previously we had reported that TM6 cells treated with MSC are delayed in S phase at about 24 hours [19,26]. In the present set of experiments the differences in Akt phosphorylation between MSC-treated and untreated control cells occur at about 24 hours. This observation was not clear because Akt phosphorylation is an immediate event, occurring within 1 hour of stimulation with growth factors and serum. Various possibilities exist: first, inhibition of Akt phosphorylation in MSC-treated cells beginning at 24 hours might require the cells to be delayed in S phase; second, there might be a requirement for MSC to be metabolized into an active molecule such as methylselenol [36] that causes inhibition; or third, there might be a slow diffusion of MSC into the cells. We have shown that MSC enters the TM6 cells within 30 min of treatment and can inhibit DNA synthesis in these cells 3 hours later [22], thus excluding the probability of slower diffusion into the cells.
To address the first two of these alternatives, different strategies were designed in TM6 cells. In the first set of experiments (scheme outlined in Fig. 1), the cells were allowed to cycle after stimulation with growth factors and serum, and MSC was added 6 hours later. In these experiments, events leading to Akt phosphorylation had already taken place before the addition of MSC. By 16 hours, although PI3-K activity was inhibited in the MSC-treated cells, the phospho-Akt levels remained unchanged in both the control and MSC-treated cells. In the TM6 synchronization model we noted that the Akt phosphorylation is stimulated again at a later time point in the cell cycle. The occurrence of this 'second wave of stimulation' is quite evident from an elevated level of phospho-p38 MAPK at 24 hours in control cells. This stimulation actually appeared at 22 hours (data not shown) in TM6 cells when examined closely. PI3-K activity was inhibited at about 16 hours, and thus its effect on Akt phosphorylation occurs only with the second wave of stimulation. This could explain why phospho-Akt levels were the same in both MSC-treated and untreated control cells at 16 hours even though the PI3-K activity was inhibited in the MSC-treated cells.
Second, the fact that PI3-K activity is inhibited earlier than Akt-phosphorylation supports the hypothesis that the upstream target of MSC-induced growth inhibition is PI3-K. When the cells were pretreated with MSC and then stimulated with growth factors and serum, there was a gradual inhibition of Akt phosphorylation. Most of the cells during this synchronization state would be predicted to be in G1 phase during this time [19], so the possibility that factors causing a delay in S phase might result in a decreased phosphorylation of Akt can be excluded.
The probable reason that the differences in the Akt phosphorylation are not observed until 24 hours is that MSC might need to be metabolized to methylselenol before it can effectively inactivate Akt. MSC can be metabolized into methylselenol, which could be dimethylated and trimethylated to dimethylselenide or trimethylselenonium respectively [37]. Other organoselenium compounds such as dimethylselenoxide and selenobetaine methyl ether can be metabolized to dimethylselenide and trimethylselenonium without the formation of methylselenol and do not have anticancer activity. It has therefore been suggested that methylselenol is the active proximal molecule of MSC [37]. MSC is capable of generating methylselenol endogenously through the action of β-lyase or related lyases [38]. As the cells in culture have low levels of β-lyase, it leads to the inefficient conversion of MSC to methylselenol [23,39,40], and so we used higher doses of MSC (100 to 400 μM) in some of our experiments. Several current studies have looked at an alternative methylselenol generator, methylseleninic acid, a compound that represents a simplified version of MSC without the amino acid moiety, thereby obviating the need for β-lyase action. There are a few reports indicating the differential effect of selenium compounds on Akt in vascular endothelial [41], prostate [42], mammary [43] and oral [44] cancer cells depending on the form of selenium. On the basis of our present results the speculated sites of MSC interaction with components of Ras–PI3-K–Akt pathway and Raf–MEK–ERK pathway are illustrated in Fig. 9.
Akt interacts with Raf and phosphorylates it at Ser259. Furthermore, phosphorylation of Raf by Akt inhibits activation of the Raf–MEK–ERK signaling pathway and has been shown to alter the cellular response in a human breast cancer cell line from cell cycle arrest to proliferation [29]. Our results indicate that this cross-talk between Akt and Raf might be altered by MSC. It has also been reported that Akt is a substrate for caspase and cleaves it into 40 and 44 kDa fragments [45]. We have recently shown that the activities of caspase-3, caspase-6 and caspase-8 are increased at 24 hours of treatment with MSC [22]. The cleaved phospho-Akt proteins were observed at 24 hours in MSC-treated cells. It is unlikely that the decrease in Akt phosphorylation at 24 hours was due to elevated caspase activity because PI3-K was inhibited at 16 hours, before the activation of these caspases could be detected in the cells.
It was recently demonstrated that certain tumor suppressor agents downregulate PI3-K by activating the expression of PTEN/MMAC1, a phosphatase that dephosphorylates phosphatidylinositol 3,4,5-trisphosphate [46]. Although MSC could inhibit PI3-K activity in the present study this inhibition was not due to elevated levels of PTEN.
PI3-K is a heterodimer with a catalytic and a regulatory subunit. The catalytic subunit possesses both lipid kinase and serine–threonine protein kinase activities. PI3-K is activated by the binding of either receptor or non-receptor tyrosine kinases to the regulatory subunit; this complex is directed to the membrane and associates with its phospholipid substrate [47]. Because the lipid kinase activity of PI3-K is inhibited on treatment with MSC before any effect on the phosphorylation of Akt, it would be interesting to examine whether MSC could block the integration of PI3-K to the membrane; this is part of an investigation currently in progress. Another important scenario might be if MSC were shown to interfere with the activity of Ras, because both phospho-Raf and phospho-Akt levels are lowered during treatment with MSC. To perform its function, the active form of Ras (GTP-Ras) must also be anchored to the cellular membrane through a post-translationally added lipophilic (iso) prenyl group [48]. Further studies are required to investigate whether MSC alters the anchoring of Ras and PI3-K into the cell membrane.
Conclusion
The present studies show that MSC blocks multiple pathways in mouse mammary tumor cells in vitro. Decreased PI3-K activity in addition to dephosphorylation of Akt by MSC contributes to the growth inhibition of TM6 mouse mammary epithelial cells. This information, along with the possibility that p38 MAPK is a target for the action of MSC on mammary cells, will provide further evidence of its mechanistic inhibition of mammary growth. These experiments need to be translated into human cell lines and xenograft model systems before this compound can be promoted for clinical trials in humans for breast cancer prevention.
Abbreviations
DMEM/F12 = Dulbecco's modified Eagle's medium/nutrient mixture F-12 Ham; ERK = extracellular signal-related kinase; HRP = horseradish peroxidase; MAPK = mitogen-activated protein kinase; MSC = Se-methylselenocysteine; PI3-K = phosphatidylinositol 3-kinase; TM6 = mouse mammary epithelial tumor cells.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EU treated cells and performed western blot analyses for the native and phosphorylated proteins. DK was responsible for PI3-K activity conducted in W-KAY's laboratory. RS established the in vitro synchronized TM6 model, performed the [3H]thymidine incorporation assay and was responsible for overall design, statistical analysis, and supervision of all the experiments. EU and RS contributed in manuscript writing. All authors read and approved the final manuscript.
Acknowledgements
The Authors thank Dr. Daniel Medina at the Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, Texas, and Dr. Karam El-Bayoumy at the Department of Biochemistry and Molecular Biology, Penn State College of Medicine, Hershey, Pennsylvania, for their helpful comments on the manuscript. This work was performed in the Department of Molecular and Cellular Biology at Baylor College of Medicine before the departure of EU and RS. The work was supported in part by grants from NCI (RO1 CA56041 to W-KAY) and from the US Army Medical Research and Materiel Command (DAMD 17-99-1-9076 to RS).
Figures and Tables
Figure 1 General scheme for synchronization and treatment of TM6 cells with Se-methylselenocysteine (MSC) including the collection times. The TM6 cells were plated at a density of 6.6 × 103 cells/cm2 in either 100 mm dishes or six-well plates. After 48 hours of growth the cells were starved in DMEM/F12 medium without growth factors and serum (minimal medium) for a further 48 hours. The cells were released from starvation with DMEM/F12 medium containing growth factors (5 ng/ml epidermal growth factor (EGF) and 10 μg/ml insulin) and serum (2% adult bovine serum). After a further 6 hours MSC was added at a final concentration of 50 to 400 μM (depending upon the experiment) to one set of cells. Untreated cells served as controls. The cells were collected after starvation (0 hours), then at 6 (before the addition of MSC), 9, 12, 16 and 24 hours time-points.
Figure 2 [3H]Thymidine incorporation into TM6 cells after Se-methylselenocysteine (MSC) treatment. (a) Asynchronous TM6 cells were grown for 24 hours and treated with various concentrations of MSC for 48 hours to determine the optimum dose for treating synchronized cells. (b) Synchronized TM6 cells were treated with 50 μM MSC at 6 hours and the DNA synthesis was measured by [3H]thymidine incorporation at the indicated time points as described in the Materials and methods section. Data are presented as means ± SEM for three observations at each given time point. MSC at 50 μM showed a greater ability to block DNA synthesis in asynchronous TM6 cells, mainly because of the extended time of treatment.
Figure 3 Effect of Se-methylselenocysteine (MSC) on Akt. Synchronized TM6 cells were treated with 50 μM MSC as described in Fig. 1. For each time point three 100 mm dishes were treated with MSC and pooled for protein content. Equal amounts of protein lysates were loaded on each lane (50 μg) for each time point. (a) Immunoblots were probed with anti-Akt, anti-phospho-Akt (Ser473) and anti-actin as described in the Materials and methods section. The phosphorylation of Akt occurring at 24 hours in control cells was inhibited in the MSC-treated cells. (b) The levels of phospho-Akt in control and MSC-treated cells were normalized with respective actin contents and plotted against various time points.
Figure 4 Effect of Se-methylselenocysteine (MSC) on phosphatidylinositol 3-kinase (PI3-K) activity in TM6 cells. Synchronized TM6 cells were treated with 100 μM MSC at 6 hours as described in Fig. 1. Another set of TM6 cells were pretreated with 1 μM wortmannin (WOR) for 30 min before the 1 hour stimulation with fresh DMEM/F12 medium containing growth factors (5 ng/ml epidermal growth factor and 10 μg/ml insulin) and serum (2% adult bovine serum). PI3-K activity was performed on 500 μg of protein lysates as described in the Materials and methods section. The kinase activity in the control cells increased within 1 hour of stimulation, and was strongly inhibited by WOR. The PI3-K activity in MSC-treated cells at 16 and 24 hours were drastically lowered compared with that of the control cells. The data are a representative of experiments performed in triplicate for each time point.
Figure 5 Effect of Se-methylselenocysteine (MSC) on PTEN levels in TM6 cells. Synchronized TM6 cells were treated with 50 μM MSC as described in Fig. 1. An equal amount of protein lysates (50 μg) was loaded on each lane. Immunoblots were probed with anti-PTEN and anti-actin as described in the Materials and methods section. The protein levels of PTEN in the control cells were not significantly different from that of MSC-treated cells at various time points.
Figure 6 Effect of Se-methylselenocysteine (MSC) on phospho-Raf, phospho-ERK, phosphorylated p38 mitogen-activated protein kinase (phospho-p38 MAPK) levels in TM6 cells. TM6 cells were synchronized and treated with 50 μM MSC as described in Fig. 1. Equal amounts of TM6 lysates (50 μg of protein) were loaded on each lane. Immunoblots were probed with anti-phospho-Raf, anti-phospho-ERK, anti-phospho-p38 MAPK and anti-actin as described in the Materials and methods section. Levels of the phosphoproteins remained unchanged at 9, 12 and 16 hours but at 24 hours the phosphorylated proteins decreased in MSC-treated cells.
Figure 7 Effect of Se-methylselenocysteine (MSC) on phospho-Akt, phospho-Raf and phospho-MEK in TM6 cells. The TM6 cells were synchronized in minimal medium as described in Fig. 1, but only for 24 hours. Then the MSC (100 to 400 μM) was added (in minimal medium) for 16 and 24 hours. The cells were stimulated with fresh DMEM/F12 containing growth factors (5 ng/ml epidermal growth factor and 10 μg/ml insulin) and serum (2% adult bovine serum) at the indicated time points for 1 hour. Equal amounts of lysates (20 μg of protein) were loaded on each lane for each time point. Immunoblots were probed with anti-phospho-Akt, anti-phospho-Raf, anti-phospho-MEK and anti-actin antibodies as described in the Materials and methods section. Both the Akt and Raf were phosphorylated within 1 hour of stimulation with growth factor and serum. At 16 hours a dose of 400 μM MSC failed to inhibit the phosphorylation of Akt or Raf and the downstream effector MEK. However, at 24 hours MSC was able to inhibit phosphorylation of all three proteins in a dose-dependent manner.
Figure 8 Effect of Se-methylselenocysteine (MSC) on Akt phosphorylation. (a) Scheme for pretreatment of TM6 cells with MSC. Cells were synchronized in minimal medium for 24 hours. Cells were then exposed to 100 μM MSC for 3, 6, 10, 18 and 24 hours in minimal medium before being stimulated with fresh DMEM/F12 medium containing growth factors (5 ng/ml epidermal growth factor and 10 μg/ml insulin) and serum (2% adult bovine serum) for 1 hour. (b) Effect of pretreatment of MSC on Akt phosphorylation in TM6 cells. Equal amounts of lysates (30 μg of protein) were loaded on each lane for each time point. After electrophoresis the immunoblots were probed with anti-phospho-Akt (Ser473) antibody as described in the Materials and methods section, and the levels were measured with Molecular Dynamics software. Each bar represents levels in TM6 tumor cells treated with MSC in three different wells. *P < 0.05 compared with 0 hours.
Figure 9 Possible sites of Se-methylselenocysteine (MSC) interaction with components of the Ras–phosphatidylinositol 3-kinase–Akt (Ras–PI3-K–Akt) and Raf–MAP kinase/ERK kinase–ERK (Raf–MEK–ERK) pathways in TM6 mouse mammary tumor cells. MAPK, mitogen-activated protein kinase.
==== Refs
Schrauzer GN White DA Schneider CJ Inhibition of the genesis of spontaneous mammary tumors in C3H mice: effects of selenium and of selenium-antagonistic elements and their possible role in human breast cancer Bioinorg Chem 1976 6 265 270 1053516 10.1016/S0006-3061(00)80232-X
Medina D Morrison DG Current ideas on selenium as a chemopreventive agent Pathol Immunopathol Res 1988 7 187 199 3057483
Ip C Hayes C Budnick RM Ganther HE Chemical form of selenium, critical metabolites and cancer prevention Cancer Res 1991 51 595 600 1824684
El-Bayoumy K Upadhyaya P Chae YH Sohn OS Rao CV Fiala E Reddy BS Chemoprevention of cancer by organoselenium compounds J Cell Biochem Suppl 1995 22 92 100 8538214 10.1002/jcb.240590812
Ip C Zhu Z Thompson HJ Lisk D Ganther HE Chemoprevention of mammary cancer with Se-allylselenocysteine and other selenoamino acids in the rat Anticancer Res 1999 19 2875 2880 10652567
Medina D Mechanisms of selenium inhibition of tumorigenesis Adv Exp Med Biol 1986 206 465 472 3591534
Ip C Thompson HJ Ganther HE Selenium modulation of cell proliferation and cell cycle biomarkers in normal and premalignant cells of the rat mammary gland Cancer Epidemiol Biomarkers Prev 2000 9 49 54 10667463
Ip C Lessons from basic research in selenium and cancer prevention J Nutr 1998 128 1845 1854 9808633
Sinha R Said TK Medina D Organic and inorganic selenium compounds inhibit mouse mammary cell growth in vitro by different cellular pathways Cancer Lett 1996 107 277 284 8947525 10.1016/0304-3835(96)04373-X
Lu J Jiang C Kaeck M Ganther H Vadhanavikit S Ip C Thompson H Dissociation of the genotoxic and growth inhibitory effects of selenium Biochem Pharmacol 1995 50 213 219 7632165 10.1016/0006-2952(95)00119-K
Sinha R El-Bayoumy K Apoptosis is a critical cellular event in cancer chemoprevention and chemotherapy by selenium compounds Curr Cancer Drug Targets 2004 4 13 28 14965264 10.2174/1568009043481614
El-Bayoumy K Sinha R Mechanisms of mammary cancer chemoprevention by organoselenium compounds Mutat Res 2004 551 181 197 15225592
McKenzie RC Arthur JR Beckett GJ Selenium and the regulation of cell signaling, growth, and survival: molecular and mechanistic aspects Antioxid Redox Signal 2002 4 339 351 12006185 10.1089/152308602753666398
Ip C Dong Y Ganther HE New concepts in selenium chemoprevention Cancer Metastasis Rev 2002 21 281 289 12549766 10.1023/A:1021263027659
Whanger PD Selenium and its relationship to cancer: an update dagger Br J Nutr 2004 91 11 28 14748935 10.1079/BJN20031015
Lu J Kaeck M Jiang C Wilson AC Thompson HJ Selenite induction of DNA strand breaks and apoptosis in mouse leukemic L1210 cells Biochem Pharmacol 1994 47 1531 1535 8185664 10.1016/0006-2952(94)90528-2
Whanger PD Selenocompounds in plants and animals and their biological significance J Am Coll Nutr 2002 21 223 232 12074249
Medina D Thompson H Ganther H Ip C Se-methylselenocysteine: a new compound for chemoprevention of breast cancer Nutr Cancer 2001 40 12 17 11799917 10.1207/S15327914NC401_5
Sinha R Medina D Inhibition of cdk2 kinase activity by methylselenocysteine in synchronized mouse mammary epithelial tumor cells Carcinogenesis 1997 18 1541 1547 9276628 10.1093/carcin/18.8.1541
Lu J Pei H Ip C Lisk DJ Ganther H Thompson HJ Effect on an aqueous extract of selenium-enriched garlic on in vitro markers and in vivo efficacy in cancer prevention Carcinogenesis 1996 17 1903 1907 8824512
Dong Y Ganther HE Stewart C Ip C Identification of molecular targets associated with selenium-induced growth inhibition in human breast cells using cDNA microarrays Cancer Res 2002 62 708 714 11830524
Unni E Singh U Ganther HE Sinha R Se-methylselenocysteine activates caspase-3 in mouse mammary epithelial tumor cells in vitro Biofactors 2001 14 169 177 11568454
Ip C Thompson HJ Zhu Z Ganther HE In vitro and in vivo studies of methylseleninic acid: evidence that a monomethylated selenium metabolite is critical for cancer chemoprevention Cancer Res 2000 60 2882 2886 10850432
Unni E Kittrell FS Singh U Sinha R Osteopontin is a potential target gene in mouse mammary cancer chemoprevention by Se-methylselenocysteine Breast Cancer Res 2004 6 R586 R592 15318939 10.1186/bcr914
Ip C Dong Y Methylselenocysteine modulates proliferation and apoptosis biomarkers in premalignant lesions of the rat mammary gland Anticancer Res 2001 21 863 867 11396176
Sinha R Kiley SC Lu JX Thompson HJ Moraes R Jaken S Medina D Effects of methylselenocysteine on PKC activity, cdk2 phosphorylation and gadd gene expression in synchronized mouse mammary epithelial tumor cells Cancer Lett 1999 146 135 145 10656618 10.1016/S0304-3835(99)00250-5
Katz ME McCormick F Signal transduction from multiple Ras effectors Curr Opin Genet Dev 1997 7 75 79 9024640 10.1016/S0959-437X(97)80112-8
Rommel C Hafen E Ras – a versatile cellular switch Curr Opin Genet Dev 1998 8 412 418 9729716 10.1016/S0959-437X(98)80111-1
Zimmermann S Moelling K Phosphorylation and regulation of Raf by Akt (Protein kinase B) Science 1999 286 1741 1744 10576742 10.1126/science.286.5445.1741
Chan TO Rittenhouse SE Tsichlis PN Akt/PKB and other D3 phosphoinositide-regulated kinases: kinase activation by phosphoinositide-dependent phosphorylation Annu Rev Biochem 1999 68 965 1014 10872470 10.1146/annurev.biochem.68.1.965
Krasilinikov MA Phosphatidylinositol-3 kinase dependent pathways: the role in control of cell growth, survival, and malignant transformation Biochemistry (Mosc) 2000 65 59 67 10702641
Rodriguez-Viciana P Warne PH Khwaja A Marte BM Pappin D Das P Waterfield MD Ridley A Downward J Role of phosphoinositide 3-OH kinase in cell transformation and control of the actin cytoskeleton by Ras Cell 1997 89 457 467 9150145 10.1016/S0092-8674(00)80226-3
Hu E Kim JB Sarraf P Spiegelman BM Inhibition of adipogenesis through MAP kinase-mediated phosphorylation of PPARgamma Science 1996 274 2100 2103 8953045 10.1126/science.274.5295.2100
Jerry JD Butel JS Medina D p53 mutations in COMMA-D cells In Vitro Cell Dev Biol Anim 1994 30A 87 89 8012661
Truitt KE Hicks CM Imboden JB Stimulation of CD28 triggers an association between CD28 and phosphatidylinositol 3-kinase in Jurkat T cells J Exp Med 1994 179 1071 1076 7509360 10.1084/jem.179.3.1071
Sinha R Unni E Ganther HE Medina D Methylseleninic acid, a potent growth inhibitor of synchronized mammary epithelial tumor cells in vitro Biochem Pharmacol 2001 61 311 317 11172735 10.1016/S0006-2952(00)00545-1
Vadhanavikit S Ip C Ganther HE Metabolites of sodium selenite and methylated selenium compounds administered at cancer chemoprevention levels in rat Xenobiotica 1993 23 731 745 8237056
Ip C Ganther HE Comparison of selenium and sulfur analogs in cancer prevention Carcinogenesis 1992 13 1167 1170 1638682
Wang Z Jiang C Lu J Induction of caspase-mediated apoptosis and cell-cycle G1 arrest by selenium metabolite methylselenol Mol Carcinog 2002 34 113 120 12112305 10.1002/mc.10056
Zhu Z Jiang W Ganther HE Ip C Thompson HJ In vitro effects of Se-allylselenocysteine and Se-propylselenocysteine on cell growth, DNA integrity, and apoptosis Biochem Pharmacol 2000 60 1467 1473 11020448 10.1016/S0006-2952(00)00461-5
Wang Z Jiang C Ganther H Lu J Antimitogenic and proapoptotic activities of methylseleninic acid in vascular endothelial cells and associated effects on PI3K-AKT, ERK, JNK and p38 MAPK signaling Cancer Res 2001 61 7171 7178 11585751
Jiang C Wang Z Ganther H Lu J Distinct effects of methylseleninic acid versus selenite on apoptosis, cell cycle, and protein kinase pathways in DU145 human prostate cancer cells Mol Cancer Ther 2002 1 1059 1066 12481429
Zhu Z Jiang W Ganther HE Thompson HJ Mechanisms of cell cycle arrest by methylseleninic acid Cancer Res 2002 62 156 164 11782373
Ghose A Fleming J El-Bayoumy K Harrison PR Enhanced sensitivity of human oral carcinomas to induction of apoptosis by selenium compounds: involvement of mitogen-activated protein kinase and Fas pathways Cancer Res 2001 61 7479 7487 11606383
Rokudai S Fujita N Hashimoto Y Tsuruo T Cleavage and inactivation of antiapoptotic Akt/PKB by caspases during apoptosis J Cell Physiol 2000 182 290 296 10623893 10.1002/(SICI)1097-4652(200002)182:2<290::AID-JCP18>3.0.CO;2-8
Maehama T Dixon JE The tumor suppressor, PTEN/MMAC1, dephosphorylates the lipid second messenger, phosphatidylinositol 3,4,5-trisphosphate J Biol Chem 1998 273 13375 13378 9593664 10.1074/jbc.273.22.13375
Rameh LE Chen CS Cantley LC Phosphatidylinositol (3,4,5)P3 interacts with SH2 domains and modulates PI 3-kinase association with tyrosine-phosphorylated proteins Cell 1995 83 821 830 8521499 10.1016/0092-8674(95)90195-7
Garbay C Liu W-Q Vidal M Roques B Inhibitors of Ras signal transduction as antitumor agents Biochem Pharmacol 2000 60 1165 1169 11007954 10.1016/S0006-2952(00)00428-7
|
16168115
|
PMC1242134
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jul 6; 7(5):R699-R707
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1276
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12771616811410.1186/bcr1277Research ArticleEarly and late skin reactions to radiotherapy for breast cancer and their correlation with radiation-induced DNA damage in lymphocytes López Escarlata [email protected] Rosario [email protected]úñez Maria Isabel [email protected] Moral Rosario [email protected] Mercedes [email protected]ínez-Galán Joaquina [email protected] Maria Teresa [email protected]ñoz-Gámez José Antonio [email protected] Francisco Javier [email protected]ín-Oliva David [email protected] Almodóvar José Mariano Ruiz [email protected] Servicio de Oncología Radioterápica, Hospital Universitario Virgen de las Nieves, Granada, Spain2 Instituto de Biopatología y Medicina Regenerativa, Centro de Investigaciones Biomédicas, Departamento de Radiología y Medicina Física, Facultad de Medicina, Universidad de Granada, Granada, Spain3 Instituto de Parasitología y Biomedicina 'López Neyra' CSIC, Parque Tecnológico de Ciencias de las Salud, Granada, Spain2005 1 7 2005 7 5 R690 R698 31 1 2005 5 5 2005 20 5 2005 29 5 2005 Copyright © 2005 López et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Radiotherapy outcomes might be further improved by a greater understanding of the individual variations in normal tissue reactions that determine tolerance. Most published studies on radiation toxicity have been performed retrospectively. Our prospective study was launched in 1996 to measure the in vitro radiosensitivity of peripheral blood lymphocytes before treatment with radical radiotherapy in patients with breast cancer, and to assess the early and the late radiation skin side effects in the same group of patients. We prospectively recruited consecutive breast cancer patients receiving radiation therapy after breast surgery. To evaluate whether early and late side effects of radiotherapy can be predicted by the assay, a study was conducted of the association between the results of in vitro radiosensitivity tests and acute and late adverse radiation effects.
Methods
Intrinsic molecular radiosensitivity was measured by using an initial radiation-induced DNA damage assay on lymphocytes obtained from breast cancer patients before radiotherapy. Acute reactions were assessed in 108 of these patients on the last treatment day. Late morbidity was assessed after 7 years of follow-up in some of these patients. The Radiation Therapy Oncology Group (RTOG) morbidity score system was used for both assessments.
Results
Radiosensitivity values obtained using the in vitro test showed no relation with the acute or late adverse skin reactions observed. There was no evidence of a relation between acute and late normal tissue reactions assessed in the same patients. A positive relation was found between the treatment volume and both early and late side effects.
Conclusion
After radiation treatment, a number of cells containing major changes can have a long survival and disappear very slowly, becoming a chronic focus of immunological system stimulation. This stimulation can produce, in a stochastic manner, late radiation-related adverse effects of varying severity. Further research is warranted to identify the major determinants of normal tissue radiation response to make it possible to individualize treatments and improve the outcome of radiotherapy in cancer patients.
==== Body
Introduction
Ionizing radiation is widely and successfully applied in oncology. However, because of dose restrictions, a definitive cure cannot be achieved for many tumour entities and localizations. Despite the advanced radiotherapy facilities available, high doses of radiation still induce early and late skin effects. Unacceptable normal tissue reactions remain the limiting factor for delivering a tumoricidal dose in radiotherapy. Radiation is an unusual toxic agent in that the timing of tissue damage expression can vary widely between one tissue or tumour and another [1]. On the other hand, recent large-scale trials of adjuvant radiotherapy for breast cancer showed that the overall survival benefit of radiotherapy can be considered an inherent characteristic of the treatment and is not influenced by the duration of follow-up [2]. Data in the literature strongly support a causal relation between better outcomes and improved radiotherapeutic techniques [3]. Changes in radiotherapy practice over the years include recognition of the importance of fraction size, fraction number, total dose, overall time for both tumour and normal tissue reactions, and the introduction of conservative therapy.
Radiotherapy outcomes might be further improved by a greater understanding of the individual variations in normal tissue reactions that determine tolerance [4]. When accurate genetic-based or cell-survival-based predictive assays are available to study tumour and normal tissue radiosensitivity, radiation therapy will become an exact science [5], allowing truly individual optimization and the prediction of adverse reactions [6]. It is of great importance to identify the variations in intrinsic (cellular) radiosensitivity and extrinsic factors that are associated with a change in the risk of morbidity. It has yet to be determined whether intrinsic cell radiosensitivity or extrinsic factors have greater influence on individual differences in damage expression [7-10]. The very high incidence of breast cancer in Western countries, partially attributable to the ageing of their populations, and the increasing use of conservative surgery and postoperative radiotherapy for its treatment make the above type of study of special interest, with the side effects of radiotherapy an increasingly important issue. Indeed, after the sweeping changes in the locoregional treatment of breast cancer during the last part of the 20th century, it appears that only a dwindling minority of patients will undergo mastectomy, at least in urban areas with a high socioeconomic level [11]. The widely varied biological characteristics of patients with breast cancer, evidenced in clinical, pathological, cellular, and molecular studies, are sufficient to explain the diversity of treatments recommended over the past two decades [12]. Recent years have seen the introduction of changes from conventional radiotherapy at 5 × 1.8 to 2.0 Gy per week to more aggressive schedules such as unconventional protocols [13] or radiochemotherapy [11]. The gradually increasing success of cancer treatments has led to longer patient survival. This also carries with it the penalty of providing a greater opportunity for late effects to appear, increasing in severity [14] and affecting the patient's quality of life [15].
With regard to radiotherapy complications, the known factors influencing normal tissue responses account for only 30% of interpatient variability in breast cancer patients under well-controlled conditions, leading to the hypothesis that most of the variability in the severity of these complications is due to differences in cellular radiosensitivity determined by genetic or epigenetic mechanisms [7,10]. Identification of the causes of this variability in radiation sensitivity could have important implications for cancer therapy. Evidence of a possible genetic basis for these differences has been provided by reports of increased cellular and tissue radiosensitivity in certain genetic syndromes [16] and of an association among the relative radiosensitivities of different normal cell types in the same individual [17]; this evidence also verifies that cellular radiosensitivity may be related to tissue response. Current radiobiological research efforts are aimed at identifying patients with abnormal radiosensitivity at risk for acute and late adverse effects of radiotherapy treatment [18,19] and detecting molecules that increase the antitumour effects of radiotherapy [20].
Most published studies on radiation toxicity were performed retrospectively. This prospective study was launched in 1996 to measure the in vitro radiosensitivity of peripheral blood lymphocytes before treatment with radical radiotherapy in patients with breast cancer, and to assess the early and the late side effects of radiation on skin in the same group of patients. We prospectively recruited consecutive breast cancer patients receiving radiation therapy after breast surgery. To evaluate whether early and late side effects of radiotherapy can be predicted by the assay, a study was conducted of the association between the results of in vitro radiosensitivity tests and acute and late adverse effects of radiation.
Materials and methods
Patients
The data analysed in this study were derived from 108 consecutive breast cancer patients who received radiotherapy and were followed up for 7 years within our departmental program for the predictive testing of the radiosensitivity of normal tissue. The investigation was approved by the local ethics committee, and written, informed consent was obtained from all patients. Patient recruitment started in March 1996. Late adverse skin effects were measured between December 2003 and June 2004. The study design and patient and treatment characteristics have been published previously [9].
The patients were treated with postoperative radiation therapy after mastectomy (54 patients) or with breast-conserving therapy using a standardized 60Co technique (54 patients). The dose delivered was 50 Gy over a period of 5 weeks, in daily fractions of 2 Gy (25 fractions at 5 per week). External radiation was delivered by the cobalt unit in almost all of the patients (98%), and only 2% received irradiation from a Linac 6-MV x-ray linear accelerator. The whole breast or chest wall was irradiated by two parallel, opposed tangential fields, with wedges used to correct dose inhomogeneities. The dose was prescribed at the ICRU (International Commission on Radiation Units and Measurements) point at the midline of the central axis. Dose homogeneity was more than 85% in the majority of the cases. Patient treatments were planned using computed tomography images and a conventional simulator. To administer regional nodal radiation, we used a direct anterior field to irradiate internal mammary nodes. Supraclavicular and axillary lymph node areas were treated by irradiation of the axillary–supraclavicular field and the posterior axillary field. The total dose was calculated at 3 cm in the supraclavicular area and at the midplane in the axilla. The conservatively treated patients also received a tumour bed boost of 16 to 25 Gy using an iridium implant (192Ir), always 15 days after external radiotherapy or electron beam therapy. The 192Ir implants were done in accordance with the rules of the Paris System of Dosimetry. The dose was calculated at the reference isodose, defined as 85% of the basal dose calculated in the central plane of application. The total dose delivered by 9-to 12-MeV electron beams was 16 Gy at 2 Gy per fraction. The dose was prescribed to the 90% isodose line. A bolus was sometimes used to optimize the homogeneity of dose distribution.
All medical records of these 108 patients were available and were reviewed. Patient files included details of surgery, clinical-pathological stage, adjuvant treatment, and the subsequent follow-up. The records also included full details of the radiotherapy treatment, and a photograph of the irradiated field was always made on the last treatment day to record the intensity of the acute radiation-induced injury on the skin of each patient.
Definitions of descriptive terms for skin reactions
The severity of skin reactions was assessed by means of a simple scale (Table 1), using scores based on the absolute side-effect scale proposed by the Radiation Therapy Oncology Group/European Organization for Research and Treatment of Cancer [9], adapted here to the nomenclature proposed by Burnet and colleagues [4] in order to facilitate communication among groups studying normal tissue radiosensitivity. The term 'normal range' refers herein to the range of normal tissue reactions observed in typical radiotherapeutic clinics that treat large numbers of patients without genetic syndromes. All of the skin reactions observed in our study fell within the normal range, and no over-reactors were found.
Radiosensitivity assay
Initial radiation-induced DNA damage in peripheral blood lymphocytes was measured as described elsewhere [17,18] and was considered an indicator of the molecular radiosensitivity of the normal cells studied. Early and late skin side effects were assessed as mentioned above.
Early side-effect data
The unit of analysis was a group of 108 patients treated with radiotherapy for curative purposes after breast surgery. The most frequent acute complications found were erythema (91.7%), dry desquamation (29.6%), and moist desquamation (35.2%). According to the score system summarized in Table 1, approximately 13% of patients were classified as highly radiosensitive. Early side effects on the skin might be considered an indicator of clinical radiation sensitivity, and their intensity, score, and distribution have been previously described [9].
Late side-effect data
Although a significant proportion of the variation in response of normal tissues could be attributed to treatment-related factors, our results showed that dose effects were not sufficient to explain the differences between patients in their skin response (data not shown). Our team previously reported an adequate correlation between scoring of radiation-induced acute skin effects by direct observation and scoring after examination of photographic images, supporting the accuracy of the direct observation of lesions of normal tissue. Therefore, this direct-observation method was used for the assessment of late normal tissue changes in the 60 patients studied, as follows: on the day programmed for the late follow-up, a single physician (EL) generated a report based on direct clinical observation of the whole treated skin, scoring the degree of reaction on the scale used (Table 1, Fig. 1).
Comparison of in vitro and in vivo results
A two-sided Student's t-test was used to compare mean values of initial radiation-induced DNA damage between the patient groups. Contingency tables and the χ2 test were used to assess any relation between early and late effects.
The relations between in vivo and in vitro results were studied using a nonparametric regression test, and Spearman's ρ correlation coefficient was calculated. The Statistical Package for Social Sciences (SPSS 11.5) was used for all data processing. Graphics and basic biostatistics were obtained using Graphpad (GraphPad Software Inc, San Diego, CA, USA).
Results
Radiosensitivity test
Initial radiation-induced DNA damage was determined in lymphocytes from 108 breast cancer patients after γ-irradiation. The parameter selected was the estimated number of dbs per Gy and per DNA unit [21]. It should be noted that the results obtained from the reference sample of patients included in this paper matched the results obtained in lymphocytes from other breast cancer patients analysed at our laboratory in ongoing studies. [6]. The mean value ± the standard error of the mean was 1.83 ± 0.18 double-strand breaks per Gy.
Early radiation-induced injury
Assessment of clinical radiation sensitivity was based on the acute skin reactions to the radiotherapy measured [9] on the last day of treatment. Five patients (4.6%) with no adverse side effects were classified as highly radioresistant; 36 (33.3%), 44 (40.7%), and 10 patients (9.3%) with mild to moderate skin reactions were classified as, respectively, moderately radioresistant, average, and moderately radiosensitive; and 13 patients (12%) with pronounced signs of radiation acute sensitivity were considered highly radiosensitive (Fig. 1). The correspondence between the descriptive terms and the radiation sensitivity data is summarized in Table 1. Acute effects on the skin included in the treatment field, such as erythema or desquamation, normally resolve rapidly in most patients. Individual variation in the level of normal tissue response could be theoretically interpreted by the classical sigmoid dose–response curve. Comparison of collateral effects between the surgical treatment subgroups (mastectomy versus breast-conserving therapy) showed that radiation-induced acute toxicity on the skin of the breast cancer patients has the same frequency and intensity regardless of the surgical approach, even when the use of concurrent chemotherapy was taken into consideration [9].
Overall survival and actuarial probabilities of normal tissue sequelae
Data of survival and late morbidity records were obtained for 87 patients who had undergone radiotherapy treatment for >7 years, of whom 51 were free of cancer disease; 9 were living with disease, 22 had died, and 5 who had received reconstructive surgery were not assessed. A total of 21 patients were missing from the follow-up. Seven years after treatment, the actuarial overall survival of the whole series of breast cancer patients was 48.84 ± 7.62% (mean ± standard error of the mean).
The actuarial probabilities of late radiation side effects, expressed as percentages ± standard errors of the mean, were 10.19 ± 2.91 for highly radioresistant; 10.19 ± 2.91 for moderately radioresistant, 21.30 ± 3.94 for average, 12.96 ± 3.23 for moderately radiosensitive, and 0.0 for highly radiosensitive. Fig. 1 depicts the distribution of the frequency of acute and late effects according to the severity. The distributions of the severity of early and late effects differed. Statistical comparison between early and late collateral effects in the same group of patients gave a χ2 value of 22.38 (P = 0.0002), demonstrating a very different distribution frequency between radiation-induced acute toxicity and radiation-related late morbidity.
Correlation between radiobiological test and early radiation skin side effects
The distribution of early normal tissue reactions observed in this study could be considered approximately normal in shape (Fig. 1). The distribution of the lymphocyte radiosensitivity measured in vitro could also be considered approximately Gaussian [6,18]. This similarity prompted us to examine whether the same relation could be found between the number of initial radiation-induced DNA double-strand breaks and the severity of acute adverse skin effects. No relation was found (Fig. 2) between the molecular radiosensitivity values in lymphocytes and the early normal tissue reactions observed in vivo (Spearman ρ = 0.076; 95% confidence interval, -0.149 to 0.293; two-tailed P = 0.497).
Correlation between radiobiological test and late radiation skin side effects
The distribution of late adverse effects observed in these patients does not appear to fit a Gaussian distribution (Fig. 1), and no statistical relation was found between the radiosensitivity test results and the late effects assessed (Fig. 3). No significant relation was found between the in vitro assay results and the severity of late side effects (Spearman ρ = 0.063; 95% confidence interval, -0.219 to 0.335; two-tailed P = 0.655). Considering the patients with tolerable late effects (highly or moderately radioresistant or with average radioresistance) separately from those with more severe effects (moderately or highly radiosensitive) in a scatter plot, it appears (Fig. 4) that the molecular radiosensitivity assay did not distinguish patients at different levels of risk of developing more severe late skin reactions after radiotherapy treatment.
Correlation between early and late skin effects
The data on the severity of early and late adverse effects after radiotherapy for breast cancer showed no relation between these toxic effects (Fig. 5). According to our results, acute and late radiation-related morbidities are independent adverse effects, (Spearman ρ = 0.032; 95% confidence interval, -0.233 to 0.293; two-tailed P = 0.809).
Correlation between early and late effects and treatment volume
It has classically been reported that patient skin tolerance may be lower with larger breast size. In the present study, this relation was studied in a group of patients treated with breast-conserving surgery, estimating the breast volume from the bra size. When acute adverse effects were considered in 47 patients, a positive relation was found (Spearman ρ = 0.497; 95% confidence interval, 0.236 to 0.691; two-tailed P < 0.001) (Fig. 6). However, the relation was weaker when late side effects were considered (Spearman ρ = 0.423; 95% confidence interval, 0.070 to 0.682; two-tailed P = 0.018), perhaps because of the smaller number of cases (n = 31) analysed (Fig. 7).
Discussion
In this study, early and late complications in normal tissue were assessed at an arbitrary single point. In this situation, a relative scale of normal tissue reactions, such as the score system proposed by Burnet [4], has a number of advantages over an absolute one. The main objectives of our study were to identify patients with extreme reactions within the normal range and to compare the results of an in vitro radiosensitivity test with the severity of acute and late reactions in the same patients. By using this relative scale, we were able to meet these objectives. The concept of the predictive testing of normal tissue reactions in order to individualize radiotherapy prescriptions is founded on a hypothetical relation between the radiosensitivity of cells and that of normal tissue. Although we are inclined to support this hypothesis, the test applied in the present study, based on the initial radiation-induced DNA damage, proved inadequate for use in the individualization of radiotherapy therapy.
Early effects such as erythema and desquamation usually appear during or immediately after radiotherapy therapy, whereas late effects develop some years afterwards. The acute side effects resolve rapidly without treatment [11]. However, in a substantial group of patients, radiation-induced fibrosis, telangiectasia, and skin pigmentation disorders appear at different times after radiotherapy. Generally, the course of radiation sequelae follows a distinct clinical pattern. An erythematous rash can develop on the skin of treated patients within a few hours of exposure and can persist or slowly worsen until the end of radiotherapy treatment. This situation is transient in nature. In severe cases, subepidermal blisters and ulcers may develop. Most of the injuries heal, although the expression of radiation-induced effects can reappear in some individuals after a latency period. Late damage becomes more severe, progresses with time, and usually cannot be halted or reversed [22]. The inability to predict the length of the latency period creates a major problem for the management of these patients. A better understanding of individual variations in normal tissue reactions, which determine tolerance, may allow the individualization of radiotherapeutic prescriptions and improve outcomes. The lag time to the onset of initial late effects might be expected to yield information on the mechanisms underlying the development of late radiation sequelae. Extreme side effects of radiotherapy, including an increased cancer risk after radiation, were observed in patients with inherited disorders such as ataxia-telangiectasia and Nijmegen syndrome [16]. According to the present results, there appears to be no mechanistic relation between the early and late adverse effects of radiation treatment. We speculate that these differences may arise because the healing of acute injuries is a deterministic process whereas late side effects may be stochastic phenomena.
Unconventional, more aggressive irradiation protocols are usually associated with an aggravation of acute reactions that might be related to more severe late effects. Therefore, amelioration of the acute response to radiation has been proposed as a useful approach to minimize late side effects of effective radiation therapy. This proposal assumes a relation between acute and late effects via a non-healing acute response component that directly progresses to a late effect [23]. However, the present results do not support the hypothesis that late effects in normal tissue can be predicted from the acute reactions observed in the same patients.
It also proved impossible in the present study to predict acute or late effects from the results of an in vitro assay to measure initial radiation-induced DNA damage. Until recently, it has been generally accepted that the genotoxic consequences of radiation exposure derive from the damage inflicted directly by radiation, producing irreversible changes during DNA replication or cell division or during the processing of DNA damage by enzymatic repair processes [24]. However, there is now considerable evidence that cells that are the progeny of exposed cells but that are not themselves exposed may divide, express delayed gene mutations, and carry chromosomal aberrations. This effect, known as radiation-induced genomic instability, may be expressed via delayed lethal mutations [25], causing prolonged perturbation of tissue volume within the radiation field [26]. Although the mechanisms of those delayed effects of ionizing radiation are unclear, excessive production of reactive oxygen species has been implicated [27]. Recent experiments showed that macrophage activation and neutrophil infiltration are consequences of the recognition and clearance of radiation-induced apoptotic cells and that increased phagocytic cell activity persists after removal of apoptotic bodies. It was demonstrated, contrary to expectations, that the recognition and clearance of apoptotic cells after exposure to radiation produces persistent macrophage activation and a genotype-dependent inflammatory-type response [28]. These phenomena and radiation-induced genetic changes may be important determinants of the longer-term consequences of radiation exposure [28]. Moreover, new evidence suggests that cytokine-mediated multicellular interactions initiate and sustain the fibrogenic process [29,30] that is a long-term effect of radiotherapy.
Initial DNA damage and post-radiation cell survival after radiation have been directly related in in vitro experiments [31]. The present findings indicated that the level of radiation-induced DNA damage in normal cells was not a major determinant of the severity of early skin injury. Moreover, no relation was found between the acute injuries and the late sequelae that, after an undetermined latency period, became a burden, lessening the quality of life of these patients [32].
However, a significant correlation has been demonstrated, using new methodologies, between five single-nucleotide polymorphisms (SNPs) and the risk of radiation-induced normal tissue reactions in a small group of breast cancer patients [33]. In fact, the completion of the human genome project and the availability of novel and powerful technologies in genomics, proteomics, and functional genomics promise to have a major impact on clinical practice. These developments are likely to change the way in which diseases will be diagnosed, treated, and monitored in the near future. Cancer, as a complex disease that affects a significant proportion of world population, has become a prime target of novel technologies, often referred to as 'omic' platforms, and it is anticipated that progress will be made towards a predictive, individualized approach to cancer care. One area of knowledge where advances are expected is on the complex variability in normal tissue radiation response, which depends on the interaction of multiple gene products. There is a growing shift from the study of single parameters of molecular or cellular radiosensitivity to the analysis of complex biological systems, and one of the main challenges we face is how best to apply the 'omic' technologies to clinically relevant samples in a well-defined clinical and pathological framework. An example of this type of venture is the European GENEPI project [34], which aims to study a large cohort of patients under highly controlled and standardized radiotherapy conditions.
Conclusion
Our first conclusion is an experimental one. These results do not support the hypothesis that the response of normal tissue to radiation can be predicted by an in vitro test. This conclusion was reached by other authors [8,10], although some results in defence of this hypothesis have also been published [35,36]. A possible explanation is that in vitro cellular radiosensitivity tests and molecular DNA damage assays do not take account of the variable degree of cytokine response, tissue remodeling, and collagen deposition that may characterize the specific normal-tissue response of each patient [29]. The paradigm that radiotherapy effects are restricted to the direct or indirect effects of radiation-induced DNA damage is challenged by the present results, which indicate that early and late effects can also be induced by unexpected interactions between irradiated and nonirradiated cells (bystander effects). This conclusion is supported by published results that showed a clear relation between the severity of late toxicity in radiotherapy treatment and the volume of normal tissue included in the field of treatment, although a significant correlation was found between breast size and dose inhomogeneities that may account for the marked changes in breast appearance reported in women with large breasts [37].
Our second conclusion is a theoretical one, and takes the form of a proposal to change the model adopted in radiobiological studies to date. Thus, for teaching and research purposes, 'direct action' could be defined as all physicochemical processes that occur after energy cession from the ionizing radiation to the tissues. Within this concept would be included actions produced by free radicals that result from the interaction of radiation with the water molecules – that is, the effects hitherto designated indirect radiation action on the DNA molecule. The cellular consequences of the direct action of radiation in terms of lethal and potentially lethal damage to DNA can be explained by linear-quadratic radiation cell survival models. However, these models cannot explain the late adverse effects of radiation, and a more general theory appears to be required.
A few days after the end of radiation treatment, cells within the irradiated volume can act in one of three ways: they can grow and divide, the basis for the healing of acute injuries; they can not grow but stay alive; or they may survive for a long time with important immunological changes, disappearing by apoptosis or apoptosis-like cell death very slowly and becoming a chronic focus of immunological system stimulation that could produce the late actions observed. Therefore, indirect action could be considered the whole immunological response of the body to the stress induced by radiation in the target volume. This may produce late side effects of varying severity that in a stochastic fashion, through a time-dependent probability relation, could lead to a lifelong risk of developing late complications [14,32,38]. In this relation, the volume of tissue irradiated may be a multiplicity constant of the frequency and severity of the late side effects. Patients and clinicians should be aware of these aspects of radiotherapy therapy. The study of these immunological changes is complex but could, given the human genome data now available, offer a key to improving radiotherapy outcomes in cancer patients.
Finally, our group supports the view that the risks of radiotherapy can be fully understood only after long-term follow-up studies. An important research aim is to develop a test that can predict late side effects.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EL was significantly involved in patient recruitment, patient treatment, and assessment of late normal tissue responses; contributed to the correlation of clinical data with experimental findings; and took a role in supervising the final version of article. RG and RdM were significantly involved in patient recruitment and patient treatment and in the assessment of skin reactions. JMG participated in patient treatment. MIN, MV, and MTV carried out the in vitro radiosensitivity test and took a role in the discussion of results. JAMG, DMO, and FJO carried out part of the laboratory work. JMRdA conceived and designed the study, interpreted the data, and revised the paper, giving final approval of the version to be submitted. All the authors read and approved the final manuscript.
Acknowledgements
This work was supported by grants from the Ministerio de Ciencia y Tecnología CICYT (SAF 2001-3533) and Ministerio de Educación y Ciencia CICYT (SAF 2004-00889). DMO and JAMG were supported by fellowships (BEFI 01/9331, BEFI 02/9029) from the Fondo de Investigaciones Sanitarias (ISCIII).
Figures and Tables
Figure 1 Frequency distribution of skin reactions in women with breast cancer treated postoperatively with radiotherapy. Reactions were classified as early (if observed at the end of the radiotherapy, 108 women) or late (if observed at the 7-year follow-up, 60 women). X-axis Radiation Therapy Oncology Group scoring system modified using the terminology proposed by Burnet (4): A, average; HRR, highly radioresistant; HRS, highly radiosensitive; MRR, moderately radioresistant; MRS, moderately radiosensitive.
Figure 2 Relation between severity of early radiotherapy-induced skin morbidity and lymphocyte molecular radiosensitivity. Skin morbidity in 108 women was assessed on the treated skin using the scoring system summarized in Table 1. Lymphocyte molecular radiosensitivity was measured as DNA double-strand breaks (dsb) by dose unit (Gy) and DNA unit (200 Mbp). bp, base pairs.
Figure 3 Relation between severity of late radiotherapy-induced skin morbidity and lymphocyte molecular radiosensitivity. Skin morbidity in 60 women was assessed on the treated skin using the scoring system summarized in Table 1. Lymphocyte molecular radiosensitivity was measured as DNA double-strand breaks (dsb) by dose unit (Gy) and DNA unit (200 Mbp). bp, base pairs.
Figure 4 Scatter-plot of quantified late skin reactions of patients and the corresponding in vitro radiosensitivity values. Horizontal solid lines are the mean values for each group of patients with breast cancer. Moderate (n = 38), women with reactions scored as highly or moderately radioresistant or as having an average response; severe (n = 12), patients with reactions scored as highly radiosensitive.
Figure 5 Relation between the severity of early and late side effects of radiotherapy for breast cancer. Women with breast cancer (n = 60) were treated postoperatively with radiotherapy and assessed using the scoring system summarized in Table 1.
Figure 6 Relation between estimated irradiation volumes and severity of early effects of radiotherapy for breast cancer. Irradiation volumes were estimated from the women's bra size, and the severity of the early effects were scored in the same women (n = 50). Dotted line shows the corresponding regression line (P < 0.001).
Figure 7 Relation between estimated irradiation volumes and severity of late effects of radiotherapy for breast cancer. Irradiation volumes were estimated from the women's bra size, and the late effects were scored in the same women (n = 33). Dotted line shows the corresponding regression line (P = 0.018).
Table 1 Scoring system used to document cutaneous and subcutaneous reactions in breast cancer patients receiving radiotherapy
Grade Early reactionsa Late reactionsb
Grade 0 – Highly radioresistant patients No toxicity observed: no erythema, desquamation, or pain Absence of differences between irradiated and nonirradiated skin
Grade 1 – Moderately radioresistant patients Faint, dull, or bright erythema, psilosis, dry desquamation, mild oedema Minimal telangiectasia, slight breast asymmetry, mild hyperpigmentation
Grade 2 – Patients with average radiosensitivity Severe erythema, at least one moist desquamation of small size, moderate oedema Marked telangiectasia, moderate hyperpigmentation, increased density and palpable firmness, mild oedema
Grade 3 – Moderately radiosensitive patients Severe or confluent moist desquamation Partially confluent telangiectasia, severe hyperpigmentation, severe oedema, subcutaneous fibrosis with fixation
Grade 4 – Highly radiosensitive patients Ulceration, haemorrhage Totally confluent telangiectasia, very marked density, retraction and fixation. Major aesthetic sequelae in treated breast
aMorbidity assessed at the end of radiotherapy treatment. bMorbidity assessed at end of 7-year follow-up period
==== Refs
Johansson S Svensson H Denenkamp J Dose response and latency for radiation induced fibrosis, edema and neuropathy in breast cancer patients Int J Radiat Oncol Biol Phys 2002 52 1207 1219 11955731 10.1016/S0360-3016(01)02743-2
Van de Steene J Vinh Hung V Cutuli B Storme G Adjuvant radiotherapy for breast cancer: effects of longer follow-up Radiother Oncol 2004 72 35 43 15236872 10.1016/j.radonc.2004.04.004
Van de Steene J Soete G Storme G Adjuvant radiotherapy for breast cancer significantly improves overall survival: The missing link Radiother Oncol 2000 55 263 272 10869741 10.1016/S0167-8140(00)00204-8
Burnet NG Johansen J Turesson I Nyman J Peacock JH Describing patients' normal tissue reactions: concerning the possibility of individualising radiotherapy dose prescriptions based on potential predictive assays of normal tissue radiosensitivity. Steering Committee of the BioMed 2 European Union Concerted Action Programme on the Development of Predictive Tests of Normal Tissue Response to Radiation Therapy Int J Cancer 1998 79 606 613 9842969 10.1002/(SICI)1097-0215(19981218)79:6<606::AID-IJC9>3.0.CO;2-Y
Brahme A Individualizing cancer treatment: biological optimization models in treatment planning and delivery Int J Radiat Oncol Biol Phys 2001 49 327 337 11173125 10.1016/S0360-3016(00)01501-7
Guirado D Ruiz de Almodóvar JM Prediction of normal tissue response and individualization of doses in radiotherapy Phys Med Biol 2003 48 3213 3233 14579861 10.1088/0031-9155/48/19/008
Turesson I Nyman J Holmberg E Oden A Prognostic factors for acute and late skin reactions in radiotherapy patients Int J Radiat Oncol Biol Phys 1996 36 1065 1075 8985028 10.1016/S0360-3016(96)00426-9
Barber JB Burrill W Spreadborough AR Levine E Warren C Kiltie AE Roberts SA Scott D Relationship between in vitro chromosomal radiosensitivity of peripheral blood lymphocytes and the expression of normal tissue damage following radiotherapy for breast cancer Radiother Oncol 2000 55 179 186 10799730 10.1016/S0167-8140(99)00158-9
López E Núñez MI Guerrero MR del Moral R de Dios Luna J del mar Rodríguez MM Valenzuela MT Villalobos M Ruiz de Almodóvar JM Breast cancer acute morbidity evaluated by different scoring systems Breast Cancer Res Treat 2002 73 127 134 12088115 10.1023/A:1015296607061
Peacock JH Ashton A Bliss J Bush C Eady J Jackson C Owen R Regan J Yarnold J Cellular radiosensitivity and complication risk after curative radiotherapy Radiother Oncol 2000 55 173 178 10799729 10.1016/S0167-8140(00)00173-0
Kurtz J The curative role of radiotherapy in the treatment of operable breast cancer Eur J Cancer 2002 38 1961 1974 12376199 10.1016/S0959-8049(02)00314-3
del Moral R López E Núñez MI Oliver FJ Valenzuela MT Villalobos M Ruiz de Almodóvar JM Interactions between radiotherapy and endocrine therapy in breast cancer Endocr Relat Cancer 2002 9 197 205 12237247 10.1677/erc.0.0090197
Vicini FA Baglan KL Kestin LL Mitchell C Chen PY Frazier RC Edmunson G Goldstein NS Benitez P Huang RR Martinez A Accelerated treatment of breast cancer J Clin Oncol 2001 19 1993 2001 11283132
Johansson S Svensson H Denenkamp J Timescale of evolution of late radiation injury after postoperative radiotherapy of breast cancer patients Int J Radiat Oncol Biol Phys 2000 48 745 750 11020571 10.1016/S0360-3016(00)00674-X
Amichetti M Caffo O Arcicasa M Roncadin M Lora D Rigon A Zini G Armaroli L Coghetto P Zorat P Quality of life in patients with ductal carcinoma in situ of breast treated with conservative surgery and postoperative irradiation Breast Cancer Res Treat 1999 54 109 115 10424401 10.1023/A:1006125602353
Gatti RA The inherited basis of human radiosensitivity Acta Oncol 2001 40 702 711 11765064 10.1080/02841860152619115
Núñez MI Guerrero R López E del Moral R Valenzuela MT Siles E Villalobos M Pedraza V Peacock JH Ruiz de Almodóvar JM DNA damage and prediction of radiation response in lymphocytes and epidermal skin cells Int J Cancer 1998 76 354 361 9579572 10.1002/(SICI)1097-0215(19980504)76:3<354::AID-IJC12>3.0.CO;2-B
Ruiz de Almodóvar JM Guirado D Núñez MI López E Guerrero R Valenzuela MT Villalobos M Ruiz de Almodóvar JM Individualisation of radiotherapy in breast cancer patients: possible usefulness of a DNA damage assay to measure normal cell radiosensitivity Radiother Oncol 2002 62 327 333 12175564 10.1016/S0167-8140(01)00490-X
Carlomagno F Burnet NG Turesson I Nyman J Peacock JH Dunning AH Ponder BA Jackson SP Comparison of DNA repair protein expression and activities between human fibroblast cell lines with different radiosensitivities Int J Cancer 2000 85 845 849 10709106 10.1002/(SICI)1097-0215(20000315)85:6<845::AID-IJC18>3.0.CO;2-C
Krause M Hessel F Szips D Hilberg F Baumann M Adjuvant inhibition of the epidermal growth factor receptor after fractionated irradiation of FaDu human squamous cell carcinoma Radiother Oncol 2004 72 95 101 15236881 10.1016/j.radonc.2004.05.003
Ruiz de Almodóvar JM Steel GG Whitaker SJ McMillan TJ A comparison of methods for calculating DNA double-strand break induction frequency in mammalian cells by pulsed-field gel electrophoresis Int J Radiat Biol 1994 65 641 649 7912713
Johansson S Svensson H Denenkamp J Timescale of evolution of late radiation injury after postoperative radiotherapy of breast cancer patients Int J Radiat Oncol Biol Phys 2000 48 745 750 11020571 10.1016/S0360-3016(00)00674-X
Dörr W Hendry JH Consequential late effects in normal tissues Radiother Oncol 2001 61 223 231 11730991 10.1016/S0167-8140(01)00429-7
McMillan TJ Tobi S Mateos S Lemon C The use of DNA double-strand break quantification in radiotherapy Int J Radiat Oncol Biol Phys 2001 49 373 377 11173130 10.1016/S0360-3016(00)01467-X
Gorgojo L Little J Expression of lethal mutations in progeny of irradiated mammalian cells Int J Radiat Biol 1989 55 619 630 2564871
Azzan EI Little JB The radiation-induced bystander effects: evidence and significance Hum Exp Toxicol 2004 23 61 65 15070061 10.1191/0960327104ht418oa
Azzan EI de Toledo SM Little JB Oxidative metabolism, gap junctions and the ionizing radiation-induced bystander effect Oncogene 2003 22 7050 7057 14557810 10.1038/sj.onc.1206961
Lorimore SA Coates PJ Scobie GE Milne G Wright EG Inflammatory-type responses after exposure to ionizing radiation in vivo: A mechanism for radiation-induced bystander effects? Oncogene 2001 20 7085 7095 11704832 10.1038/sj.onc.1204903
Dickson J Magee B Stewart A West CM Relationship between residual radiation-induced DNA double strand breaks in cultured fibroblasts and late radiation reactions: a comparison of training and validation cohorts of breast cancer patients Radiother Oncol 2002 62 321 326 12175563 10.1016/S0167-8140(01)00432-7
Herskind C Johansen J Bentzen SM Overgaard M Overgaard J Bamberg M Rodemann HP Fibroblast differentiation in subcutaneous fibrosis after postmastectomy radiotherapy Acta Oncol 2000 39 383 388 10987235 10.1080/028418600750013159
Núñez MI McMillan TJ Valenzuela MT Ruiz de Almodóvar JM Pedraza V Relationship between DNA damage, rejoining and cell killing by radiation in mammalian cells Radiother Oncol 1996 39 155 165 8735483 10.1016/0167-8140(96)01732-X
Fehlauer F Tribius S Höller U Rades D Kuhlmey A Bajorovic A Alberti W Long-term radiation sequelae after breast conserving therapy in women with early-stage breast cancer: an observational study using the LENT-SOMA scoring system Int J Radiat Oncol Biol Phys 2003 55 651 658 12573752 10.1016/S0360-3016(02)04120-2
Andreassen CN Alsner J Overgaard M Overgaard J Prediction of normal tissue radiosensitivity from polymorphisms in candidate genes Radiother Oncol 2003 69 127 135 14643949 10.1016/j.radonc.2003.09.010
Baumann M Hölscher T Begg AC Towards genetic prediction of radiation responses: ESTRO's GENEPI project Radiother Oncol 2003 69 121 125 14643948 10.1016/j.radonc.2003.08.006
West CM Davidson SE Elyan SA Valentine H Roberts SA Swindell R Hunter RD Lymphocyte radiosensitivity is a significant prognostic factor for morbidity in carcinoma of the cervix Int J Radiat Oncol Biol Phys 2001 51 10 15 11516845 10.1016/S0360-3016(01)01575-9
Hoeller U Borgmann K Bonacker M Kuhlmey A Bojrovic A Jung H Alberti W Dikomey E Individual radiosensitivity measured with lymphocytes may predict the risk of fibrosis after radiotherapy for breast cancer Radiother Oncol 2003 69 137 144 14643950 10.1016/j.radonc.2003.10.001
Moody AM Mayles WPM Bliss JM A'Hern PP Owen JR Regan J Broad B Yarnold JR The influence of breast size on late radiation effects and association with radiotherapy dose inhomogeneity Radiother Oncol 1994 33 106 112 7708953 10.1016/0167-8140(94)90063-9
Jung H Beck-Borholdt HP Svoboda V Alberti W Herrmann T Quantification of late complications after radiation therapy Radiother Oncol 2001 61 233 246 11730992 10.1016/S0167-8140(01)00457-1
|
16168114
|
PMC1242135
|
CC BY
|
2021-01-04 16:54:51
|
no
|
Breast Cancer Res. 2005 Jul 1; 7(5):R690-R698
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1277
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12781616811910.1186/bcr1278Research ArticleThe expression of the ubiquitin ligase subunit Cks1 in human breast cancer Slotky Merav [email protected] Ma'anit [email protected] Ofer [email protected] Shai [email protected] Boris [email protected] Medy [email protected] Dan D [email protected] Department of Surgery A, Rambam Medical Center and the Technion–Israel Institute of Technology, Haifa, Israel2 Department of Pathology, Rambam Medical Center and the Technion–Israel Institute of Technology, Haifa, Israel3 Department of Medical Oncology, Rambam Medical Center and the Technion–Israel Institute of Technology, Haifa, Israel4 Unit of Clinical Epidemiology, Rambam Medical Center and the Technion–Israel Institute of Technology, Haifa, Israel5 Breast Health Institute, Rambam Medical Center and the Technion–Israel Institute of Technology, Haifa, Israel2005 19 7 2005 7 5 R737 R744 4 4 2005 4 5 2005 31 5 2005 Copyright © 2005 Slotky et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Loss of the cell-cycle inhibitory protein p27Kip1 is associated with a poor prognosis in breast cancer. The decrease in the levels of this protein is the result of increased proteasome-dependent degradation, mediated and rate-limited by its specific ubiquitin ligase subunits S-phase kinase protein 2 (Skp2) and cyclin-dependent kinase subunit 1 (Cks1). Skp2 was recently found to be overexpressed in breast cancers, but the role of Cks1 in these cancers is unknown. The present study was undertaken to examine the role of Cks1 expression in breast cancer and its relation to p27Kip1 and Skp2 expression and to tumor aggressiveness.
Methods
The expressions of Cks1, Skp2, and p27Kip1 were examined immunohistochemically on formalin-fixed, paraffin-wax-embedded tissue sections from 50 patients with breast cancer and by immunoblot analysis on breast cancer cell lines. The relation between Cks1 levels and patients' clinical and histological parameters were examined by Cox regression and the Kaplan–Meier method.
Results
The expression of Cks1 was strongly associated with Skp2 expression (r = 0.477; P = 0.001) and inversely with p27Kip1 (r = -0.726; P < 0.0001). Overexpression of Cks1 was associated with loss of tumor differentiation, young age, lack of expression of estrogen receptors and of progesterone receptors, and decreased disease-free (P = 0.0007) and overall (P = 0.041) survival. In addition, Cks1 and Skp2 expression were increased by estradiol in estrogen-dependent cell lines but were down-regulated by tamoxifen.
Conclusion
These results suggest that Cks1 is involved in p27Kip1 down-regulation and may have an important role in the development of aggressive tumor behavior in breast cancer.
==== Body
Introduction
The prognosis and clinical management of patients with breast cancer are commonly determined by traditional clinicopathological factors such as tumor size and grade, lymph node status, and the expression of receptors to estrogen (ERs) and to progesterone (PRs) and of Her2/neu [1]. Nevertheless, patients may have significantly different clinical outcomes despite similar clinicopathological features. Some of these differences may be attributed to alterations in the normal regulation of the cell cycle that ultimately lead to aggressive tumor behavior.
Among the various cell-cycle proteins, deregulation of p27Kip1 expression was found to have a particularly important role in cancer [2]. Numerous studies have shown that down regulation of p27Kip1, an inhibitor of cyclin-dependent kinases, is associated with poor prognosis in many cancers, including breast, colorectal, prostate, and lung carcinomas [3-8]. The decrease in p27Kip1 levels in these cancers was found to result from its rapid degradation by the ubiquitin–proteasome pathway, rather than from decreased protein synthesis or gene mutation [6,9,10].
The main rate-limiting regulator for p27Kip1 degradation was identified as an SCF-type ubiquitin ligase complex that contains S-phase kinase protein 2 (Skp2) as the specific substrate-recognition subunit [11-13]. Skp2 specifically binds p27Kip1 and targets it for degradation by the ubiquitin proteolytic system. The important role of Skp2 in controlling p27Kip1 levels in some human cancers, including breast, prostate, colorectal, and oral squamous cell carcinomas, was recently emphasized [14-17]. Thus, increased expression of Skp2 in these tumors was associated with low p27Kip1 levels, aggressive tumor behavior, and poor overall survival.
More recently, the essential role of cyclin kinase subunit 1 (Cks1) in facilitating the ubiquitin-mediated proteolysis of p27Kip1 through interaction with Skp2 was discovered [18,19]. Cks1 is a member of the highly conserved family of Cks/Suc1 (Schizosaccharomyces pombe cell-cycle regulatory protein) proteins, which interact with Cdks (cyclin-dependent kinase), but its exact mechanism of action remained poorly understood until recently. Cks1 was found to be an essential cofactor for efficient Skp2-dependent ubiquitination of p27Kip1. The critical role of Cks1 in targeting p27Kip1 for efficient degradation by Skp2 was emphasized by demonstrating the lack of p27Kip1 ubiquitination and breakdown in the absence of Cks1 in vitro and the slow proliferation and accumulation of p27Kip1 in Cks1 nullizyous mice in vivo [18,19]. Recent studies from our and other laboratories have investigated the expression of Cks1 levels in different human cancers and its relation to Skp2 and p27Kip1 expression [20-24]. For example, we have shown that in colorectal cancer the expression of Cks1 protein levels correlated strongly with the expression of Skp2 protein levels and inversely with those of p27Kip1 [23]. Furthermore, Cks1 expression was increased in poorly differentiated tumors and was strongly and independently associated with poor overall survival [24]. The role of Cks1 in breast cancer, however, is unknown.
In the present study, we investigated the expression of Cks1 in relation to Skp2 and p27Kip1 and prognosis in breast cancer. We show that Cks1 expression is strongly associated with Skp2 expression and inversely associated with p27Kip1 expression. Furthermore, our results suggest that Cks1 expression may be used as an independent prognostic marker for disease-free and overall survival in breast cancer.
Materials and methods
Patients and tissue samples
Tissue samples from 50 patients with primary breast carcinomas that had been operated on more than 6 years before the beginning of the study were collected for immunohistochemical studies, after we had obtained the approval of the institution's Human Investigation Committee. Patients presenting with metastatic disease were excluded from this study. Clinical and histological data were available for all of these patients. Long-term follow-up data were provided by patients' medical records and the Israel Cancer Registry.
Cell cultures
Human breast cancer cell lines (MCF-7, T47D, and MDA-MB-231) were generously provided by Dr H Degani (Weizmann Institute of Science, Rehovot, Israel). Because Skp2 levels change during the cell cycle (being normally highest in the S phase and lowest in the G1 phase) we cultured the cells in different media under conditions of similar proliferation rate in all cell lines. MDA-MB-231 cells were grown in RPMI medium (Biological Industries, Beth Ha'emek, Israel) supplemented with 10% fetal calf serum, 100 Units penicillin, and 100 μg streptomycin per ml and 1 mM sodium pyruvate. MCF-7 breast carcinoma cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Biological Industries) supplemented with 10% fetal bovine serum, 4.5 g/l glucose, antibiotics as described above, and 4 mM glutamine. The T47D cell line was cultured in RPMI medium supplemented with 10% fetal calf serum, 100 Units/ml penicillin, 100 μg/ml streptomycin, and 10 μg/ml insulin. All cell lines were cultured at 37°C in 5% CO2. Under these conditions, the proliferation rate of all cell lines was similar (21 to 22 hours doubling time).
Immunoblotting
Samples containing 30 μg of protein were resolved by electrophoresis on a 12.5% SDS-polyacrylamide gel and were transferred to nitrocellulose membranes. The membranes were probed with affinity-purified rabbit polyclonal antibody directed against a 12-amino-acid synthetic peptide from the extreme C terminus of human Cks1 (gift of Dr A Hershko, Technion, Haifa, Israel). This antibody does not cross-react with Cks2. The antibody was diluted 1:200 in TBST (Tris-buffered saline and 0.1% Tween 20) containing 10% (w/v) nonfat dry milk. Membranes were also probed against the mouse monoclonal antibody directed against Skp2 (Zymed Laboratories Inc, South San Francisco, CA, USA) diluted 1:500, or with a rabbit polyclonal antibody directed against p27Kip1 (Zymed) diluted at 1:250. Membranes were also probed with a mouse monoclonal antibody directed against Skp1 (Transduction Laboratories, Lexington, KY, USA; 1:1000 dilution). Since levels of Skp1 do not change in the cell cycle, this protein served as an internal control to normalize for loading of cellular protein. After being washed with TBST, the immunoreactive proteins were visualized with horseradish-conjugated IgG (Pierce, Rockford, IL, USA) diluted 1:10,000 and an enhanced chemiluminescence system (SuperSignal West Pico; Pierce). All immunoblot analyses were done at least twice.
Immunohistochemistry
Immunohistochemical studies were performed on formalin-fixed tissue sections embedded in paraffin wax. Five-micron sections were deparaffinized with xylene and rehydrated in a series of ethanols.
For epitope retrieval, slides were heated in 1 mM ethylenediaminetetraacetic acid buffer (pH 8), either in a microwave oven at 92°C for 20 min (p27Kip1) or in an Antigen Retrieval Processor (Milestone Inc, Sorsiole, Italy) at 120°C for 8 min (Skp2 and Cks1). After they had cooled, the slides were washed in distilled water.
Skp2 staining was carried out in the NexES IHC Immunostainer (Ventana Medical Systems, Tucson, AZ, USA), in accordance with the manufacturer's instructions, using a monoclonal antibody (Zymed) diluted 1:100.
Slides for p27Kip1and Cks1 staining were treated for 10 min with 3% H2O2 in methanol to block endogenous peroxidase and for 30 min with 10% nonimmune rabbit serum to block nonspecific protein binding. The slides were then washed in water and soaked in washing buffer (pH 7.4, Optimax; Biogenex, San Ramon, CA, USA) for 5 min. For p27Kip1 staining, slides were incubated overnight at 4°C with the monoclonal p27Kip1 antibody diluted 1:500. Staining was completed with a Histostain-plus kit (Zymed) in accordance with the manufacturer's instructions. For Cks1 staining, slides were incubated with an affinity-purified rabbit polyclonal antibody directed against a 12-amino-acid synthetic peptide from the extreme C terminus of human Cks1 (described above) diluted 1:40 for 2 hours at room temperature. Staining was completed using a Super-sensitive Multilink Immunodetection kit (Biogenex) in accordance with the manufacturer's instructions. The color reaction product was developed with aminoethyl carbazole (AEC). All sections were counterstained with hematoxylin.
For negative controls, slide sections that were positive for staining were treated with 10% nonimmune rabbit serum (Zymed) instead of the primary antibody. No staining was observed in any of these controls.
The immunohistochemical slides were scored according to the percentage of tumor cells exhibiting nuclear staining. To define high and low protein expression, we used a cutoff of 50% p27Kip1 and 10% for Skp2, which is the cutoff used in most previous studies [14,15]. The cutoff for Cks1 was 10%, similar to the level used for Skp2, as in previous studies. When stained, cells displayed similar intensities of nuclear staining regardless of the percentage of cells stained, and therefore the intensity of staining was not included in the score. The specificity of immunohistochemistry staining procedures for Skp2, Cks1, and p27Kip1 was previously verified by comparing the protein levels as determined by immunohistochemistry with the protein levels determined by immunoblot analysis from the same tumor specimens [15,23].
Statistical analysis
Statistical data analyses were performed using SPSS 11.0 statistical software package (SPSS Inc, Chicago, IL, USA). First, the relations between Skp2, Cks1, and p27Kip1 and those between protein levels and various clinical and pathological features were studied using cross tabulation and Pearson's χ2. Survival curves were constructed using the Kaplan–Meier method and multivariate analysis by Cox regression; P values less than 0.05 were considered significant.
Results
Cks1 protein expression is directly related to Skp2 expression and inversely related to p27Kip1 expression in human breast cancer
We studied the expression of Cks1, Skp2, and p27Kip1 by immunohistochemistry in 50 tumor samples obtained from patients with breast cancer. The quality of immunostaining was good, with minimal background reactivity. Cks1 levels were very low or absent in all normal breast tissues adjacent to tumors. However, Cks1 levels were not uniformly high in all tumor specimens, being very high in some but virtually absent in others (Table 1). Given the important biochemical link between Cks1, Skp2, and p27Kip1, we examined the correlation between these proteins. Regression analysis of all 50 cases showed an inverse correlation between Skp2 and p27Kip1 levels (r = -0.395; P = 0.005). Cks1 expression was strongly related to Skp2 expression (r = 0.477; P = 0.001) and inversely related to p27Kip1 levels (r = -0.726; P < 0.001). Thus, strong relations were found between the expressions of all three proteins in 74% of the patients. High levels of Cks1 and Skp2 with low levels of p27Kip1 were observed in 15 patients, and low levels of Cks1 and Skp2 with high levels of p27Kip1 were observed in 22 patients. A typical representative immunohistochemical sample is shown in Fig. 1. In this tumor sample, cancer cells show strong nuclear staining for Cks1 and Skp2 and weak staining for p27Kip1, but high p27Kip1 staining and low Cks1 and Skp2 staining in the normal surrounding breast tissue. In 13 of the 50 cases, the above relation was not found. In eight cases an inverse relation was found between Cks1 and p27Kip1 expression but not between Skp2 and p27Kip1 expression; in three cases, levels of both p27Kip1 and Skp2 were high whereas Cks1 expression was low; and in another five cases, p27Kip1 levels were low despite high Cks1 levels and low Skp2 levels. In another five cases, the differences between the three proteins were not significant.
Cks1 overexpression is associated with poor tumor differentiation, young age, and negative ER/PR status
To examine the relation between Cks1 expression and common parameters associated with aggressive tumor behavior, we compared Cks1 levels with various clinicopathological features such as age, tumor size, pathological grade, lymph node status, and ER and PR expression (Table 1). A significant direct correlation was found between high Cks1 levels and loss of tumor differentiation (r = -0.421; P = 0.012), negative ER expression (r = -0.356; P = 0.013), and negative PR expression (r = -0.404; P = 0.006). We also found a strong relation between Cks1 overexpression and patient age less than than 50 years (r = 0.327; P = 0.049), but did not observe a significant correlation between Cks1 levels and lymph node status (r = 0.029; P = 0.508) or tumor size (r = 0.192; P = 0.333).
Examination of the association between Skp2 or p27Kip1 and patients' clinicopathological characteristics showed a strong association between Skp2 expression and loss of tumor differentiation (r = -0.400; P = 0.005), young age (r = -0.337; P = 0.022), and negative ER or negative PR status (r = -0.589; P = 0.001 and r = -0.674; P = 0.001), respectively. Loss of p27Kip1 was associated with loss of tumor differentiation (r = -0.323; P = 0.032) and, respectively, loss of ER and PR (r = 0.298, P = 0.048; r = -0.316, P = 0.030 (Table 1)).
Because of the strong association between Cks1, Skp2, and p27Kip1 levels and the ER status of the tumors, we next examined the expression of these proteins in estrogen-dependent (MCF-7 and T47D) and estrogen-independent (MDA-MB-231) breast carcinoma cell lines, and the possible effects of estrogen modulation on the regulation of Skp2 and Cks1 expression. Basal levels of Cks1 and Skp2 were higher in estrogen-independent cells than in estrogen-dependent cells, whereas p27Kip1 levels were lower (Fig. 2a). Interestingly, between the two estrogen-dependent cell lines, levels of Cks1 and Skp2 were higher in T47D cells, which phenotypicallly have poorer cellular differentiation (Fig. 2b). Treatment of T47D cells with estradiol increased Cks1 and Skp2 levels, whereas treatment with tamoxifen down regulated Cks1 and Skp2, resulting in an upregulation of p27Kip1 levels (Fig. 2b).
Cks1 overexpression is associated with poor disease-free and overall survival in human breast cancer
To determine the association between Cks1 expression and prognosis, we plotted Kaplan–Meier curves for disease-free survival and overall survival. Complete 80-month follow-up data were available for all the patients. Patients presenting with high tumor Cks1 levels had significantly shorter disease-free survival rates than patients with low Cks1 levels (54 months vs 76 months, respectively; P = 0.0007; Fig. 3a). There was only one case of disease recurrence in patients with low Cks1 levels, whereas in patients with high Cks1 levels, disease recurrence was observed in 11 (42% of patients). Similarly, Skp2 and p27Kip1 were also good predictors for disease-free survival (P = 0.0014 and P = 0.0191, respectively; data not shown).
We also examined the prognostic role of Cks1 in predicting local or distant relapse disease-free survival (Fig. 3b,c, respectively). Patients with high Cks1 levels had both higher local recurrence rates (P = 0.0352) and more distant relapse rates (P = 0.0006) than patients with low Cks1 levels. Skp2 was also found to be a good predictor for local (P = 0.0051) and distal (P = 0.003) relapse rates, while p27Kip1 expression was a predictor for distant relapse (P = 0.031) but not for local recurrence (P = 0.136).
There was also a strong association between Cks1 expression and overall survival (Fig. 4a). The mean survival rate for patients with low Cks1 expression was 79 months, significantly higher than that (72 months) of patients with high Cks1 levels (P = 0.0415). Similarly, Skp2 expression was also found to be a strong predictor of overall survival (69 months vs 79 months P = 0.0224) (Fig. 4b), but p27Kip1 expression was not (P = 0.0831; Fig. 4c).
Discussion
Recent studies have clearly shown that Cks1 plays an essential role in Skp2-mediated degradation of p27Kip1. The mechanistic model that emerges from these studies suggests that Cks1 is associated with Skp2 and confers an allosteric change in Skp2 that increases its affinity for phosphorylated p27Kip1 [18]. As a result, the Cks1-Skp2 interaction enables efficient transfer of ubiquitin to p27Kip1, resulting in rapid proteasome-mediated degradation. Whereas the role of Skp2 as an oncogene responsible for down regulation of p27Kip1 protein levels has been well established in a wide variety of cancers, the role of Cks1 in many of these cancers, and in particular breast cancer, is unknown. This study provides new insights into its role in human breast cancer. Using immunohistochemical studies obtained from tumor specimens of patients with nonmetastatic breast cancer, we found that Cks1 was overexpressed in subsets of patients with unfavorable histological features. Furthermore, its expression was strongly and independently associated with poor disease-free survival and poor overall survival. Similar findings were recently reported by our and other laboratories in aggressive gastric, lung, oral, and colorectal carcinomas [20-24]. Taken together, these results suggest that Cks1 has an important role in the decrease in p27Kip1 levels in aggressive cancers.
The interaction between the three proteins was evident in a significant proportion of the study patients, whereby the expression of Cks1 was strongly associated with Skp2 expression and inversely with p27Kip1protein levels. However, in five patients high Cks1 levels were associated with low p27Kip1 despite low Skp2 levels, and in three patients Cks1 levels were high but Skp2 and p27Kip1 levels were low. These findings, in which the inverse relation between Cks1 and p27Kip1 was stronger than that between Skp2 and p27Kip1 levels, are particularly interesting, considering that Skp2 is the prime mediator of p27Kip1 degradation. Although we do not yet know the reason for this observation, we postulate that in the presence of high Cks1 levels Skp2 degradation of p27Kip1 is more efficient and that lower levels of Skp2 are therefore required to down-regulate p27Kip1.
Cks1 overexpression was also strongly associated with various clinical and pathological features that are commonly used to determine aggressive tumor behavior, including poor tumor differentiation and lack of receptors to estrogen and progesterone. The association between Cks1, Skp2, and p27Kip1 and ER expression was also observed in breast carcinoma cell lines. Estrogen-dependent cell lines displayed lower levels of Cks1 and Skp2 than estrogen-independent cell lines. Nevertheless, estrogen modulation affected the levels of these proteins, suggesting that tamoxifen may have an important therapeutic role in ER-positive cancers expressing high levels of Cks1 or Skp2.
We found a strong correlation between young age of the patient and overexpression of either Cks1 or Skp2. Thus, these results support the notion that breast cancer has more aggressive tumor biology in young patients than in older ones, and that this difference may account, at least in part, for the poorer survival rates of younger patients than of older patients with similar clinicopathological features. Overall, Cks1 or Skp2 were each found to be good prognostic markers for disease-free and overall survival. Interestingly, there was no significant association between p27Kip1 expression and overall survival (P = 0.083). A possible reason for this finding is the relatively small number of patients in this study. Larger studies may be required to resolve this issue.
Conclusion
The results of the present study provide additional insights into the importance of alterations in cell-cycle regulatory proteins and tumor progression in breast cancer. Our findings suggest that Cks1 has an important role in the deregulation of the cell-cycle protein p27Kip1 in breast cancer. Overexpression of Cks1 was strongly associated with increased Skp2 expression and down regulation of p27Kip1 levels, resulting in aggressive tumor behavior and poor prognosis. Thus, both Cks1 and Skp2 may be considered as potential novel prognostic markers and targets for the future development of specific therapeutic interventions in breast cancer.
Abbreviations
Cks = cyclin kinase subunit; ER = estrogen receptor; PR = progesterone receptor; Skp = S-phase kinase protein; TBST = Tris-buffered saline and 0.1% Tween 20.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MS (Merav Slotky) and MS (Ma'anit Shapira) contributed equally in the preparation of the study. Both carried out the immunohistochemical studies and the immunobot analysis, participated in the sequence alignment, and drafted the manuscript. OBI carried out the pathological and immunohistochemical analysis and interpretation. SL and BF participated in the design of the study and performed the statistical analysis. MT collected and analyzed the clinical data and helped draft the manuscript. DDH conceived of the study, participated in its design and coordination, and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We thank Dr Avram Hershko for critical review and Dr Micha Bar-Hanna and Mrs Miriam Alon from the Israel Cancer Registry for their assistance. This study was supported by the Israel Cancer Research Fund and the Israel ministry of Health.
Figures and Tables
Figure 1 Representative immunohistochemistry slides for Cks1, Skp2, and p27Kip1 staining in breast cancer. Tumor cells of a grade II invasive duct cancer exhibiting strong and diffuse nuclear staining for Skp2 (S-phase kinase protein 2) and Cks1 (cyclin kinase subunit 1) (× 300) and low p27Kip1 staining. Note the inversely low Skp2 and Cks1 staining with high p27Kip1 staining of normal adjacent breast tissue (× 400).
Figure 2 Immunoblots showing Cks1, Skp2, and p27Kip1 proteins in breast cancer cell lines. (a) Levels of the three proteins Cks1 (cyclin kinase subunit 1), Skp2 (S-phase kinase protein 2), and p27Kip1 were measured in MDA-MB-231 (an estrogen-independent cell line) and MCF-7 (an estrogen-dependent cell line). Skp1 levels were determined to confirm equal protein loading. (b)Effect of estrogen modulation on Cks1, Skp2, and p27Kip1 expression in T47D cells (an estrogen-dependent breast cancer cell line) treated with estradiol (0.5 μM) or tamoxifen (10 μM). P27, p27Kip1.
Figure 3 The association between Cks1 expression and disease-free survival. The expression of Cks1 (cyclin kinase subunit 1) was plotted in relation to overall (a), local (b), or distant (c) recurrence rates. Analyses were constructed using the Kaplan–Meier method. Protein expression was considered low if fewer than 10% of cell nuclei were stained and high if more than 10% of nuclei were stained.
Figure 4 The association between expression of cell cycle proteins and survival of patients with breast cancer. Relation between Cks1 (cyclin kinase subunit 1)(a), Skp2 (S-phase kinase protein 2) (b), and p27Kip1 (c) expression and overall survival (death). Analyses were constructed using the Kaplan–Meier method. Protein expression was considered low if fewer than 10% of cell nuclei were stained and high if more than 10% of nuclei were stained.
Table 1 Relations between proteina expression and clinicopathological characteristics in patients with breast cancer
Characteristic Cks1 Skp2 p27Kip1
High Low
P
b
High Low
P
b
High Low
P
b
Grade
1 2 8 0.012 2 8 0.005 2 8 0.032
2 14 14 0.012 7 21 0.005 14 14 0.032
3 10 2 0.012 9 3 0.005 8 4 0.032
Tumor sizec
T1 15 18 0.333 10 23 0.265 19 14 0.485
T2 10 6 0.333 7 9 0.265 8 8 0.485
T3 1 0 0.333 1 0 0.265 0 1 0.485
Lymph node involvementc
N0 16 15 0.508 11 20 0.803 18 13 0.387
N1 6 3 0.508 4 5 0.803 3 6 0.387
N2 4 6 0.508 3 7 0.803 6 4 0.387
Estrogen receptor expression
Negative 17 6 0.013 16 7 0.001 9 14 0.048
Positive 10 17 0.013 3 24 0.001 17 10 0.048
Progesterone receptor expression
Negative 15 3 0.006 14 4 0.001 5 13 0.032
Positive 12 20 0.006 4 27 0.001 20 12 0.032
Age (years)
>50 14 5 0.017 12 7 0.002 7 12 0.051
< 50 12 19 0.017 25 6 0.002 20 11 0.051
aCks1, cyclin kinase subunit 1; p27Kip1, a cell-cycle inhibitory protein; Skp2, S-phase kinase protein 2. bRefers to the differences between protein expression and the variables. cAccording to the TNM (tumor, node, metastases) staging system of the American Joint Committee on Cancer. P ≤ 0.05 (values shown in bold type) was considered statistically significant.
==== Refs
Morabito A Magnani E Gion M Sarmiento R Capaccetti B Longo R Gattuso D Gasparini G Prognostic and predictive indicators in operable breast cancer Clin Breast Cancer 2003 3 381 390 12636883
Tsihlias J Kapusta L Slingerland J The prognostic significance of altered cyclin-dependent kinase inhibitors in human cancer Annu Rev Med 1999 50 401 423 10073286 10.1146/annurev.med.50.1.401
Alkarain A Jordan R Slingerland J p27 deregulation in breast cancer: prognostic significance and implications for therapy J Mammary Gland Biol Neoplasia 2004 9 67 80 15082919 10.1023/B:JOMG.0000023589.00994.5e
Chiarle R Pagano M Inghirami G The cyclin dependent kinase inhibitor p27 and its prognostic role in breast cancer Breast Cancer Res 2001 3 91 94 11250752 10.1186/bcr277
Catzavelos C Bhattacharya N Ung YC Wilson JA Roncari L Sandhu C Shaw P Yeger H Morava-Protzner I Kapusta L Franssen E Decreased levels of the cell-cycle inhibitor p27Kip1 protein: prognostic implications in primary breast cancer Nat Med 1997 3 227 230 9018244 10.1038/nm0297-227
Loda M Cukor B Tam SW Lavin P Fiorentino M Draetta GF Jessup JM Pagano M Increased proteasome-dependent degradation of the cyclin-dependent kinase inhibitor p27 in aggressive colorectal carcinomas Nat Med 1997 3 231 234 9018245 10.1038/nm0297-231
Tsihlias J Kapusta LR DeBoer G Morava-Protzner I Zbieranowski I Bhattacharya N Catzavelos GC Klotz LH Slingerland JM Loss of cyclin-dependent kinase inhibitor p27Kip1 is a novel prognostic factor in localized human prostate adenocarcinoma Cancer Res 1998 58 542 548 9458103
Hommura F Dosaka-Akita H Mishina T Nishi M Kojima T Hiroumi H Ogura S Shimizu M Katoh H Kawakami Y Prognostic significance of p27KIP1 protein and ki-67 growth fraction in non-small cell lung cancers Clin Cancer Res 2000 6 4073 4081 11051259
Pagano M Tam SW Theodoras AM Beer-Romero P Del Sal G Chau V Yew PR Draetta GF Rolfe M Role of the ubiquitin-proteasome pathway in regulating abundance of the cyclin-dependent kinase inhibitor p27 Science 1995 269 682 685 7624798
Hershko A Ciechanover A The ubiquitin system Annu Rev Biochem 1998 67 425 479 9759494 10.1146/annurev.biochem.67.1.425
Carrano AC Eytan E Hershko A Pagano M SKP2 is required for ubiquitin-mediated degradation of the CDK inhibitor p27 Nat Cell Biol 1999 1 193 199 10559916 10.1038/12013
Sutterluty H Chatelain E Marti A Wirbelauer C Senften M Muller U Krek W p45SKP2 promotes p27Kip1 degradation and induces S phase in quiescent cells Nat Cell Biol 1999 1 207 214 10559918 10.1038/12027
Tsvetkov LM Yeh KH Lee SJ Sun H Zhang H p27(Kip1) ubiquitination and degradation is regulated by the SCF(Skp2) complex through phosphorylated Thr187 in p27 Curr Biol 1999 9 661 664 10375532 10.1016/S0960-9822(99)80290-5
Signoretti S Di Marcotullio L Richardson A Ramaswamy S Isaac B Rue M Monti F Loda M Pagano M Oncogenic role of the ubiquitin ligase subunit Skp2 in human breast cancer J Clin Invest 2002 110 633 641 12208864 10.1172/JCI200215795
Hershko D Bornstein G Ben-Izhak O Carrano A Pagano M Krausz MM Hershko A Inverse relation between levels of p27(Kip1) and of its ubiquitin ligase subunit Skp2 in colorectal carcinomas Cancer 2001 91 1745 1751 11335900 10.1002/1097-0142(20010501)91:9<1745::AID-CNCR1193>3.0.CO;2-H
Gstaiger M Jordan R Lim M Catzavelos C Mestan J Slingerland J Krek W Skp2 is oncogenic and overexpressed in human cancers Proc Natl Acad Sci USA 2001 98 5043 5048 11309491 10.1073/pnas.081474898
Yang G Ayala G De Marzo A Tian W Frolov A Wheeler TM Thompson TC Harper JW Elevated Skp2 protein expression in human prostate cancer: association with loss of the cyclin-dependent kinase inhibitor p27 and PTEN and with reduced recurrence-free survival Clin Cancer Res 2002 8 3419 3426 12429629
Ganoth D Bornstein G Ko TK Larsen B Tyers M Pagano M Hershko A The cell-cycle regulatory protein Cks1 is required for SCF(Skp2)-mediated ubiquitinylation of p27 Nat Cell Biol 2001 3 321 324 11231585 10.1038/35060126
Spruck C Strohmaier H Watson M Smith AP Ryan A Krek TW Reed SI A CDK-independent function of mammalian Cks1: targeting of SCF(Skp2) to the CDK inhibitor p27Kip1 Mol Cell 2001 7 639 650 11463388 10.1016/S1097-2765(01)00210-6
Masuda TA Inoue H Nishida K Sonoda H Yoshikawa Y Kakeji Y Utsunomiya T Mori M Cyclin-dependent kinase 1 gene expression is associated with poor prognosis in gastric carcinoma Clin Cancer Res 2003 9 5693 5698 14654553
Inui N Kitagawa K Miwa S Hattori T Chida K Nakamura H Kitagawa M High expression of Cks1 in human non-small cell lung carcinomas Biochem Biophys Res Commun 2003 303 978 984 12670508 10.1016/S0006-291X(03)00469-8
Kitajima S Kudo Y Ogawa I Bashir T Kitagawa M Miyauchi M Pagano M Takata T Role of Cks1 overexpression in oral squamous cell carcinomas: cooperation with Skp2 in promoting p27 degradation Am J Pathol 2004 165 2147 2155 15579456
Shapira M Ben-Izhak O Bishara B Futerman B Minkov I Krausz MM Pagano M Hershko DD Alterations in the expression of the cell cycle regulatory protein cyclin kinase subunit 1 in colorectal carcinoma Cancer 2004 100 1615 1621 15073847 10.1002/cncr.20172
Shapira M Ben-Izhak O Linn S Futerman B Minkov I Hershko DD The prognostic impact of the ubiquitin ligase subunits Skp2 and Cks1 in colorectal carcinoma Cancer 2005 103 1336 1346 15717322 10.1002/cncr.20917
|
16168119
|
PMC1242136
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jul 19; 7(5):R737-R744
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1278
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12791616811710.1186/bcr1279Research ArticleIncreased copy number at 3p14 in breast cancer Ljuslinder Ingrid [email protected] Beatrice [email protected] Irina 2Thomasson Marcus 1Grankvist Kjell 3Höckenström Thomas 4Emdin Stefan 5Jonsson Yvonne 1Hedman Håkan 1Henriksson Roger 11 Department of Radiation Sciences, Oncology, Umeå University, Umeå, Sweden2 Department of Medical Biosciences, Medical and Clinical Genetics, Umeå, Sweden3 Department of Medical Biosciences, Clinical Chemistry, Umeå University, Umeå, Sweden4 Department of Pathology, Umeå University, Umeå, Sweden5 Department of Surgical and Perioperative Sciences, Division of Surgery, Umeå University, Sweden2005 6 7 2005 7 5 R719 R727 12 11 2004 4 4 2005 2 5 2005 8 6 2005 Copyright © 2005 Ljuslinder et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
The present study was conducted to investigate if chromosome band 3p14 is of any pathogenic significance in the malignant process of breast cancer. Genetic studies have implicated a tumour suppressor gene on chromosome arm 3p and we have proposed LRIG1 at 3p14 as a candidate tumour suppressor. The LRIG1 gene encodes an integral membrane protein that counteracts signalling by receptor tyrosine kinases belonging to the ERBB family. LRIG1 mRNA and protein are expressed in many tissues, including breast tissue.
Methods
In the present report we analysed the LRIG1 gene by fluorescence in situ hybridisation (FISH), LRIG1 mRNA by quantitative RT-PCR, and LRIG1 protein by western blot analysis. Two tumour series were analysed; one series consisted of 19 tumour samples collected between 1987 and 1995 and the other series consisted of 9 tumour samples with corresponding non-neoplastic breast tissues collected consecutively.
Results
The LRIG1 gene showed increased copy number in 11 out of 28 tumours (39%) and only one tumour showed a deletion at this locus. Increased LRIG1 copy number was associated with increased levels of LRIG1 mRNA (two of three tumours) and protein (four of four tumours) in the tumours compared to matched non-neoplastic breast tissue, as assessed by RT-PCR and western blot analysis.
Conclusion
The molecular function of LRIG1 as a negative regulator of ERBB receptors questions the biological significance of increased LRIG1 copy number in breast cancer. We propose that a common, but hitherto unrecognised, breast cancer linked gene is located within an amplicon containing the LRIG1 locus at 3p14.3.
==== Body
Introduction
Breast cancer is a major cause of death among women. In order to provide optimal treatment, prognostic factors, such as lymph node status and steroid receptor expression, are widely used. In recent years, genetic approaches studying chromosomal aberrations have been suggested as a tool in the process to individualise the adjuvant treatment given to patients. Several studies have been published during the past years with a focus on identifying the genes that contribute to initiation and clinical progression of breast cancer [1,2].
Identification of a germline mutation of BRCA1 at 17q21 [3] and BRCA2 at 13q12-13 [4] has been an important finding in studies of hereditary breast cancer. Epidermal growth factor receptor (EGFR/ERBB1) and ERBB2 (also known as HER2) overexpression [5-7], p53 inactivation [8,9] and nm23 overexpression [10] also seem to be of clinical prognostic importance. Chromosomal amplifications have been described in breast cancer for several genes, including MYC at 8q24 and ERBB2 at 17q11.2 [11,12]. Other amplified chromosomal areas detected in breast cancer are 13q31, 17q22-24, 1q41-44 and 20q13. In general, gene amplifications are considered late events in cancerogenesis, even though much is still unknown about the importance of amplifications of specific genes. In breast cancer, amplification of ERBB2 correlates with a worse prognosis [13] and amplification of C-MYC is associated with progression from carcinoma in situ to invasive breast cancer [14]. Cytogenetic analyses of tumours have shown that chromosome 1 is the most frequently altered chromosome in breast cancer [15]. In other breast cancer studies, loss of heterozygosity (LOH) at 3p was the most common chromosomal aberration [16-18]. In a study by Maitra et al. [19], LOH in the 3p area was apparent in 87% of breast tumours, and LOH at 3p14.3 in 41% of the tumours. The short arm of chromosome 3 thus likely harbours at least one tumour suppressor gene [20]. The FHIT gene localized to 3p14.2, which frequently shows LOH, is also suggested to be a prognostic factor in breast cancer [21,22].
Recently, the human gene LRIG1 (leucine-rich and immunoglobulin-like domains 1) was described and localised to chromosome 3p14.3 [23,24]. The LRIG1 gene encodes a protein with extracellular leucine-rich repeats and immunoglobulin-like domains, a transmembrane part, and a cytoplasmic tail. LRIG1 acts as a negative regulator of ERBB1-4 by enhancing receptor ubiquitylation and degradation [25,26]. The mechanism involves the recruitment of c-Cbl, an E3 ubiquitin ligase that simultaneously ubiquitylates EGFR and LRIG1 and sorts them for degradation [25].
The role of LRIG1 as a part of a group of proteins that help desensitize receptor tyrosine kinase (RTK) signalling makes it important to study the expression and role of LRIG1 in tumours in which the ERBB receptors have clinical relevance.
The present study was conducted to investigate if the LRIG1 gene, mRNA, or protein was deleted or dysregulated in human breast cancer. The LRIG1 locus was analysed by fluorescence in situ hybridisation (FISH), mRNA was quantified by real-time RT-PCR and protein was analysed by western blot analysis. To further explore how LRIG1 expression was related to growth factor receptor expressions, quantitative RT-PCR of EGFR and ERBB2 was performed. We report an unexpected increase in copy number of the LRIG1 locus in 39% of the breast tumours, implicating a breast cancer gene at, or close to, 3p14.3.
Materials and methods
Patients and sample preparation
Previously collected (1986 to 1995) samples from 19 patients were included in a first examination (group A). Tumour samples and non-neoplastic breast tissue were then collected from nine patients with breast carcinoma (group B). Clinical characteristics of the patients are presented in Table 1. The study was approved by the local ethics committee. None of the patients had received any treatment prior to specimen collection. In group B, samples of the tumour and a piece of the non-neoplastic breast tissue were collected immediately after excision, one part of each frozen in liquid nitrogen and stored at -80°C, and another part stored in RNAlater (Ambion inc, Austin, Texas, USA). The other adjacent parts of the tissue samples were fixed in formalin, paraffin embedded and used for routine morphological examination and tumour grading (according to Page et al. [27]), immunohistochemical staining and tumour tissue array construction. The preparation of RNA was performed as previously described [23].
FISH
Freshly frozen breast cancer tissues were disintegrated in methanol:acetic acid solution (3:1; Carnoy's solution) on ice. The nuclei were collected by passing the disintegrated tissues through a nylon mesh (pore size 70 μm) and then centrifuged. Cells were washed in methanol:acetic acid solution (3:1) two to three times at room temperature. FISH slides were prepared by dropping the cell suspension onto glass slides. After air-drying, FISH-slides were immediately used or stored at -20°C. Before hybridisation, FISH-slides were incubated in 75 mM KCl for 20 minutes at 37°C and fixed in Carnoy's solution for 5 minutes at room temperature. After fixation, FISH slides were treated with RNAase (100 μg/ml) for 1 h, followed by washing in 2 × SSC (saline sodium citrate) three times for 2 minutes each time. Finally, the slides were incubated in solution containing 100 μg/ml pepsin in 10 mM HCl for 10 minutes, followed by incubation in PBS for 5 minutes at room temperature and stepwise dehydration in alcohol (70%, 80%, 95%). The BAC clone 751k5 (Invitrogene, Carlsbad, USA), containing the LRIG1, was used as the FISH probe. DNA was labelled by nick translation using Spectrum Orange according to the manufacturer's protocol (Abbot Diagnostics, Wiesbaden-Delkenheim, Germany). Probe (10 μl) containing 100 ng DNA, 5 μl Cot-1 DNA in 60% formamide was pre-incubated for 1 h at 37°C and then applied to each slide. Probe and target DNA were denatured simultaneously for 3 minutes at 72°C. Slides were hybridised overnight at 37°C in a humidity chamber. Post-hybridisation washing was performed in 2 × SSC containing 0.3% NP-40. Nuclear counterstaining was done with DAPI solution for 2 minutes. As control, a centromere probe for chromosome 3 was included in the hybridisation solution.
In each case, LRIG1 and CEP3 signals were counted in 100 to 200 nuclei by two independent investigators. The presence of at least three signals in more than 20% of the nuclei was the criteria for scoring an increased copy number of LRIG1. Analysis was performed using an Axioplan 2 microscope (Carl Zeiss Vision, Hallbergmoos, Germany.) Digital images were captured and stored using Cytovision software (Applied Imaging Corporation, San Jose, USA).
Cell lines
The breast cancer cell lines MDA-MB-231, MDA-MB-415 and HS 578T were obtained from American type culture collection (Manassas, VA, USA) and ZR-75-1 was kindly provided by Dr J Bergh (Uppsala University, Sweden). The breast cancer cell lines were cultivated in Dulbecco's modified Eagles medium, supplemented with 10% w/v fetal bovine serum and 50 μg/ml gentamicin from Invitrogen AB (Täby, Sweden). The immortalised mammary epithelial cell line hTERT-HME1 was obtained from BD Bioscencies Clontech (Stockholm, Sweden) and cultivated according to the manufacurer's instructions by using media and supplements from Clonetics, Bio Whittaker (Walkersville, MD, USA).
Quantitative RNA analysis
RNA was prepared from tissue samples by using RNAqueous kit (Ambion inc, Austin, Texas, USA), according to the manufacturer's instructions. Real-time quantitative RT-PCR was performed as previously described [28].
Western blot analysis
Cell lysates, protein concentrates and immunoprecipitated material were incubated in LDS (lithium dodecyl sulfate) sample buffer for 10 minutes at 70°C followed by electrophoresis on 3% to 8% TRIS-acetate NuPAGE gradient gel. The proteins were thereafter transferred to polyvinylidene difluoride membranes by using an Xcell II Mini-Gel blot module. Gel apparatus, gels, buffers, blotting module, and membranes were from Invitrogen. Non-specific binding was blocked by using incubation of the membranes with 5% w/v non-fat milk powder in TBS containing 0.1% w/v Tween-20. The membranes were thereafter incubated with the primary antibodies at 1 μg/ml followed by peroxidase-conjugated secondary antibodies (Amersham Pharmacia Biotech, Amersham Biosciences, New Jersey, USA). The primary antibodies used were LRIG1-151 [24] and rabbit anti-actin (Sigma-Aldrich St. Louis, Missouri, USA). Visualization was performed by using the enhanced chemiluminescense system ECL-plus, (ECL-advanced and hyperfilm ECL Blotting Detection system kit, Amersham Biosciences, New Jersey, USA). The samples were diluted stepwise by approximately 50% in 3 to 4 steps. The results were analysed visually by three separate investigators and an apparent change between tumour and non-neoplastic tissue of at least 50% was considered convincing.
Results
FISH analysis of archived breast cancer samples
To evaluate the number of LRIG1 gene copies, FISH was performed on cell nuclei from the archived breast cancer samples (group A). An increased copy number of LRIG1 was seen in more than 20% of the nuclei in 7 of the 19 tumours (in most cells three to five signals). The fraction of tumour cells with increased copy number varied between 23% and 79%. Normal signal pattern corresponding to two copies per nucleus was detected in 11 of the 19 tumours, and 1 tumour demonstrated decreased copy number of LRIG1 (Table 2).
FISH Analysis of Fresh Tumour Samples and Breast Cell lines
FISH analysis revealed increased copy numbers of LRIG1 in four of the nine tumours from group B (example shown in Fig. 1a). The fraction of tumour cells with increased copy number varied between 21% and 49%. Normal signal pattern corresponding to two LRIG1 copies per nucleus was detected in the remaining five tumours (Table 2). A parallel FISH analysis including 10 tumours of a different tissue origin showed no aberrations of LRIG1 gene copy numbers in these tumours (ongoing study, data not shown). In one of the breast tumours with increased copy number of LRIG1 (patient B8), a more detailed FISH analysis was performed to assess the chromosome 3 status and the ploidity of the tumour cells. This showed that the LRIG1 copy number was increased (Fig. 1a) but the chromosome 3 centromere was not (Fig. 1b). Furthermore, by using a mixture of LRIG1 probe and a specific 3p subtelomere probe (probe position 30 tel (D3S4559); Abbot Vysis), no increased copy number was found for the 3p subtelomeric region either (Fig. 1c). No evidence of aneuploidy was found, as analysed by using centromere probes for chromosomes 3, 18 and X (Fig. 1d).
FISH analysis was also performed on the breast cancer cell lines MDA-MB-231, MDA-MB-415, HS 578T and ZR-75-1, and the immortalised mammary epithelial cell line hTERT-HME1. Increased copy number of LRIG1 was found in three of the five cell lines (MDA-MB-231, HS 578T and hTERT-HME1), whereas decreased copy number was found in the MDA-MB-415 cell line. A normal FISH signal pattern for LRIG1 was present in ZR-75-1.
Quantitative RT-PCR
Quantitative RT-PCR was performed on RNA extracted from tumour tissue and non-neoplastic tissue from seven of the nine patients in group B (the quality of the samples from two of the patients was not adequate for quantitative RT-PCR analysis). A fibroadenosis was also examined, with collected pieces both from the fibroadenosis and the surrounding tissue. The expression levels in different parts of the healthy breast tissue from the same individual did not differ by more than 20% (data not shown) and, therefore, a 20% cut-off level was used for both overexpression and underexpression. The ratio between the expression in the tumour and non-neoplastic tissue was calculated and ratios >1.2 were regarded as significant tumour overexpression and ratios <0.8 were regarded as significant tumour underexpression. LRIG1 mRNA was significantly overexpressed in two of the seven tumours and significantly underexpressed in two of the seven tumours (Table 3). The three tumours with increased LRIG1 copy number (FISH analysis) that were able to be analysed by RT-PCR for LRIG1 showed significant overexpression of LRIG1 mRNA in two cases. Four tumours with increased LRIG1 copy number were analysed by RT-PCR for EGFR/ERBB1, and all four showed significantly lower expression of EGFR/ERBB1 and three showed significantly higher expression of ERBB2 than their matched normal controls (Table 4). Two of the tumours without increased LRIG1 copy number also had lowered EGFR expression.
Western blot analysis of fresh tumour samples and their matched non-neoplastic breast tissue
In group B, five of the nine tumours overexpressed the LRIG1 protein compared to their matched non-neoplastic tissues as analysed by western blotting. Four of these five tumours also displayed increased LRIG1 copy number (Fig. 2, Table 3). Thus, all of the tumours with increased LRIG1 copy number overexpressed LRIG1 as determined by western blot analysis, but also one tumour with normal LRIG1 copy number showed high levels of the protein by western blotting.
Combined analysis of group A and B
In total, 28 breast tumours were analysed by FISH with LRIG1 specific probe. A normal signal pattern, corresponding to two LRIG1 copies per nucleus, was detected in 16 cases. In 11 out of 28 tumours (39%), an increased number of LRIG1 signals were found (Table 2, Fig. 1a). The fraction of tumour cells with increased copy number varied between 21% and 79%. Complementary FISH analyses showed that there was no increase in the copy number of the entire 3p arm (Fig. 1b–d). As seen by quantitative RT-PCR analysis, two out the three analysed tumours with increased LRIG1 copy number (in group B) showed higher expression of LRIG1 mRNA than the matched non-neoplastic breast tissues. In all four tumours with increased LRIG1 copy number, expression of LRIG1 protein was higher than in the matched non-neoplastic breast tissue, as assessed by western blot analysis.
Complementary analysis of five transformed breast cancer cell lines showed similar results, with three of them showing an increased copy number of the 3p14 locus.
Discussion
This novel investigation of 3p14 demonstrated unexpectedly an increased copy number of the proposed tumour suppressor gene LRIG1 in 39% (11/28) of the breast tumours and in 60% (3/5) of the breast cancer cell lines. The malignant process is believed to be driven by genetic diversification through mutations, deletions and amplifications followed by natural selection of surviving and proliferating cancer cells. One result of this process is the enrichment of amplicons harbouring tumour promoting genetic elements, that is, cancer associated genes. Accordingly, the presented results imply that a common, but hitherto unrecognised, breast cancer related gene was located within an amplicon that included the LRIG1 locus at 3p14.
Despite numerous genetic studies, increased copy number at the 3p14 locus has, to our knowledge, never previously been reported in primary human breast tumours. Interestingly, however, amplifications at 3p14 have recently been reported in breast cancer-derived cell lines [29,30]. Previous comparative genomic hybridisation (CGH) studies of 3p have generally shown losses and only rarely gains [31]. There are at least four possible explanations why the herein demonstrated increased copy number at 3p14 has not previously been described. First, the area of increased copy number could be relatively small, and so escaped detection by conventional analyses. Second, most studies of this chromosomal area have analysed LOH, and thus have not addressed possible gene amplifications. Third, results obtained by conventional CGH, a method frequently used to detect both gains and losses, are usually difficult to interpret in chromosomal regions close to the centromere, such as the LRIG1 locus at 3p14. By employing an alternative CGH methodology using cDNA arrays, an amplicon at 3p14 was described in breast cancer-derived cell lines [29]. Whether this region is co-duplicated with LRIG1 at 3p14 has not been addressed but will be the subject of future studies. Finally, a further limitation of conventional and array based CGH methodologies is that they evaluate the mean gene copy number in the analysed sample. FISH, in contrast, has single cell resolution and, thus, is more sensitive and able to detect modest gene copy number changes that could involve only a minority of the tumour cells.
An important question regarding the herein discovered amplicon is its size and the identity of the underlying possible breast cancer gene(s). We have shown by FISH that the area of increased copy number contained LRIG1 at 3p14 but was lacking the centromere and the subtelomeric region of chromosome 3. In addition, as discussed above, chromosome 3 has previously been extensively studied by CGH, which rules out the possibility of a common amplicon spanning centromere-distal regions of 3p. From this, we estimate that the putative breast cancer gene is located on chromosome 3, somewhere between the centromere and 3p21. Obviously, the only gene directly demonstrated so far to be duplicated in the analysed breast tumours was LRIG1. This raises the question of whether LRIG1 itself is a breast cancer gene. LRIG1 has been proposed to interact with and counteract the effects of growth factor receptors such as EGFR/ERBB1 [23,32], thereby functioning as a tumour suppressor. This hypothesis was recently confirmed by molecular studies showing that LRIG1 downregulates ERBB1-4 by enhancing receptor degradation [25,26]. Because EGFR/ERBB1 and ERBB2 are important and frequently overexpressed breast cancer genes, it is unlikely that LRIG1, as an ERBB antagonist, is a tumour promoter. Of course, we cannot exclude that LRIG1 might have other functions, which for tumour promotion could dominate over its ERBB-antagonising effects. According to a recent estimate [33], however, there are 80 genes in addition to LRIG1 in the region between the centromere and 3p21 (coordinates 64M-92M), the gene copy numbers of which could potentially have been increased in conjuction with LRIG1. These genes encode a variety of different kinds of proteins, of known and unknown functions, including a tyrosine kinase receptor (EPHA3), a protein phosphatase regulatory subunit (PPP4R2), an ubiquitin-conjugating enzyme (UBE1C), and different transcription factors (e.g. TMF1, FOXP1 and POU1F1). The amplicon described by Hyman et al. [29] is confined to a 2.7 mb region (coordinates 72M-75M), which include 13 genes but not LRIG1 at 66M. Moreover, an amplicon close to the LRIG1 locus with the coordinates 60M-64M was recently described in breast cancer cell line MCF-7 [30]. This amplicon contains about 17 genes. Whether the regions at 72M-75M and 60M-64M are increased in copy number in conjunction with the LRIG1 locus at 66M was not addressed in the present study but will be the subject of future studies. Clearly, a more refined mapping of the area of increased copy number and functional studies of candidate genes are needed for defining the hypothesised breast cancer gene(s). We conclude, nevertheless, that a common but hitherto unrecognised breast cancer gene is located at or near the LRIG1 locus.
Increased LRIG1 copy number as detected by FISH was associated with increased mRNA and protein levels in tumours compared to non-neoplastic breast tissue, as determined by quantitative RT-PCR and western blot analysis (Table 3). A concordance between gene overexpression and enhanced mRNA levels is often, but not always, observed [29].
Because ERBB family members are strongly implicated in the aetiology of breast cancer, and because ERBB proteins and LRIG1 interact at the molecular level, we analysed the expression of EGFR/ERBB1 and ERBB2 mRNA. Intriguingly, EGFR/ERBB1 mRNA was significantly underexpressed in all of the four tumours analysed with increased LRIG1 copy number. Of these four tumours, three showed significant overexpression of ERBB2 mRNA. EGFR/ERBB1 is overexpressed in 35% to 60% of breast cancers, which correlates with a negative steroid receptor status, increased ERBB2 and VEGF (Vascular endothelial growth factor) expression [7]. The impact of EGFR/ERBB1 overexpression on clinical outcome has not been completely clarified, but in most studies it is considered to be a negative prognostic factor [34]. The combination of increased LRIG1 copy number and protein expression and low EGFR/ERBB1 expression might represent a subtype of breast cancer with its own clinical features. To reveal such a biological subtype, a larger number of tumours must be examined.
Conclusion
We have described, for the first time, frequent increased copy number at 3p14.3 in breast cancer. This attributes a breast cancer associated gene to 3p14 or surrounding plausibly co-amplified regions. In future studies, it will be important to define the genetic element, that is, the breast cancer gene(s) underlying the observed copy number increase, and to examine a greater number of tumours in order to evaluate its clinical significance.
Abbreviations
CGH = comparative genomic hybridisation; EGFR = epidermal growth factor receptor; FISH = fluorescence in situ hybridisation; LOH = loss of heterozygosity; LRIG1 = leucine-rich and immunoglobulin-like domains 1; PBS = phosphate buffered saline; RT-PCR = reverse transcriptase polymerise chain reaction; SSC = saline sodium citrate.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
IL is a PhD student in the project and participated in the FISH analysis and the overall analysis of the results. IL was also responsible for writing the manuscript. TH participated in the collection of new breast cancer samples and investigating whether the samples were representative. MT carried out the quantitative RT-PCR analysis, YJ carried out the western blot analysis, IG was responsible for the FISH analysis, SE participated in contact with patients for the collection of breast cancer tumours, and KG was responsible for the collection of the breast cancer tumours from 1987 to 1995. BM, HH and RH were responsible for the overall design and implementation of the study and also for the overall analysis of the material and they helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We would like to thank Kerstin Bergh for help with the immunohistochemistry and Charlotte Andersson for help with the FISH analysis. This study was supported by grants from The Swedish Cancer Society and the Cancer Research Foundation, Northern Sweden.
Figures and Tables
Figure 1 FISH analysis of breast cancer cells from patient B8. (a) Analysis with a specific LRIG1 (red) probe showing increased gene copy number (more than two copies) of the LRIG1 gene at 3p14. (b) Analysis with a specific CEP3 (centromeric chromosome 3; red) probe, showing no additional chromosome 3. (c) Analysis with a specific 3p subtelomeric probe (green) and LRIG1 (red) mixture showing increased gene copy number of the LRIG1 gene but only two copies of the 3p arm. (d) Analysis with a mixture of probes for CEP3 (red), X chromosome (green) and chromosome 18 (blue), showing no aneuploidy for these chromosomes.
Figure 2 Western blot analyses of LRIG1 in the nine breast cancer patients in group B. Tumours (T) versus non-neoplastic breast tissue (NN). Western blot analysis was performed on samples with primary antibodies LRIG1-151 and anti-rabbit anti-actin. A visual change of at least 50% was considered convincing, as determined by three different investigators. Pat, patient number.
Table 1 Clinical characteristics of the 28 breast tumours included in the study
n % of total Tumours with increased LRIG1 copy number (n = 11) Tumours without increased LRIG1 copy number (n = 17)
Tumour type
Ductal 21 75 10 (82 %) 11(65%)
Lobular 2 7 1 1
Mixed group 4 14 0 4
Medullar 1 4 0 1
Axillar lymph node
Negative 17 61 6 (54%) 11 (65%)
Positive 10 36 5 (45%) 5 (29%)
Missing value 1 4 0 1
Estrogen receptor
Positive 21 75 8 (73%) 13 (76%)
Negative 4 14 2 (9%) 2 (12%)
Missing value 3 11 1 2 (12%)
Progesterone receptor
Positive 18 64 6 (54%) 12 (70%)
Negative 5 18 3 (27%) 2 (28%)
Missing value 5 18 2 (18%) 3 (18%)
Outcome=deceased 10 36 3 (27%) 7 (41%)
Table 2 FISH analysis of the LRIG1 locus in 28 breast cancer tumours
FISH 3p14.3 Group A (n = 19) Group B (n = 9)
Increased copy number 7 (36 %) 4 (44%)
Normal copy number 11 (57%) 5 (55%)
Decreased copy number 1 (5%) 0
Group A: tumours collected 1986 to 1996 retrospectively. Group B: tumours collected 2002 to 2003 prospectively.
Table 3 Summary of FISH, western blot and RT-PCR analysis of LRIG1 in nine breast cancer tumours
Patient no. FISHa Western blotb LRIG1 RT-PCR (T/NN)c
B1 I In 1.45d
B2e I In -
B3e NL E -
B4 NL D 0.85
B5 I In 1.63d
B6 NL E 0.50d
B7 NL In 0.36d
B8 I In 1.13
B9 NL E 1.17
aFISH: I, increased LRIG1 copy number; NL, normal LRIG1 copy number. bWestern blot: comparison of the staining in tumour tissue (T) and non-neoplastic tissue (NN). A visual increase (In)/decrease (D) of at least 50% was considered convincing. E, equal levels of protein. cLRIG1 RT-PCR T/NN; LRIG1 mRNA expression level in neoplastic (T) tissue samples divided by levels in matched non-neoplastic (NN) tissue samples. dAs described in the results, at p.9, ratios >1.2 are regarded as significant overexpression and ratios <0. 8 are regarded as significant underexpression in neoplastic tissue compared to non-neoplastic tissue. Samples were consecutively collected. eTwo tumours did not yield RNA of sufficient quality for the RT-PCR analysis.
Table 4 EGFR and ERBB2 quantitative RT-PCR mRNA results in eight patients and matched controls (group B)
Patient number T NN T/NNa
1b
EGFR 4,677 12,493 0.37
ERBB2 24,066 21,817 1.1
2b
EGFR 3,321 14,519 0.22
ERBB2 61,389 6,666 9.2
4
EGFR 20,778 14,842 1.39
ERBB2 9,143 6,742 1.35
5b
EGFR 9,730 15,867 0.61
ERBB2 13,506 3,056 4.41
6
EGFR 1,305 15,871 0.08
ERBB2 10,677 10,080 1.06
7
EGFR 1,024 12,162 0.08
ERBB2 7,639 5,871 1.30
8b
EGFR 4,936 28,111 0.17
ERBB2 17,106 9,263 1.85
9
EGFR 5,540 1,437 3.85
ERBB2 7,330 5,540 1.32
aT/NN: mRNA expression levels of LRIG1 in tumour tissue (T) samples divided by values in matched non-neoplastic (NN) tissue samples. As described in the results, at p.9, ratios >1. 2 are regarded as significant overexpression and ratios <0. 8 are regarded as significant underexpression in neoplastic tissue compared to non-neoplastic tissue. bPatients with increased LRIG1 copy number. No RNA from patient 3 was available for EGFR/ERBB2 analysis.
==== Refs
Bieche I Liderau R Genetic alterations in breast cancer Genes Chromosomes Cancer 1995 14 227 251 8605112
Ross JS Linette GP Stec J Clark E Ayers M Leschly N Symmans WF Hortobagyi GN Pusztai L Breast cancer biomarkers and molecular medicine Expert Rev Mol Diagn 2003 3 573 585 14510178 10.1586/14737159.3.5.573
Friedman LS Ostermeyer EA Lynch ED Szabo CI Anderson LA Dowd P Lee MK Rowell SE Boyd J King MC The search for BRCA1 Cancer Res 1994 54 6374 6382 7987831
Futreal PA Liu Q Shattuck-Eidens D Cochran C Harshman K Tavtigian S Bennett LM Haugen-Strano A Swensen J Miki Y BRCA1 mutations in primary breast and ovarian carcinomas Science 1994 266 120 122 7939630
Klijn JG Berns PM Schmitz PI Foekens JA The clinical significance of epidermal growth factor receptor (EGF-R) in human breast cancer: a review on 5232 patients Endocr Rev 1992 13 3 17 1313356 10.1210/er.13.1.3
Borg A Tandon AK Sigurdsson H Clark GM Ferno M Fuqua SA Killander D McGuire WL HER-2/neu amplification predicts poor survival in node-positive breast cancer Cancer Res 1990 50 4332 4337 1973070
Linderholm BK Lindh B Tavelin B Erlanson M Beckman L Grankvist K Henriksson R Expression of epidermal growth factor receptor 1 (EGFR1) is correlated with over-expression of c-erbB-2, and higher expression of the angiogenic factors basic fibroblast growth factor (BFGF) and vascular endothelial growth factor (VEGF), but do not add prognostic information in primary breast cancer Breast 2003 12 suppl 1 S25 10.1016/S0960-9776(03)80072-3
Elledge RM Allred DC p53 tumor suppressor gene in breast cancer Breast Cancer Res Treat 1994 32 39 47 7819584 10.1007/BF00666204
Linderholm BK Lindahl T Holmberg L Klaar S Lennerstrand J Henriksson R Bergh J The expression of vascular endothelial growth factor correlates with mutant p53 and poor prognosis in human breast cancer Cancer Res 2001 61 2256 2260 11280795
Heimann R Ferguson DJ Hellman S Relationship between nm23, angiogenesis, and the metastatic proclivity of node-negative breast cancer Cancer Res 1998 58 2766 2771 9661889
Nesbit CE Tersak JM Prochownik EV MYC oncogenes and human neoplastic disease Oncogene 1999 18 3004 3016 10378696 10.1038/sj.onc.1202746
Isola J Chu L DeVries S Matsumura K Chew K Ljung BM Waldman FM Alterations in ERBB2-amplified breast carcinomas Clin Cancer Res 1999 5 4140 4145 10632352
Menard S Pupa SM Campiglio M Tagliabue E Biologic and therapeutic role of HER2 in cancer Oncogene 2003 22 6570 6578 14528282 10.1038/sj.onc.1206779
Robanus-Maandag EC Bosch CA Kristel PM Hart AA Faneyte IF Nederlof PM Peterse JL van de Vijver MJ Association of C-MYC amplification with progression from the in situ to the invasive stage in C-MYC-amplified breast carcinomas J Pathol 2003 201 75 82 12950019 10.1002/path.1385
Pandis N Heim S Bardi G Idvall I Mandahl N Mitelman Arm t (1;16) and i (1q) as sole anomalies identify gain of 1q as a primary chromosomal abnormality in breast cancer Genes Chromosomes Cancer 1992 5 235 238 1384676
Malamou-Mitsi VD Syrrou M Georgiou I Pagoulatos G Agnatis NJ Analysis of chromosomal aberrations in breast cancer by CGH. Correlations with Histoprognostic Variables and c-erB-2 immunoexpression J Exp Clin Cancer Res 1999 18 357 361 10606182
Teixeira MR Pandis N Bardi G Andersen JA Heim S Karyotypic comparisons of multiple tumorous and macroscopically normal surrounding tissue samples from patients with breast cancer Cancer Res 1996 56 855 859 8631024
Sato T Akiyama F Sakamoto G Kasumi F Nakamura Y Accumulation of genetic alterations and progression of primary breast cancer Cancer Res 1991 51 5794 5799 1682035
Maitra A Wistuba II Washington C Virmani AK Ashfaq R Milchgrub S Gazdar AF Minna JD High-resolution chromosome 3p allelotyping of breast carcinomas and precursor lesions demonstrates frequent loss of heterozygosity and discontinuous pattern of allele loss Am J Pathol 2001 159 119 130 11438460
Matsumoto S Kasumi F Sakamoto G Onda M Nakamura Y Emi M Deletion mapping of chromosome arm 3p in breast cancers: a 2-cM region on 3p14.3-21.1 and a 5-cM region on 3p24.3-25.1 commonly deleted in tumours Genes Chromosomes Cancer 1997 20 268 274 9365834 10.1002/(SICI)1098-2264(199711)20:3<268::AID-GCC7>3.0.CO;2-0
Yang Q Nakamura M Nakamura Y Yoshimura G Suzuma T Umemura T Shimizu Y Mori I Sakurai T Kakudo K 2-hit inactivation of FHIT by loss of heterozygosity and hypermethylation in breast cancer Clin Cancer Res 2002 8 2890 2893 12231533
Ingvarsson S Sigbjornsdottir BI Huiping C Jonasson JG Agnarsson BA Alterations of the FHIT gene in breast cancer: association with tumor progression and patient survival Cancer Detect Prev 2001 25 292 298 11425271
Nilsson J Vallbo C Guo D Golovleva I Hallberg B Henriksson R Hedman H Cloning, characterization and expression of human LIG1 Biochem Biophys Res Commun 2001 284 1155 1161 11414704 10.1006/bbrc.2001.5092
Nilsson J Starefeldt A Henriksson R Hedman H LRIG1 protein in human cells and tissues Cell Tissue Res 2003 312 65 71 12684867
Gur G Rubin C Katz M Amit I Citri A Nilsson J Amariglio N Henriksson R Rechavi G Hedman H LRIG1 restricts growth factor signaling by enhancing receptor ubiquitylation and degradation EMBO J 2004 23 3270 3281 15282549 10.1038/sj.emboj.7600342
Laederich MB Funes-Duran M Yen L Ingalla E Wu X Carraway KL 3rdSweeney C The leucine-rich repeat protein LRIG1 is a negative regulator of ErbB family receptor tyrosine kinases J Biol Chem 2004 279 47050 47056 15345710 10.1074/jbc.M409703200
Page DL Dupont WD Rogers LW Landenberger M Intraductal carcinoma of the breast: follow-up after biopsy only Cancer 1982 49 751 8 6275978
Thomasson M Hedman H Guo D Ljungberg B Henriksson R LRIG1 and epidermal growth factor receptor in renal cell carcinoma: a quantitative RT-PCR and immunohistochemical analysis Br J Cancer 2003 89 1285 1289 14520461 10.1038/sj.bjc.6601208
Hyman E Kauraniemi P Hautaniemi S Wolf M Mousses S Rozenblum E Ringner M Sauter G Monni O Elkahloun A Impact of DNA amplification on gene expression patterns in breast cancer Cancer Res 2002 62 6240 6245 12414653
Zhao X Li C Paez JG Chin K Janne PA Chen TH Girard L Minna J Christiani D Leo C An integrated view of copy number and allelic alterations in the cancer genome using single nucleotide polymorphism arrays Cancer Res 2004 64 3060 3071 15126342
Rennstam K Ahlstedt-Soini M Baldetorp B Bendahl PO Borg A Karhu R Tanner M Tirkkonen M Isola J Patterns of chromosomal imbalances defines subgroups of breast cancer with distinct clinical features and prognosis. A study of 305 tumours by comparative genomic hybridization Cancer Res 2003 63 8861 8868 14695203
Hedman H Nilsson J Guo D Henriksson R Is LRIG1 a tumour suppressor gene at chromosome 3p14.3? Acta Oncol 2002 41 352 354 12234026 10.1080/028418602760169398
Map Viewer
Tsutsui S Ohno S Murakami S Hachitanda Y Oda S Prognostic value of epidermal growth factor receptor (EGFR) and its relationship to the estrogen receptor status in 1029 patients with breast cancer Breast Cancer Res Treat 2002 71 67 75 11859875 10.1023/A:1013397232011
|
16168117
|
PMC1242137
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jul 6; 7(5):R719-R727
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1279
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12811616811610.1186/bcr1281Research ArticleFunctional interaction between mouse erbB3 and wild-type rat c-neu in transgenic mouse mammary tumor cells Kim Aeree [email protected] Bolin [email protected] Dalia [email protected] Kathy M [email protected] Lynn D [email protected] Christine [email protected] Susan M [email protected] XiaoHe [email protected] Ann D [email protected] Department of Pathology and College of Medicine, Oklahoma University Health Sciences Center (OUHSC), Oklahoma City, OK, USA2 Department of Pathology, College of Medicine, Korea University, Seoul, Korea2005 6 7 2005 7 5 R708 R718 7 12 2004 14 2 2005 24 5 2005 10 6 2005 Copyright © Kim et al. licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License () which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Co-expression of several receptor tyrosine kinases (RTKs), including erbB2 and erbB3, is frequently identified in breast cancers. A member of the RTK family, the kinase-deficient erbB3 can activate downstream signaling via heterodimer formation with erbB2. We studied the expression of RTK receptors in mammary tumors from the wild-type (wt) rat c-neu transgenic model. We hypothesized that physical and functional interactions between the wt rat neu/ErbB2 transgene and mouse ErbB3-encoded proteins could occur, activating downstream signaling and promoting mammary oncogenesis.
Methods
Immunohistochemical and Western blot analyses were performed to study the expression of rat c-neu/ErbB2 and mouse erbB3 in mammary tumors and tumor-derived cell lines from the wt rat c-neu transgenic mice. Co-immunoprecipitation methods were employed to quantitate heterodimerization between the transgene-encoded protein erbB2 and the endogenous mouse erbB3. Tumor cell growth in response to growth factors, such as Heregulin (HRG), epidermal growth factor (EGF), or insulin-like growth factor-1 (IGF-1), was also studied. Post-HRG stimulation, activation of the RTK downstream signaling was determined by Western blot analyses using antibodies against phosphorylated Akt and mitogen-activated protein kinase (MAPK), respectively. Specific inhibitors were then used with cell proliferation assays to study the phosphoinositide-3 kinase (PI-3K)/Akt and MAPK kinase (MEK)/MAPK pathways as possible mechanisms of HRG-induced tumor cell proliferation.
Results
Mammary tumors and tumor-derived cell lines frequently exhibited elevated co-expression of erbB2 and erbB3. The transgene-encoded protein erbB2 formed a stable heterodimer complex with endogenous mouse erbB3. HRG stimulation promoted physical and functional erbB2/erbB3 interactions and tumor cell growth, whereas no response to EGF or IGF-1 was observed. HRG treatment activated both the Akt and MAPK pathways in a dose- and time-dependent manner. Both the PI-3K inhibitor LY 294002 and MEK inhibitor PD 98059 significantly decreased the stimulatory effect of HRG on tumor cell proliferation.
Conclusion
The co-expression of wt rat neu/ErbB2 transgene and mouse ErbB3, with physical and functional interactions between these two species of RTK receptors, was demonstrated. These data strongly suggest a role for erbB3 in c-neu (ErbB2)-associated mammary tumorigenesis, as has been reported in human breast cancers.
==== Body
Introduction
The erbB or epidermal growth factor receptor (EGFR) family forms subclass I of the receptor tyrosine kinase (RTK) superfamily. Type I RTKs are expressed by epithelial, mesenchymal and neural tissues to regulate cell proliferation, differentiation and other important biological functions critical to species development [1]. Dysregulated expression of erbB receptors or mutational events thereof have been implicated in diverse types of human cancers [1,2]. Members of the family include: ErbB1 (also known as EGFR), ErbB2 (also known as Her-2 or neu), ErbB3 (or Her-3) and ErbB4 (or Her-4) [3-7]. erbB2 is an orphan receptor whereas other family members directly bind ligands (like the epidermal growth factor (EGF) and transforming growth factor-α (TGF-α) for EGFR, and HRG for erbB3 and erbB4) to initiate intracellular signaling [8].
ErbB2 may be activated via either ligand-dependent heterodimeric, or ligand-independent homodimeric processes. In the former, erbB2 is the preferred heterodimerization partner for other erbB family receptors with bound ligand [9]. In ligand-independent signaling, erbB2 may be upregulated as a result of gene amplification, promoting homodimerization, or be activated through mutational events. ErbB2 amplification with enhanced protein expression is noted in approximately one-third of invasive human breast cancers [10]. Selected heterodimers may enhance receptor activation and downstream signaling as compared with homodimers [1,11,12]. Although erbB3 lacks a functional kinase to initiate cell signaling [13,14], the erbB2/erbB3 heterodimer complex is believed to be the most biologically active and pro-tumorigenic form of these receptor complexes [15,16].
The erbB receptors and their respective ligands influence a wide range of cellular processes such as proliferation, maturation, survival, apoptosis and angiogenesis [11,17-19]. In general, activated RTKs add phosphorylated tyrosine residues to downstream signaling molecules, such as the p85 subunit of phosphatidylinositol 3-kinase (PI-3K), Shc and/or Grb2 of the mitogen-activated protein kinase (MAPK) pathway. However, because of the complexity of RTK ligand-dependent and -independent mechanisms, the downstream signaling effects may be highly diverse and interactive. RTK-induced signaling is also influenced by, and may modulate, other molecular factors and signaling pathways.
The ErbB2 gene-encoded protein is over-expressed in 25 to 30% of invasive breast and ovarian cancers and has been associated with a poor clinical outcome [20-25].
Evidence of a causal relationship in human breast cancer has been derived from numerous prognostic studies and clinical trials. In vivo and in vitro model systems including transgenic mouse models support a relationship between erbB2 alterations and mammary tumorigenesis. Overexpression of erbB3 is also frequently reported in erbB2 altered breast, ovarian and bladder cancers [23,26,27]. Human breast cancer cell lines commonly co-overexpress both erbB2 and erbB3, further supporting their role in breast carcinogenesis [2,11].
To investigate the role of RTKs in mammary tumorigenesis, transgenic mice bearing the wild-type (wt) or mutated, activated rat c-neu (ErbB2) were generated, and have been widely studied [28-31]. Transgenic mice expressing the activated rat c-neu (with deletion mutations) bear mammary tumors with elevated co-expression of the mutant c-neu/ErbB2 and the endogenous mouse ErbB3-encoded protein [32]. Functional and physical interactions between these two cross-species receptors have not been reported, although interactions have been widely speculated. Transgenic mice bearing the wt-rat c-neu, under control of the mouse mammary tumor virus promoter (MMTV-LTR), typically develop unifocal, well-circumscribed, low-grade tumors after a long latency [29]. In addition to transgene expression and, in some cases, mutation, upregulation of EGFR and p53 have been reported in derived tumors [33,34].
We have used the wt-erbB2 transgenic mouse model to study the effects of exogenous pharmacological or dietary estrogens and anti-estrogens. In particular, we have studied interactions between RTK-associated mammary tumorigenesis and steroid hormones. From the derived mouse tumors, we have established over 150 novel murine cell lines which have proven useful for in vitro studies [35,36]. Most tumor-derived cell lines express significant mouse ErbB3-encoded protein, in addition to high levels of the rat c-neu/ErbB2 transgene. These are also typically negative for ERα but show ERβ protein expression. A similar pattern of receptor expression has also been detected in the mouse mammary tumors.
The co-expression of erbB3 with erbB2 in both the activated and wt-neu/ErbB2 transgenic model systems suggested a biological role for erbB3 in mammary tumor pathogenesis. We hypothesized that physical and functional interactions between these RTK receptors should occur, despite their cross-species molecular structures. Signaling initiated by activated erbB2/erbB3 heterodimers should provide a more potent oncogenic signal than erbB2 homodimers alone. This would require ligand binding, most likely HRG, to activate erbB3. To test this hypothesis, we studied the responsiveness of tumor-derived cell lines to growth factors, including HRG, EGF and insulin-like growth factor-1 (IGF-1); we evaluated the effects of ligand stimulation and heterodimer formation on downstream signaling activation; and we sought evidence of physical interactions between the wt-rat c-neu/erbB2 and the endogenous mouse erbB3.
Materials and methods
Cells and cell culture
Human breast cancer cell lines SKBR-3 and BT-474 were obtained from the American Type Culture Collection (Rockville, MD, USA) and maintained in DMEM and Ham's F-12 medium (1:1, v/v) (Invitrogen Corp, Grand Island, NY, USA) supplemented with 10% FBS (Invitrogen Corp). These cell lines were cultured in a 37°C humidified atmosphere containing 95% air and 5% CO2 and were split twice a week. These human breast cancer cells were used primarily as controls.
Establishment of novel, mouse mammary tumor cell lines
Mammary tumors were obtained from the transgenic mice by surgical removal immediately following euthanasia, according to our approved IACUC protocol. The histological pattern and tumor diagnoses were confirmed by microscopic analysis. These methods have been previously described in detail [35], although the specific cell lines described in this work have not been previously published. In brief, solid tumor tissue was transferred into a tissue culture dish containing PBS. After removal of mammary fat and connective tissues, tumors were minced into small pieces and treated with 0.25% trypsin-EDTA (Invitrogen Corp) at 37°C for 30 min. Cells were subsequently centrifuged at 1,200 rpm for 5 min. After discarding supernatant, cells were suspended in DMEM/F12 medium supplemented with 10% FBS and 1% antibiotics and antimycotics (Invitrogen Corp). These mammary tumor cells (~1.0 × 106 cells/plate) were seeded in tissue culture dishes and kept in a 37°C humidified atmosphere containing 95% air and 5% CO2. The media was changed twice a week to maintain cells in culture. Each line was passaged approximately 20 times before stability was assumed.
Soft agar cloning assays
Soft agar cloning was performed as described previously [35] with some modification. The bottom agar was prepared with a mixture of 1.6 ml of 1 × DMEM/F12 (complete medium), 3.2 ml of 2 × DMEM/F12 (complete medium), and 3.2 ml of 1.25% Noble agar (Sigma Co, St Louis, MO, USA) and maintained at 42°C. From this, 2 ml was pipetted into each well of six-well cell culture plates and agar was allowed harden in the hood. For each well, top agar was a mixture of 0.2 ml of 1 × DMEM/F12, 0.4 ml of 2 × DMEM/F12, and 0.4 ml of 0.95% Noble agar. Five thousand cells (in 80 μl complete medium) were added into the top agar mixture. After vortexing gently, the cell containing top agar was added in a drop-wise fashion onto the bottom agar containing six-well plates (in triplicate per cell line). After resting for 10 min in the hood, the six-well plates were cultured in a 37°C incubator for 3 weeks. Colony counts were obtained under an inverted microscope, from three wells per cell line counting all colonies >50 μM in diameter.
Doubling time in culture
Measurement of cell growth rate in culture was determined using sulforhodamine B (SRB; Sigma Co) assays as previously described [35]. Two thousand cells were seeded into each well of a six-well plate with complete medium. Cells were fixed with 50% trichloroacetic acid (TCA) at 24 h intervals for 3 days. TCA-fixed cells were then stained with 0.4% SRB for 30 min followed by three washes. Protein-bound dye was dissolved in 10 mM Tris base. Plates were read at 565 nM using a micro-plate reader. Cell-doubling time was calculated based on proliferation curves that resulted from changes in SRB absorbance over time. Data represent the means of at least three independent experiments.
Cell proliferation assay
A CellTiter96™ AQ non-radioactive cell proliferation kit (Promega Corp, Madison, WI, USA) was used to determine the responsiveness of cells to various growth factors. Cells were plated onto 96-well plates, 5,000 cells/well for each cell line. Twenty-four hours later, the culture media were replaced by 0.5% FBS in DMEM/F12 fresh medium or the same medium containing 25 ng/ml HRG (R&D Systems, Inc, Minneapolis, MN, USA), 10 ng/ml EGF (Sigma Chemical Co), or 40 ng/ml IGF-1 (R&D Systems, Inc) for another 72 h incubation with 5% CO2 at 37°C. After reading at 490 nM with the micro-plate reader, the percentages of viable cells were determined by reduction of MTS (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium; inner salt) relative to controls. Data reflect the means of at least three independent experiments.
RT-PCR and DNA sequencing analysis
RT-PCR analyses were performed as previously described [37]. The primers specific for rat neu were synthesized according to the literature [38]. Forward primer AB2913, 5'-CGG AAC CCA CAT CAG GCC-3' and reverse primer AB1310, 5'-TTT CCT GCA GCA GCC TAC GC-3' amplify the region corresponding to nucleotides 1492 to 2117 of rat neu cDNA [38]. The PCR products purified from agarose gel using QIAquick Gel Extraction Kit (Qiagen, Inc, San Valencia, CA, USA) were submitted to the core facility at the Oklahoma Medical Research Foundation for direct sequencing.
Immunohistochemistry
Immunohistochemical staining of mammary tumor tissues was performed as previously described [39]. Briefly, after deparaffinization and rehydration, tissue sections were steamed in a 10 mM citrate buffer, pH 6.0, for 30 min. Non-specific reactivity was blocked with 0.3% H2O2 in buffer. For erbB3 immunoassays, CAS Block (Zymed Laboratories, Inc, South San Francisco, CA, USA) and 10% normal horse serum (Vector Laboratories, Inc, Burlingame, CA, USA) were used sequentially. For phospho-Akt immunostaining, we used 1% H2O2 and 5% normal goat serum (Vector Laboratories, Inc) sequentially. Primary antibodies included an anti-erbB2 (reactive with rat c-neu/erbB2 rabbit polyclonal, dilution 1:1000 (Dako, Carpinteria, CA, USA) for 2 h incubation at room temperature), anti-erbB3 (cross-reacts with mouse and human, mouse monoclonal, dilution 1:50 (NeoMarkers, Inc, Fremont, CA, USA), overnight incubation at 4°C), anti-phospho-Akt (rabbit polyclonal, diluted in 5% normal goat serum 1:12.5 (Cell Signaling Technology, Beverly, MA, USA), overnight at 4°C), or anti-phospho-MAPK (E10 monoclonal antibody, diluted in 5% normal goat serum 1:25 (Cell Signaling Technology), overnight at 4°C). After multiple washes with buffer, tissue sections were sequentially incubated for 30 min at room temperature with diluted biotinylated secondary antibody (1:500; Dako) and VECTASTAIN Elite ABC reagent (Vector Laboratories, Inc) diluted in PBS. After reaction with diaminobenzidine (Dako) and counterstaining with hematoxylin, tumors were individually examined. Each slide was evaluated in its entirety for antigen expression, cell type and histopathological diagnoses.
Immunoprecipitation and Western blot analysis
Immunoprecipitation and Western blot assays were performed as previously described [40]. Briefly, cells were lysed in NP-40 lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5% NP-40, 50 mM NaF, 1 mM Na3VO4, 1 mM phenylmethylsulfonyl fluoride, 25 μg/ml leupeptin, 25 μg/ml aprotinin). The supernatants were cleared by centrifugation. Protein concentrations were measured using the Coomassie plus protein assay reagent (Pierce Chemical Co, Rockford, IL, USA). Total cell lysates containing 200 μg of protein were subjected to immunoprecipitation in the presence of 1 μg anti-erbB2 antibody (mouse monoclonal antibody, Ab-4; Oncogene Science Products, Cambridge, MA, USA) for 2 h at 4°C, followed by incubation with immobilized protein A-agarose (Roche Diagnostics Corp, Indianapolis, IN, USA) at 4°C overnight with rotation. For Western blot analyses, the immunoprecipitates or equal amounts of crude extracts were boiled in Laemmli SDS-sample buffer, resolved by SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose (Bio-Rad Laboratories, Hercules, CA, USA), and probed with different primary antibodies. After the blots were incubated for another 1 h at room temperature with horseradish peroxidase-labeled secondary antibody (goat anti-rabbit IgG or goat anti-mouse IgG; Perkin Elmer, Boston, MA, USA), the signals were detected using the Enhanced Chemiluminance assay (Amersham Life Science Inc., Arlington Heights, IL, USA) according to the manufacturer's instructions.
Results
Co-expression of erbB2 and erbB3 protein in tumor-derived cell lines and tumors
Western blot analyses were used to determine erbB2 and erbB3 protein expression in tumor-derived cell lines (and the control SKBR-3 human breast cancer cell line). The majority of tumor-derived cell lines expressed moderate to high levels of both erbB3 and erbB2 (Fig. 1). In general, lines with the highest erbB2 expression showed the highest levels of erbB3 protein. Tyrosine phosphorylation (activation) of these receptors was examined by Western blots using antibodies specific for phophorylated erbB2 (P-erbB2) or phosphorylated erbB3 (P-erbB3). Tumor lines with co-overexpression of both proteins showed higher P-erbB2 and P-erbB3 levels (Fig. 1). The intensity of P-erbB2 and P-erbB3 signals did not necessarily correlate with their corresponding protein levels. The expression of either receptor protein was undetectable in only one of our novel, derived tumor cell lines (78423). AIB-1 (also called SRC3, RAC3, ACTR and p/CIP), a co-activator of estrogen receptor commonly amplified in breast cancer cells [41], was used as a loading control. Expression of AIB-1 further established the origin of these cells as mammary-derived.
To confirm the transformed characteristics of these lines, soft agar cloning assays (which quantitate anchorage-independent cloning capability) were used. All six tumor-derived cell lines formed colonies in soft agar. Colony formation was variable when comparing one cell line with another (range 17–180, Table 1). There was no correlation between the ability of a cell line to form anchorage-independent clones and the expression levels of erbB2 or erbB3.
Immunohistochemical methods were used to visualize RTK expression and downstream signaling (protein activation) by tumors in situ. Tumors showed strong and typically diffuse co-expression of both erbB2 and erbB3. The only exception to this was the mammary tumor 78423 R1, the progenitor of the cell line that did not co-express erbB2 and erbB3 discussed above. We also studied RTK signaling activation in situ, using phosphospecific antibodies. Phosphorylated-Akt (P-Akt) showed cytoplasmic and membranous staining, which was less diffuse than the erbB-2 expression. Phosphorylated-MAPK (P-MAPK) was the most selectively expressed, typically expressed by clustered or isolated tumor cells as shown in Fig. 2 (left panel) with tumor 78617 R3. The majority of tumor cells from 78423 R1 were erbB3 negative, although some cells showed weak erbB2 protein expression. In this later tumor, P-Akt staining was weak with clustered or isolated tumor cells and no staining for P-MAPK was observed (Fig. 2, right panel). The histological, cytological and biological features of these tumors have been reported elsewhere [36,42]. As a control, we also studied cytokeratin expression and all tumors were positive. This confirmed the epithelial nature of these tumors (data not shown).
Sequencing analyses of the transgene neu in established mammary tumor-derived cell lines
As alluded to above, in-frame deletions of 7–12 amino acids have been reported in the extracellular region of the transgene, proximal to the transmembrane domain [38]. To study the mutational status of tumor-derived lines, we performed RT-PCR amplification of exactly the same region followed by direct sequencing analysis. The PCR primers used were specific for rat neu and were designed to amplify the 603 bp extracellular region [38]. Of six tumor-derived cell lines used in this manuscript and therefore studied for mutation, only four showed PCR gene amplification (Fig. 3a). Of these, the strongest PCR signal was seen in 85819 cells. These data are consistent with our Western blot results that showed overexpression of the rat neu/erbB2 in only the four PCR-positive lines (Fig. 1). Direct sequencing of the PCR products revealed no deletion mutations in the amplified product. Sequencing showed three of the four were wt rat neu cDNA sequence. Sequencing data from the 83923 cells indicated a mixture of two kinds of neu cDNA. Using a reverse primer, we verified that both wt and point mutation neu transcripts co-existed in 83923 cells (Fig. 3b). This suggests biclonal populations or a heterozygous mutation. Further studies and sub-cloning are in process.
Mammary tumor cell response to growth factors corresponds with erbB receptor data
To study the functionality and interactions of the erbB receptors, 78423 and other three representative mouse mammary tumor-derived lines with the highest expression of wt erbB2 and co-expression of erbB3 were chosen for further study. Baseline proliferation was determined using monolayer culture conditions and the SRB assay (Fig. 4a). Some variability in the basal doubling time was observed between these cell lines. The mouse mammary tumor cell lines 78423, 78617, 85815 and 85819 showed population doubling times of 15.15 ± 1.10, 16.25 ± 1.40, 30.85 ± 2.31 and 20.35 ± 1.89 h, respectively. Using an MTS assay, we then tested the response of these lines to EGF, HRG and insulin-like growth factor (IGF)-1 (Fig. 4b). HRG strongly stimulated the proliferation of three of the four mouse mammary tumor cell lines (78617, 85815, 85819) with overexpression of both erbB2 and erbB3. Proliferation was not induced by EGF or IGF-1, which bind to EGFR and IGF-1 receptor, respectively. HRG also promoted the growth of SKBR-3 and BT-474 human breast cancer cells (controls). These data strongly support a functional interaction between the wt-rat neu/ErbB2 and endogenous mouse erbB3.
HRG activation of PI-3K/Akt and MAPK kinase (MEK)/MAPK signaling promotes mammary tumor cell growth
It is well documented that the MEK/MAPK and PI-3K/Akt pathways are the two major signal transduction pathways downstream of the erbB receptors [11,17-19]. To determine which signaling pathways were activated in the mouse-derived mammary tumor cells exposed to HRG, we performed Western blots to detect P-MAPK or P-Akt. With 2 h of HRG treatment, both P-Akt and P-MAPK increased in the 85815 and 85819 mouse mammary tumor cell lines (Fig. 5a). This study included a series of HRG concentrations, and stimulation was maximal at a concentration of 2.5 ng/ml. Next, we performed a time-course analysis to further verify these results. HRG stimulated both Akt and MAPK in 85815 and 85819 cells, whereas it had no effect on Akt or MAPK activation in the 78423 cells (Fig. 5b). These data were consistent with the results of minimal stimulation by HRG in this cell line (Fig. 4b). In aggregate, these data suggest that HRG induces activation of both MEK/MAPK and PI-3K/Akt signaling transduction pathways in mammary tumor cells with elevated expression levels of both the transgene rat c-neu/ErbB2 and the endogenous mouse ErbB3 gene. This activation was both dose- and time-dependent.
To study cross-species functional interactions between the rat c-neu/ErbB2 transgene and mouse ErbB3, we evaluated tumor and tissue expression in vivo, ligand-associated interactions, and signaling in vitro. Immunohistochemical studies showed cytoplasmic P-Akt and P-MAPK expression in tumor cells with erbB2 and erbB3 co-expression, predominantly a perivascular distribution. In rare tumors without erbB2 and erbB3 expression (e.g. 78423 R1), the perivascular distribution was not identified and only rare cells showed immunoreactivity. This evidence of perivascular pathway activation suggests that ligand-associated signaling via erbB3 may be involved. Ligand-associated signaling probably provides enhanced growth or pro-tumorigenic signaling, in addition to ligand-independent, transgene activation. Our data, and those from others showing frequent erbB3 upregulation in transgenic mice bearing activated neu/ErbB2, suggest that the concomitant upregulation of erbB3 and ligand-associated signaling may be an important additional factor in both wt and activated neu/ErbB2-associated mammary tumor development. To further define the role of HRG (ligand)-associated signaling, we utilized derived cell lines and specific inhibitors in vitro. The PI-3K inhibitor LY294002 was significantly more potent than the MEK inhibitor PD98059 in blocking the stimulatory effects of HRG (Fig. 7). Hence, while the MEK/MAPK and PI-3K/Akt signaling cascades both contribute HRG induced proliferation, the PI-3K/Akt pathway appears to provide the dominant response.
Physical interaction between wt-rat-c-neu/ErbB2 and endogenous mouse erbB3
The erbB2/erbB3 complex is believed to be the most biologically active erbB heterodimer [43,44], with potent activation of the downstream signaling cascade [13,14]. Since both erbB2 and erbB3 were highly expressed by our mammary tumor cell lines and HRG-promoted tumor cell proliferation, we sought physical evidence that the wt-rat-neu/ErbB2 could form a complex with the endogenous mouse erbB3. Immunoprecipitation of erbB2, followed by Western blot analysis for erbB2 and erbB3 (Fig. 8a) showed a low level of complex formation between these receptors in untreated cell lines. HRG treatment significantly increased the physical interaction between the rat transgene and mouse erbB3 in two out of three cell lines. The antibody we used for immunoprecipitation (c-neu Ab-4) appeared to be wt-rat-neu/ErbB2-specific, because human erbB2 was not immunoprecipitated from SKBR-3 cell lysates (Fig. 8a, upper panel), although it was expressed by SKBR-3 cells (Fig. 8b, Western blot analysis with c-neu Ab-3). HRG treatment (for 2 h) did not increase the total protein levels of erbB2 or erbB3 as compared with untreated cell lines (Fig. 8b).
Discussion
We have shown that transgenic mice bearing the wt-rat c-neu gene, under control of the MMTV promoter, develop mammary tumors that overexpress the rat c-neu transgene [45,46] and the endogenous mouse erbB3 protein, in the vast majority of cases. We have shown a functional interaction between these two important RTK receptors and a role for ligand-induced signaling in vitro and in vivo. While others have reported that transgenic mice bearing activated forms of rat c-neu/erbB2 have co-expression of erbB2 and endogenous erbB3 in mammary tumors [32], direct physical and functional interactions between these two species receptors have not previously been reported.
Deletion mutants of the neu oncogene have been reported in two out of three of the mammary tumors derived from this wt-rat c-neu transgenic model [38]. We did not find the same mutation rate or type in selected tumor-derived cell lines. However, we have identified a potential point mutation in 83923 cells (Fig. 3). This missense mutation is located inside the same extracellular region of neu where the deletion mutations have been reported. This particular mutation changes the amino acid 654 serine (codon AGC) into cysteine (codon TGC). It is different from the active neu mutation G664V reported in the transmembrane domain [47]. The biological significance of the newly discovered S654C mutant neu is not yet known.
Using ligand stimulation with or without specific inhibitors, we have studied RTK-induced signaling in response to HRG and have shown activation of both PI-3K/Akt and the MEK/MAPK signal transduction pathways. A greater role for PI-3K/Akt signaling was suggested in response to HRG treatment (Fig. 7). PI-3K/Akt signaling is known to be regulated by erbB2-mediated tyrosine kinase activity. This pathway plays a crucial role in cell proliferation and survival [18] and has been associated with the pathogenesis of human breast cancers. PI-3K/Akt activation has also been cited as a key pathway that influences chemo-resistance patterns [48,49]. Akt is frequently upregulated in ErbB2 amplified or overexpressing human breast cancer cells. These similarities between our transgenic model and human breast carcinogenesis suggest that the model and derived tumor cell lines may be a useful resource to study ligand dependent and independent RTK signaling in vivo and in vitro.
As a major ligand for erbB3, HRG is known to bind to erbB3, foster heterodimer complex formation and promote potent downstream signaling [12]. HRG can thus promote mammary tumorigenesis, cell growth, differentiation and phenotypic aggression [50]. Our immunohistochemical studies of tumors for phosphorylated proteins facilitated studies of the cellular location and architectural context of signaling. We noted enhanced phosphorylated Akt and MAPK in a perivascular distribution in mammary tumors, with overexpression of both erbB2 and erbB3 (Fig. 6), suggesting that circulating HRG may enhance the physical and functional erbB2/erbB3 interactions in vivo, similar to what we observe in vitro. This study has focused primarily on erbB3, whereas others have demonstrated upregulation of EGFR in tumors (by immunohistochemistry and Western blot) in the same model system [33]. Low and variable expression of EGFR has also been found in mammary tumors that develop in transgenic mice bearing activated forms of rat c-neu/ErbB2 [32]. Using in vitro analyses of the tumor-derived cell lines, we have found no significant physical or functional interaction between EGFR (erbB1) and erbB2 in the presence of EGF (data not shown). However, by immunohistochemical study, we also detected erbB1 expression at the tumor periphery as reported by DiGiovanna [33]. These data suggest to us that erbB3 plays a more significant role in tumorigenesis than erbB1 in this model system.
These data and this model probably have relevance to human breast cancer biology and treatment strategies. We have reported that only a minority of erbB2-altered invasive human breast cancers have overexpression of erbB1 (EGFR) and activation of erbB2 [51]. Given the complexity of the RTK receptors, various ligands and downstream signaling, it is likely that combinations of these factors including erbB3 contribute to cell signaling, biological behavior and treatment response [52,53]. To date, the role of erbB3 in human breast carcinogenesis is not well defined, although many investigators have suggested that HRG-associated signaling may be important. In view of these complexities, it is not surprising that erbB2 aberrant breast cancers have shown variable responses to anti-erbB2 therapeutics [52,53]. It is widely believed that co-expression of other erbB RTK family members may be one mechanism of Herceptin resistance [54]. Ligand-induced heterodimerization between erbB3 and erbB2, the most potent signaling complex amongst the various heterodimers, is one likely mechanism of Herceptin resistance [55]. More detailed investigations using banks of human tumors and clinical trial-associated specimens, to define the incidence of erbB3 abnormalities, functional complex formation and downstream signaling, may provide important new clues regarding these interactions and their role in breast carcinogenesis.
Conclusion
Our results indicate that over-expression of endogenous mouse erbB3 plays an important role in the development and progression of mammary tumors that arise in mice bearing the wt-rat c-neu transgene. The functional and physical interactions between these important cross-species erbB receptors result in activation of both PI-3K/Akt and MEK/MAPK signaling. These data support the concept that ligand-dependent and -independent signaling through erbB2 may promote mammary tumorigenesis in these transgenic mice, similar to what is observed in human breast cancers.
Abbreviations
DMEM = Dulbecco's modified Eagle's medium; EGF = epidermal growth factor; EGFR = EGF receptor; ER = estrogen receptor; FBS = fetal bovine serum; HRG = heregulin; IGF-1 = insulin-like growth factor-1; mAb = monoclonal antibody; MAPK = mitogen-activated protein kinase; MEK = MAPK kinase; MMTV = mouse mammary tumor virus; PBS = phosphate-buffered saline; RTK = receptor tyrosine kinase; PI-3K = phosphoinositide 3-kinase; RT-PCR = reverse transcription-polymerase chain reaction; SRB = sulforhodamine B; TCA = trichloroacetic acid; TGF-α = transforming growth factor-α.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
The authors' contributions to this research work are reflected in the order shown, with the exception of ADT who supervised the research and finalized the report. AK, BL and DOE carried out most of the experiments. AK and BL drafted the manuscript. KMA collected mammary tumors from the transgenic mice. LDJ performed immunohistochemistry analysis. CM maintained tumor cell culture. SME, XY and ADT conceived the study and participated in its design and coordination. All authors read and approved the final manuscript.
Acknowledgements
This work was supported by NIH 1RO1 CA82848 and 1P50 CA89018 to ADT.
Figures and Tables
Figure 1 Protein expression of erbB2, P-erbB2, erbB3, P-erbB3 and AIB1 in breast tumor cell lines. The cell lysates from human breast cancer cell line SKBR-3 and the six novel, mammary tumor-derived cell lines were prepared as described in Materials and methods. 50 μg total cell lysates were used for Western blot analyses with antibodies directed against erbB2 (c-neu Ab-3, Oncogene Research Products), P-erbB2 (clone PN2A, NeoMarkers, Inc), erbB3 (Ab-7, NeoMarkers, Inc), P-erbB3 (clone 21D3, Cell Signaling Technology) and AIB1 (clone 34, BD Biosciences Pharmingen, San Diego, CA, USA).
Figure 2 Immunohistochemical staining for erbB2, erbB3, phospho-Akt (P-Akt) and phospho-MAPK (P-MAPK) in mammary tumor tissues. Representative photomicrographs were taken from the similar area of 78617 R3 and 78423 R1 mammary tumor sections (40X).
Figure 3 RT-PCR and sequencing analyses of partial extracellular domain of rat neu in the tumor-derived cell lines. (a) RT-PCR analyses of rat neu. Total RNA isolated from the indicated tumor cells by the TRIZOL reagent was analyzed with electrophoresis using a 1% agarose gel containing ethidium bromide and visualized under UV light (bottom). First-strand cDNA was synthesized using a kit from Roche Diagnostics Corp. The partial extracellular domain of rat neu was amplified with specific primers. The PCR products were separated on a 1.2% agarose gel containing ethidium bromide and visualized under UV light (top). (b) Partial sequencing of the PCR product from 83923 cells with reverse primer AB1310. The mixture of wt nucleotide T (red) and mutant nucleotide A (green) is indicated by an arrow.
Figure 4 Proliferation of the tumor-derived cell lines and their responsiveness to growth factors. (a) The tumor-derived cell lines 78423, 78617, 85815 and 85819 were subjected to SRB assays for measurement of cell growth rate as described in Materials and methods. The means of at least three independent experiments were plotted. SD for each point was less than 10%. (b) The indicated breast cancer cells (5 × 103) in 0.1 ml culture media were plated onto 96-well plates. After 24 h incubation, cells were grown in either 0.1 ml fresh medium with 0.5% FBS as control, or 0.1 ml same medium containing either 25 ng/ml HRG or 10 ng/ml EGF, and 40 ng/ml IGF-1. Cells were incubated at 37°C with 5% CO2 for another 72 h, and the percentages of surviving cells from each group relative to controls, defined as 100% survival, were determined by reduction of MTS. Data reflect the means of at least three independent experiments.
Figure 5 Effects of HRG on the phosphorylation of Akt and MAPK in tumor-derived cells. (a) 85815 and 85819 cells were cultured overnight in medium containing 0.5% FBS before being exposed to HRG at the indicated concentrations for 2 h. Cells were harvested and 50 μg total cell lysates were subjected to Western blot analysis for total Akt, phosphorylated Akt, total ERK2 (polyclonal antibody C-14; Santa Cruz Biotechnology, Inc, Santa Cruz, CA, USA), and phosphorylated MAPK (E10 mAb; Cell Signaling Technology) expression. β-actin was used as loading control. (b) 78423, 85815 and 85819 cells were cultured overnight in medium containing 0.5% FBS before being exposed to 2.5 ng/ml HRG for the indicated time intervals. At each time point, cells were harvested and 50 μg total cell lysates were subjected to Western blot analysis for total Akt, phosphorylated Akt, total ERK2, and phosphorylated MAPK expression. β-actin was used as loading control.
Figure 6 Immunohistochemical staining for phospho-Akt (P-Akt) and phospho-MAPK (P-MAPK) in 78617 R3 and 78423 R1 mammary tumor tissues. Procedure used was similar to Fig. 2. Representative photomicrographs were taken from perivascular areas of the tumor sections (20X).
Figure 7 Inhibitory effects of PD98059 and LY294002 blocking HRG-mediated tumor cell proliferation. The indicated breast cancer cells (5 × 103) in 0.1 ml culture media were plated onto 96-well plates. After 24 h incubation, cells were grown in either 0.1 ml fresh medium with 0.5% FBS as control, or 0.1 ml same medium containing PD98059 (6.7 μM for 78617 and 85815, 5 μM for 85819), or LY294002 (3.3 μM for 78617 and 85815, 2.5 μM for 85819), or 25 ng/ml HRG alone or in combination of the same concentrations of HRG and PD98059, or HRG and LY294002. Cells were incubated at 37°C with 5% CO2 for another 72 h, and the percentages of surviving cells from each group relative to controls, defined as 100% survival, were determined by reduction of MTS. Data reflect the means of at least three independent experiments.
Figure 8 Increased physical associations between wt-rat c-neu protein and mouse erbB3 by HRG treatment. (a) The indicated breast cancer cells were cultured overnight in medium containing 0.5% FBS, and then incubated with or without 2.5 ng/ml HRG for another 2 h. Cell lysates were prepared and 200 μg of total protein was subjected to immunoprecipitation with anti-erbB2 specific antibody (c-neu Ab-4), followed by Western blot analysis for erbB2 (c-neu Ab-3) and erbB3. The bar graph underneath was obtained by densitometry analysis. The relative signal intensities of erbB3 were measured by EAGLE EYE™ II (Stratagene, La Jolla, CA, USA). (b) The same bath of 50 μg total cell lysates was subjected to Western blot analysis with antibodies directed against erbB2 and erbB3.
Table 1 Anchorage-independent cloning of mammary tumor-derived cell lines
Cell Lines SKBR-3a 78423 78617 78717 83923 85815 85819
Colonies in soft agarb 422 87 49 24 41 180 17
aHuman breast cancer cell line SKBR-3 was used as positive control.
bColony numbers represent average of triplicates for each cell line.
==== Refs
Olayioye MA Neve RM Lane HA Hynes NE The ErbB signaling network: receptor heterodimerization in development and cancer EMBO J 2000 19 3159 3167 10880430 10.1093/emboj/19.13.3159
deFazio A Chiew YE Sini RL Janes PW Sutherland RL Expression of c-erbB receptors, heregulin and oestrogen receptor in human breast cell lines Int J Cancer 2000 87 487 498 10918187 10.1002/1097-0215(20000815)87:4<487::AID-IJC5>3.3.CO;2-A
Coussens L Yang-Feng TL Liao YC Chen E Gray A McGrath J Seeburg PH Libermann TA Schlessinger J Francke U Tyrosine kinase receptor with extensive homology to EGF receptor shares chromosomal location with neu oncogene Science 1985 230 1132 1139 2999974
King CR Kraus MH Aaronson SA Amplification of a novel v-erbB-related gene in a human mammary carcinoma Science 1985 229 974 976 2992089
Plowman GD Culouscou JM Whitney GS Green JM Carlton GW Foy L Neubauer MG Shoyab M Ligand-specific activation of HER4/p180erbB4, a fourth member of the epidermal growth factor receptor family Proc Natl Acad Sci USA 1993 90 1746 1750 8383326
Kraus MH Issing W Miki T Popescu NC Aaronson SA Isolation and characterization of ERBB3, a third member of the ERBB/epidermal growth factor receptor family: evidence for overexpression in a subset of human mammary tumors Proc Natl Acad Sci USA 1989 86 9193 9197 2687875
Plowman GD Whitney GS Neubauer MG Green JM McDonald VL Todaro GJ Shoyab M Molecular cloning and expression of an additional epidermal growth factor receptor-related gene Proc Natl Acad Sci USA 1990 87 4905 4909 2164210
Riese DJ 2ndStern DF Specificity within the EGF family/ErbB receptor family signaling network Bioessays 1998 20 41 48 9504046 10.1002/(SICI)1521-1878(199801)20:1<41::AID-BIES7>3.0.CO;2-V
Graus-Porta D Beerli RR Daly JM Hynes NE ErbB-2, the preferred heterodimerization partner of all ErbB receptors, is a mediator of lateral signaling EMBO J 1997 16 1647 1655 9130710 10.1093/emboj/16.7.1647
Penuel E Akita RW Sliwkowski MX Identification of a region within the ErbB2/HER2 intracellular domain that is necessary for ligand-independent association J Biol Chem 2002 277 28468 28473 12000754 10.1074/jbc.M202510200
Holbro T Beerli RR Maurer F Koziczak M Barbas CF 3rdHynes NE The ErbB2/ErbB3 heterodimer functions as an oncogenic unit: ErbB2 requires ErbB3 to drive breast tumor cell proliferation Proc Natl Acad Sci USA 2003 100 8933 8938 12853564 10.1073/pnas.1537685100
Tzahar E Waterman H Chen X Levkowitz G Karunagaran D Lavi S Ratzkin BJ Yarden Y A hierarchical network of interreceptor interactions determines signal transduction by Neu differentiation factor/neuregulin and epidermal growth factor Mol Cell Biol 1996 16 5276 5287 8816440
Harari D Yarden Y Molecular mechanisms underlying ErbB2/HER2 action in breast cancer Oncogene 2000 19 6102 6114 11156523 10.1038/sj.onc.1203973
Yarden Y Sliwkowski MX Untangling the ErbB signalling network Nat Rev Mol Cell Biol 2001 2 127 137 11252954 10.1038/35052073
Alimandi M Romano A Curia MC Muraro R Fedi P Aaronson SA Di Fiore PP Kraus MH Cooperative signaling of ErbB3 and ErbB2 in neoplastic transformation and human mammary carcinomas Oncogene 1995 10 1813 1821 7538656
Wallasch C Weiss FU Niederfellner G Jallal B Issing W Ullrich A Heregulin-dependent regulation of HER2/neu oncogenic signaling by heterodimerization with HER3 EMBO J 1995 14 4267 4275 7556068
Chang HW Aoki M Fruman D Auger KR Bellacosa A Tsichlis PN Cantley LC Roberts TM Vogt PK Transformation of chicken cells by the gene encoding the catalytic subunit of PI 3-kinase Science 1997 276 1848 1850 9188528 10.1126/science.276.5320.1848
Hermanto U Zong CS Wang LH ErbB2-overexpressing human mammary carcinoma cells display an increased requirement for the phosphatidylinositol 3-kinase signaling pathway in anchorage-independent growth Oncogene 2001 20 7551 7562 11709727 10.1038/sj.onc.1204964
Pearson G Robinson F Beers Gibson T Xu BE Karandikar M Berman K Cobb MH Mitogen-activated protein (MAP) kinase pathways: regulation and physiological functions Endocr Rev 2001 22 153 183 11294822 10.1210/er.22.2.153
Slamon DJ Clark GM Wong SG Levin WJ Ullrich A McGuire WL Human breast cancer: correlation of relapse and survival with amplification of the HER-2/neu oncogene Science 1987 235 177 182 3798106
Slamon DJ Godolphin W Jones LA Holt JA Wong SG Keith DE Levin WJ Stuart SG Udove J Ullrich A Studies of the HER-2/neu proto-oncogene in human breast and ovarian cancer Science 1989 244 707 712 2470152
Paterson MC Dietrich KD Danyluk J Paterson AH Lees AW Jamil N Hanson J Jenkins H Krause BE McBlain WA Correlation between c-erbB-2 amplification and risk of recurrent disease in node-negative breast cancer Cancer Res 1991 51 556 567 1670762
Gullick WJ Love SB Wright C Barnes DM Gusterson B Harris AL Altman DG c-erbB-2 protein overexpression in breast cancer is a risk factor in patients with involved and uninvolved lymph nodes Br J Cancer 1991 63 434 438 1672253
Andrulis IL Bull SB Blackstein ME Sutherland D Mak C Sidlofsky S Pritzker KP Hartwick RW Hanna W Lickley L neu/erbB-2 amplification identifies a poor-prognosis group of women with node-negative breast cancer. Toronto Breast Cancer Study Group J Clin Oncol 1998 16 1340 1349 9552035
Thor AD Liu S Edgerton S Moore D 2ndKasowitz KM Benz CC Stern DF DiGiovanna MP Activation (tyrosine phosphorylation) of ErbB-2 (HER-2/neu): a study of incidence and correlation with outcome in breast cancer J Clin Oncol 2000 18 3230 3239 10986055
Bieche I Onody P Tozlu S Driouch K Vidaud M Lidereau R Prognostic value of ERBB family mRNA expression in breast carcinomas Int J Cancer 2003 106 758 765 12866037 10.1002/ijc.11273
Chow NH Chan SH Tzai TS Ho CL Liu HS Expression profiles of ErbB family receptors and prognosis in primary transitional cell carcinoma of the urinary bladder Clin Cancer Res 2001 7 1957 1962 11448910
Bouchard L Lamarre L Tremblay PJ Jolicoeur P Stochastic appearance of mammary tumors in transgenic mice carrying the MMTV/c-neu oncogene Cell 1989 57 931 936 2567634 10.1016/0092-8674(89)90331-0
Guy CT Webster MA Schaller M Parsons TJ Cardiff RD Muller WJ Expression of the neu protooncogene in the mammary epithelium of transgenic mice induces metastatic disease Proc Natl Acad Sci USA 1992 89 10578 10582 1359541
Muller WJ Sinn E Pattengale PK Wallace R Leder P Single-step induction of mammary adenocarcinoma in transgenic mice bearing the activated c-neu oncogene Cell 1988 54 105 115 2898299 10.1016/0092-8674(88)90184-5
Andrechek ER Hardy WR Siegel PM Rudnicki MA Cardiff RD Muller WJ Amplification of the neu/erbB-2 oncogene in a mouse model of mammary tumorigenesis Proc Natl Acad Sci USA 2000 97 3444 3449 10716706 10.1073/pnas.050408497
Siegel PM Ryan ED Cardiff RD Muller WJ Elevated expression of activated forms of Neu/ErbB-2 and ErbB-3 are involved in the induction of mammary tumors in transgenic mice: implications for human breast cancer EMBO J 1999 18 2149 2164 10205169 10.1093/emboj/18.8.2149
DiGiovanna MP Lerman MA Coffey RJ Muller WJ Cardiff RD Stern DF Active signaling by Neu in transgenic mice Oncogene 1998 17 1877 1884 9778054 10.1038/sj.onc.1202091
Li B Rosen JM McMenamin-Balano J Muller WJ Perkins AS neu/ERBB2 cooperates with p53-172H during mammary tumorigenesis in transgenic mice Mol Cell Biol 1997 17 3155 3163 9154814
Jeruss JS Liu NX Chung Y Magrane G Waldman F Edgerton S Yang X Thor AD Characterization and chromosomal instability of novel derived cell lines from a wt-erbB-2 transgenic mouse model Carcinogenesis 2003 24 659 664 12727793 10.1093/carcin/bgg001
Yang X Edgerton SM Kosanke SD Mason TL Alvarez KM Liu N Chatterton RT Liu B Wang Q Kim A Hormonal and dietary modulation of mammary carcinogenesis in mouse mammary tumor virus-c-erbB-2 transgenic mice Cancer Res 2003 63 2425 2433 12750262
Liu B Edgerton SM Yang X Kim A Ordonez-Ercan D Mason TL Alvarez KM McKimmey C Liu N Thor AD Low dose dietary phytoestrogen abrogates tamoxifen associated mammary tumor prevention Cancer Res 2005 65 879 886 15705886
Siegel P Dankort D Hardy W Muller W Novel activating mutations in the neu proto-oncogene involved in induction of mammary tumors Mol Cell Biol 1994 14 7068 7077 7935422
Thor AD Berry DA Budman DR Muss HB Kute T Henderson IC Barcos M Cirrincione C Edgerton S Allred C erbB-2, p53, and efficacy of adjuvant therapy in lymph node-positive breast cancer [comment] J Natl Cancer Inst 1998 90 1346 1360 9747866 10.1093/jnci/90.18.1346
Liu B Fang M Lu Y Mendelsohn J Fan Z Fibroblast growth factor and insulin-like growth factor differentially modulate the apoptosis and G1 arrest induced by anti-epidermal growth factor receptor monoclonal antibody Oncogene 2001 20 1913 1922 11313939 10.1038/sj.onc.1204277
Anzick SL Kononen J Walker RL Azorsa DO Tanner MM Guan XY Sauter G Kallioniemi OP Trent JM Meltzer PS AIB1, a steroid receptor coactivator amplified in breast and ovarian cancer Science 1997 277 965 968 9252329 10.1126/science.277.5328.965
Kosanke S Edgerton SM Moore D 2nd Yang X Mason T Alvarez K Jones L Kim A Thor AD Mammary tumor heterogeneity in wt-ErbB-2 transgenic mice Comp Med 2004 54 280 287 15253274
Riese DJ 2ndvan Raaij TM Plowman GD Andrews GC Stern DF The cellular response to neuregulins is governed by complex interactions of the erbB receptor family Mol Cell Biol 1995 15 5770 5776 [Erratum Mol Cell Biol 1996, 16:735] 7565730
Pinkas-Kramarski R Soussan L Waterman H Levkowitz G Alroy I Klapper L Lavi S Seger R Ratzkin BJ Sela M Yarden Y Diversification of Neu differentiation factor and epidermal growth factor signaling by combinatorial receptor interactions EMBO J 1996 15 2452 2467 8665853
Sacco MG Gribaldo L Barbieri O Turchi G Zucchi I Collotta A Bagnasco L Barone D Montagna C Villa A Establishment and characterization of a new mammary adenocarcinoma cell line derived from MMTV neu transgenic mice Breast Cancer Res Treat 1998 47 171 180 9497105 10.1023/A:1005988715285
Campbell MJ Wollish WS Lobo M Esserman LJ Epithelial and fibroblast cell lines derived from a spontaneous mammary carcinoma in a MMTV/neu transgenic mouse In Vitro Cell Dev Biol Anim 2002 38 326 333 12513120 10.1290/1071-2690(2002)038<0326:EAFCLD>2.0.CO;2
Bargmann CI Hung MC Weinberg RA Multiple independent activations of the neu oncogene by a point mutation altering the transmembrane domain of p185 Cell 1986 45 649 657 2871941 10.1016/0092-8674(86)90779-8
Bacus SS Altomare DA Lyass L Chin DM Farrell MP Gurova K Gudkov A Testa JR AKT2 is frequently upregulated in HER-2/neu-positive breast cancers and may contribute to tumor aggressiveness by enhancing cell survival Oncogene 2002 21 3532 3540 12032855 10.1038/sj.onc.1205438
Knuefermann C Lu Y Liu B Jin W Liang K Wu L Schmidt M Mills GB Mendelsohn J Fan Z HER2/PI-3K/Akt activation leads to a multidrug resistance in human breast adenocarcinoma cells Oncogene 2003 22 3205 3212 12761490 10.1038/sj.onc.1206394
Lupu R Cardillo M Cho C Harris L Hijazi M Perez C Rosenberg K Yang D Tang C The significance of heregulin in breast cancer tumor progression and drug resistance Breast Cancer Res Treat 1996 38 57 66 8825123 10.1007/BF01803784
DiGiovanna MP Stern DF Edgerton S Whalen SG Moore D 2ndThor AD Relationship of epidermal growth factor receptor expression to ErbB2-signaling activity and prognosis in breast cancer patients J Clin Oncol 2005 23 1152 1160 15718311 10.1200/JCO.2005.09.055
Baselga J Tripathy D Mendelsohn J Baughman S Benz CC Dantis L Sklarin NT Seidman AD Hudis CA Moore J Phase II study of weekly intravenous recombinant humanized anti-p185HER2 monoclonal antibody in patients with HER2/neu-overexpressing metastatic breast cancer [comment] J Clin Oncol 1996 14 737 744 8622019
Cobleigh MA Vogel CL Tripathy D Robert NJ Scholl S Fehrenbacher L Wolter JM Paton V Shak S Lieberman G Slamon DJ Multinational study of the efficacy and safety of humanized anti-HER2 monoclonal antibody in women who have HER2-overexpressing metastatic breast cancer that has progressed after chemotherapy for metastatic disease J Clin Oncol 1999 17 2639 2648 10561337
Cardoso F Piccart MJ Durbecq V Di Leo A Resistance to trastuzumab: a necessary evil or a temporary challenge? Clin Breast Cancer 2002 3 247 257 12425752
Agus DB Akita RW Fox WD Lewis GD Higgins B Pisacane PI Lofgren JA Tindell C Evans DP Maiese K Targeting ligand-activated ErbB2 signaling inhibits breast and prostate tumor growth [comment] Cancer Cell 2002 2 127 137 12204533 10.1016/S1535-6108(02)00097-1
|
16168116
|
PMC1242139
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jul 6; 7(5):R708-R718
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1281
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12821616811810.1186/bcr1282Research ArticleHigh proportion of recurrent germline mutations in the BRCA1 gene in breast and ovarian cancer patients from the Prague area Pohlreich Petr [email protected] Michal [email protected] Jana [email protected] Zdenek [email protected] Marketa [email protected] Jaroslav [email protected] Jana [email protected] Jan [email protected] Lubos [email protected] Csilla [email protected] Bohuslav [email protected] Department of Biochemistry and Experimental Oncology, First Faculty of Medicine, Charles University, Prague, Czech Republic2 Institute of Biology and Medical Genetics, First Faculty of Medicine, Charles University, Prague, Czech Republic3 Department of Oncology, First Faculty of Medicine, Charles University, Prague, Czech Republic4 Unit of Genetic Epidemiology, International Agency for Research on Cancer, Lyon, France2005 19 7 2005 7 5 R728 R736 27 1 2005 11 3 2005 10 6 2005 22 6 2005 Copyright © 2005 Pohlreich et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Germline mutations in the BRCA1 and BRCA2 genes have been shown to account for the majority of hereditary breast and ovarian cancers. The purpose of our study was to estimate the incidence and spectrum of pathogenic mutations in BRCA1/2 genes in high-risk Czech families.
Methods
A total of 96 Czech families with recurrent breast and/or ovarian cancer and 55 patients considered to be at high-risk but with no reported family history of cancer were screened for mutations in the BRCA1/2 genes. The entire coding sequence of each gene was analyzed using a combination of the protein truncation test and direct DNA sequencing.
Results
A total of 35 mutations in the BRCA1/2 genes were identified in high-risk families (36.5%). Pathogenic mutations were found in 23.3% of breast cancer families and in 59.4% of families with the occurrence of both breast and ovarian cancer. In addition, four mutations were detected in 31 (12.9%) women with early onset breast cancer. One mutation was detected in seven (14.3%) patients affected with both a primary breast and ovarian cancer and another in three (33.3%) patients with a bilateral breast cancer. A total of 3 mutations in BRCA1 were identified among 14 (21.4%) women with a medullary breast carcinoma. Of 151 analyzed individuals, 35 (23.2%) carried a BRCA1 mutation and 9 (6.0%) a BRCA2 mutation. One novel truncating mutation was found in BRCA1 (c.1747A>T) and two in BRCA2 (c.3939delC and c.5763dupT). The 35 identified BRCA1 mutations comprised 13 different alterations. Three recurrent mutations accounted for 71.4% of unrelated individuals with detected gene alterations. The BRCA1 c.5266dupC (5382insC) was detected in 51.4% of mutation positive women. The mutations c.3700_3704del5 and c.181T>G (300T>G) contributed to 11.4% and 8.6% of pathogenic mutations, respectively. A total of eight different mutations were identified in BRCA2. The novel c.5763dupT mutation, which appeared in two unrelated families, was the only recurrent alteration of the BRCA2 gene identified in this study.
Conclusion
Mutational analysis of BRCA1/2 genes in 151 high-risk patients characterized the spectrum of gene alterations and demonstrated the dominant role of the BRCA1 c.5266dupC allele in hereditary breast and ovarian cancer.
==== Body
Introduction
Breast cancer (BC) is the most common malignancy affecting western women. About 5% to 10% of all BC cases are due to inheritance of a susceptibility allele, consistent with transmission in an autosomal dominant fashion, and a substantial proportion of these are due to germline mutations of the two major highly penetrant cancer susceptibility genes, BRCA1 (OMIM, 113705; GenBank, U14680.1) [1,2] and BRCA2 (OMIM, 600185; GenBank, U43746.1) [3-5]. Hereditary BC is characterized by an early age of onset, high incidence of bilateral disease and frequent association with ovarian cancer (OC). An increased incidence of other malignancies, such as colorectal, prostate and pancreatic cancer is also observed among BRCA1/2 mutation carriers [6-8]. The proportion of described mutations in BRCA1 relative to BRCA2 varies between populations. With the exception of a strong BRCA2 founder effect in Iceland [9], however, BRCA1 mutations are generally more frequently reported. In the majority (>80%) of families with BC and OC, the diseases are linked to the BRCA1 gene. Conversely, in the majority (>75%) of families with male and female BC, the disease is linked to BRCA2. Among families with female BC only, proportions of diseases due to mutations in BRCA1, BRCA2 and other genes are similar [10].
A large number of distinct mutations, polymorphisms and genetic variants of uncertain significance in the BRCA1 and BRCA2 genes is described in the Breast Cancer Information Core Database (BIC Database) [11]. The majority of mutations known to be disease causing result in a truncated protein due to frameshift, nonsense or splice site alterations. The spectrum of mutations varies between populations, with some showing a high frequency of unique mutations, for example in Italy [12,13], whereas a small number of founder mutations is more common in other ethnic groups. Notably, a single founder mutation in BRCA2 (c.771_775del5; commonly referred to as 999del5) accounts for the majority of hereditary cancer cases in Iceland [9], and three ancestral mutations (c.68_69delAG and c.5266dupC in BRCA1 and c.5946delT in BRCA2; 185delAG, 5382insC and 6174delT, respectively) were identified in the vast majority of families with a history of BC and OC in Ashkenazi Jews [14]. Population specific mutations have also been described in the Netherlands [15], Sweden [16], France [17], Spain [18] and other countries [19]. Two BRCA1 founder mutations, c.5266dupC and c.181T>G (300T>G), occur most frequently in countries of Central and Eastern Europe [20-25], including the Czech Republic [26].
The aim of this study was to estimate the incidence, spectrum and possible clustering of disease phenotypes associated with BRCA1 and BRCA2 mutations in the Prague area and Central Bohemia. The analysis was performed in families with a history of BC/OC and in high-risk patients not selected on the basis of their family history of cancer.
Materials and methods
Patients and families
Women with BC or OC considered to be at high risk of carrying a BRCA1 or BRCA2 mutation were selected for genetic testing between 1998 and 2003 at the Department of Oncology and at the Department of Gynecology and Obstetrics of the First Faculty of Medicine Charles University in Prague. The testing was performed immediately after confirming the pathologic diagnosis. All patients had Czech ancestries and were living in the Prague area and Central Bohemia. Patients were selected from cancer families that met the following criteria in first- or second-degree relatives: two cases of either BC diagnosed before the age of 50 or OC diagnosed at any age; and three or more cases of breast or ovarian cancer diagnosed at any age. A total of 60 families had a history positive for BC only (HBC families), 4 families had OC only (HOC families), and 32 families had both breast and ovarian cancer (HBOC families). Genetic material for analysis was obtained from the youngest affected individual from each family. Genetic testing was further offered to patients diagnosed with BC before the age of 36 (31 women) or with bilateral BC before the age of 51 (3 women) and patients with both primary breast and ovarian cancer (7 women) or medullary breast carcinoma diagnosed at any age (14 women), regardless of absence of reported family history of cancer (non-familial patients). All women in the study gave their informed consent prior to genetic testing. The protocol of investigation was approved by the Ethical Committee at the First Faculty of Medicine.
DNA and RNA isolation
Genomic DNA was isolated from EDTA blood samples using the Wizard genomic DNA purification kit (Promega, Madison, USA), according to the manufacturer's instructions. Total RNA was obtained from peripheral blood lymphocytes and reverse transcribed into cDNA as described [27].
Screening for BRCA1 and BRCA2 mutations
Most disease-associated mutations lead to premature termination of protein translation. Mutations are classified as deleterious if they truncate either the BRCA1 protein at least 10 amino acids from the C-terminus or the BRCA2 protein at least 110 amino acids from the C-terminus [28]. The nomenclature of mutations used is according to den Dunnen and Paalman [29], with nucleotides numbered from the A of the ATG translation initiation codon of GenBank reference sequences U14680.1 for BRCA1 cDNA and U43746.1 for BRCA2 cDNA. Original designations for BRCA1/2 mutations commonly referred to in the literature are included for ease of cross-referencing.
Mutational analysis was carried out by the protein truncation test (PTT) and direct DNA sequencing. The entire coding region of BRCA1 and BRCA2 was divided into overlapping fragments with sizes of 880 to 1569 bp and amplified by PCR. The PTT assay was used for pre-screening of amplified fragments; the final analysis of identified gene alterations was done by sequencing appropriate PCR fragments. Large exons (exon 11 of BRCA1 and exons 10 and 11 of BRCA2) were amplified on genomic DNA (exon 11 of BRCA1 and BRCA2 in three and four fragments, respectively; exon 10 of BRCA2 as a single fragment), whereas the remaining coding exons were amplified by two-step PCR (nested PCR) from cDNA (exons 2 to 10 and 12 to 24 of BRCA1 and exons 2 to 9 of BRCA2 as single fragments; exons 12 to 27 of BRCA2 in two fragments). Primer sequences used for amplification of BRCA1 and BRCA2 fragments were as described [16,30,31].
Amplifications were performed in 12.5 μl reaction mixtures containing PCR buffer (10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl2), 0.2 mM dNTPs, 0.4 μM of each primer, 0.5U of LA Taq DNA polymerase (Takara Shuzo Co., Shiga, Japan) and 30–50 ng of genomic DNA. Following initial denaturation (at 93°C for 2 minutes), 32 cycles (at 93°C for 1 minute, at 58°C for 1 minute, and at 72°C for 4minutes) and final extension (at 72°C for 5minutes) were performed. In the nested PCR procedure, a 2 μl aliquot of the reverse transcription reaction was used in the first round of amplification with external primers. A 1 μl aliquot of the first PCR reaction was removed and used as a template with internal primers in the second round of amplification. Reaction conditions and cycling parameters were as described above.
PTT-analysis was carried out by incubating PCR fragments (0.5 μl) in the TnT/T7 coupled transcription/translation system (Promega, Madison, USA) containing 0.5 to 1.0 μCi of L-[35S]methionine (Amersham Biosciences, Buckinghamshire, UK) for 90 minutes at 30°C in a total volume of 3 μl. Labeled protein products were analyzed on 12% SDS/polyacrylamide minigels (Bio-Rad Laboratories, Hercules, USA). The gels were fixed and prepared for fluorography by washing in Amplify (Amersham Biosciences, Buckinghamshire, UK). Dried gels were exposed for 24 to 48 h to X-ray film at -80°C.
DNA sequencing
PCR products that gave rise to truncated proteins by PTT analysis were gel purified and directly sequenced in forward and reverse directions using the BigDye 3.1 terminator cycle sequencing kit in a model 310 automated DNA sequencer (Applied Biosystems, Foster City, USA). Mutations detected by RNA-based analysis were confirmed by DNA sequencing. Each identified sequence alteration was confirmed by the analysis of a second blood sample.
Frequently occurring mutations at the beginning and at the end of BRCA1 (c.68_69delAG, c.181T>G and c.5266dupC) were identified by sequencing of RT-PCR fragments corresponding to exons 2 to 8 and 18 to 24. Sequence analyses of amplified genomic fragments containing the end of exon 11 in BRCA1 and exons 17 and 20 in BRCA2 were done as direct tests for other recurrent variants (c.4034delA in BRCA1 and c.7910_7914del5 and c.8537_8538delAG in BRCA2) known to occur in the Czech Republic and in populations of Central and Eastern Europe. In addition, in families with negative results from mutation analysis and with a strong history of cancer (families with three or more BC and/or OC cases), short exons of BRCA1 and BRCA2 were further screened for mutations by direct sequencing of RT-PCR fragments or by radioactive heteroduplex analysis following the protocol described by Gayther et al. [32] and Serova et al. [33].
Statistical analysis
Differences in mutation frequencies among groups of analyzed families were statistically evaluated by the Chi-square test.
Results
Mutation analysis was performed in 96 women from BC/OC families and in 55 non-familial patients. Analysis revealed 44 pathogenic cancer predisposing mutations, 6 of which have been previously reported elsewhere [27]. Within 151 analyzed individuals, 35 (23.2%) carried a BRCA1 mutation and 9 (6.0%) a BRCA2 mutation.
The BRCA1 mutations comprised 13 distinct alterations distributed widely across the coding sequence of the gene (Table 1). Twelve gene alterations caused a premature protein termination: eight were frame-shift alterations, with the majority of small deletions and insertions occurring in stretches of mononucleotide or dinucleotide repeats, and four were nonsense mutations. The c.181T>G mutation leading to a substitution of conserved cysteine 61 with glycine (p.Cys61Gly) in the RING finger domain of the BRCA1 protein was the only missense mutation identified in the gene.
Of the 10 BRCA1 mutations observed only once in our series, the c.1747A>T nonsense mutation is a novel gene alteration (not reported to the BIC by June 2004) found in family 397 with two cases of OC diagnosed at the ages of 39 and 43. Four additional mutations (c.1016delA, c.3331C>T, c.1127delA and c.2263G>T) belong to rare gene alterations (with one to four entries in the BIC database), whereas the others (c.68_69delAG, c.1687C>T, c.2411_2412delAG, c.3756_3759delGTCT and c.4165_4166delAG) occur frequently in various European regions, including Central Europe.
Three recurrent mutations were found in 25 (71.4%) of the 35 women with detected alterations in BRCA1. The mutation c.5266dupC (5382insC) was a highly prominent mutation detected in 18 patients, which accounted for 51.4% of all identified alterations in BRCA1. The mutation c.3700_3704del5 found in four families was the second most commonly identified alteration, which contributed to 11.4% of mutations detected in BRCA1. The mutation c.181T>G (300T>G; p.C61G) identified in three families comprised 8.6% of detected mutations.
The BRCA2 mutations included eight different gene abnormalities (Table 2). All alterations were localized to exon 11 and led to a truncated protein product: five were frameshift alterations and three were nonsense mutations. The mutation c.3939delC is a novel frameshift mutation, which results in a termination of translation at codon 1313. This mutation was detected in family 348 with two cases of OC diagnosed at the ages of 46 and 58. Both the c.3076A>T nonsense mutation (occurring in conjunction with the c.3075G>T missense mutation) and the frameshift mutation c.5238dupT belong to rare, infrequently reported gene alterations. Other identified mutations (c.2808_2811delACAA, c.3975_3978dupTGCT, c.5645C>A and c.5682C>G) occur frequently throughout Western Europe.
Recurrent mutations represented over two-thirds of all the BRCA1 mutations identified in this series. By contrast, the alterations in BRCA2 were mostly unique. The c.5763dupT mutation, which causes a premature termination of translation at codon 1923, appeared in two families and was the only recurrent alteration of the BRCA2 gene identified in this study. The mutation was first detected in family 67 with breast and ovarian cancer diagnosed before the age of 50. In the second unrelated family (F-327), only one affected woman who developed BC at the age of 32 was found. This gene alteration has since been reported once in Western Europe, as indicated in the BIC database.
Table 3 shows the prevalence of mutations in BC/OC families and in non-familial risk patients who had no reported family history of cancer. Pathogenic mutations were revealed in 35 (36.5%) of the 96 analyzed families. The incidence of mutations differed significantly between HBC (14/60; 23.3%) and HBOC (19/32; 59.4%) families (p = 0.0006). Two mutations were found in four HOC families.
Non-familial patients included a group of 31 women diagnosed with BC between ages 22 years and 35 years without history of BC or OC in their family. Pathogenic germline mutations in predisposing genes were detected in four (12.9%) women. Three mutations were identified in BRCA1 (c.68_69delAG, c.181T>G and c.5266dupC) and one in BRCA2 (c.5763dupT). Screening of seven patients with both breast and ovarian cancer and no family history identified one mutation (14.3%) in the BRCA1 gene (c.5266dupC). The mutation c.5266dupC was also detected in a group of three patients (33.3%) with bilateral BC.
A high incidence of medullary carcinoma has been reported among women with BRCA1-associated BC [34]. We performed the analysis of BRCA1 and BRCA2 genes in 14 women with this histological tumor subtype and found three truncating mutations (21.4%) in exon 11 of the BRCA1 gene. The c.3331C>T nonsense mutation was detected once, whereas the c.3700_3704del5 frameshift mutation was identified in two cases. No alteration in the BRCA2 gene was found among these analyzed patients.
Discussion
In our study, 35 mutations in the BRCA1 and BRCA2 genes were detected in 96 BC/OC families. In addition, we found four pathogenic mutations in patients with early onset BC, one mutation in a case of a bilateral BC, one mutation in a woman with both BC and OC, and three mutations in women with a medullary breast carcinoma. The majority of mutations identified in our study lead to protein truncations. Although short coding exons of both BRCA1 and BRCA2 were analyzed by either direct sequencing or heteroduplex analysis, only the BRCA1 c.181T>G (c.300T>G; p.C61G) missense mutation and BRCA2 c.3075G>T (p.K1025N) missense alteration (present in conjunction with the K1026X nonsense mutation) of unknown clinical significance were observed. One novel mutation was found in BRCA1 (c.1747A>T); two novel mutations were identified in BRCA2 (c.3939delC and c.5763dupT).
The prevalence of inherited BRCA1/2 mutations observed in different studies varies according to the ethnic origin of analyzed individuals, criteria of selection for genetic testing and techniques used for mutational screening. Although the techniques we applied in our study compromised our ability to detect missense mutations in regions of the BRCA1/2 genes screened only by PTT, at present the majority of reported missense alterations are difficult to interpret with respect to potential clinical significance, as the effect of the amino acid substitution on protein function is not yet well understood. Conversely, most protein truncating BRCA1/2 mutations are presumed to be pathogenic, thereby permitting women harboring such mutations to be provided appropriate clinical counseling and management. Our study may not have detected large genomic rearrangements encompassing regions outside the primer sets used for RT-PCR amplification. On the other hand, in the absence of rapid degradation of aberrant transcripts by nonsense-mediated mRNA decay [35], any rearrangements encompassing the exons within the regions amplified from cDNA would have been observed as aberrantly sized PCR products, as has also been found in other studies [36]. Further, the prevalence of genomic rearrangements varies between populations. To our knowledge, large deletions and rearrangements in BRCA1/2 genes have not been reported in countries of Central and Eastern Europe. In our series, analysis of truncated RT-PCR products and sequencing of corresponding genomic fragments identified one case with the rearrangement that involved exons 21 and 22 of the BRCA1 gene (Zikan, unpublished results). Despite the limitations of our approach, the prevalence of mutations in our group of high-risk families was comparable to that observed in Central Europe [21,23,26]. Further, the distribution of germline BRCA1/2 mutations in our high-risk families is consistent with a higher prevalence in the context of OC [10]; mutations were detected significantly more frequently in HBOC and HOC families than in HBC families (58.3% of 36 versus 23.3% of 60; p = 0.0006).
Interestingly, both BRCA1 and BRCA2 mutationswere more prevalent in families with OC (Table 3). Of the 27 BRCA1 mutations detected in high-risk families, 11 were present in HBC families (18.3%), whereas 16 were identified in HBOC and HOC families (44.4%). Of the eight mutations detected in BRCA2, three were present in HBC families (5%), relative to five in HBOC and HOC families (13.9%). This observation is in contrast to observations in larger series of examined families, where the risk of OC was significantly greater among women with BRCA1 mutations compared to women with BRCA2 mutations [10,37]. The higher prevalence of BRCA2 mutations among families with OC in our study may be due to the preponderance of mutations identified in the ovarian cancer cluster region, and lend support to the increased risk of OC suggested to be conferred by mutations within this region relative to other BRCA2 mutations [38,39].
Gayther et al. [40] have reported that mutations in the 3' third of the BRCA1 gene are associated with a lower proportion of ovarian cancer. The border for this phenotype correlation was located at exon 13, between codons 1435 and 1443. Further studies provided proof for this genotype-phenotype correlation [16,41], although other authors failed to replicate this observation for BRCA1 mutations [17,20]. Despite the higher occurrence of ovarian cancer in high-risk families with mutations located in exons 2 to 12, the relative frequency (12/37 cancer cases; 32.4%) did not differ significantly in our study from the frequency of ovarian cancer (11/51 cancer cases; 21.6%) in patients with mutations in exons 14 to 24 (p = 0.25).
The surprising finding of our investigation was a high predominance of recurrent mutations in the BRCA1 gene, which contributed to a substantial proportion of hereditary BC and OC cases. The three repeatedly occurring mutations in BRCA1 were detected in more than 56% of women with identified alterations in BRCA1/2 genes and in more than 71% of women with alterations in BRCA1. The most frequent mutation was c.5266dupC (5382insC) found in 15 (15.6%) of 96 analyzed BC/OC families and in 18 (40.9%) of 44 BRCA1/2 mutation positive patients. The occurrence of this mutation is comparable to that found in Polish [22,23,25,42] and Russian [20] populations, but significantly higher than that described in Germany [43] and Austria [21]. The c.5266dupC mutation is the most prevalent BRCA1 alteration in Europe and a geographic distribution of this mutation is consistent with its Baltic origin [19]. The c.5266dupC allele also occurs at a high frequency (0.11%) in the Ashkenazi Jewish population [14], although the 18 Czech patients carrying this gene defect did not report an Ashkenazi Jewish heritage. The other two recurrent mutations detected in our group of patients, c.3700_3704del5 and c.181T>G (300T>G), also belong to gene alterations that have been repeatedly detected in the Central European population [23-26].
The BRCA1 c.5266dupC and c.181T>G mutations are prevalent in Poland and in the Czech Republic, although spectra of mutations display significant differences in these countries [26,42]. The mutation c.4034delA (4153delA), contributing to 9.8% of mutations identified in BRCA1 in Poland [42], did not occur either in our or in Moravian families [26]. We identified 10 unique mutations in BRCA1, which suggests that the spectrum of alterations in this gene is more heterogeneous than that reported in Poland [42].
In contrast to the BRCA1 gene, there were few recurrent mutations within BRCA2. With the exception of the c.5763dupT mutation detected in two unrelated individuals, each alteration identified in BRCA2 was found in only one family.
In a set of pathogenic mutations in BRCA1 identified in Brno (Moravia) [26], the prevalence of the three most common mutations (c.5266dupC, c.3700_3704del5 and c.181T>G) was 37.3% (22/59), 13.6% (8/59) and 10.2% (6/59), respectively, whereas in our group of patients, the prevalence of these alterations was 51.4% (18/35), 11.4% (4/35) and 8.6% (3/35). Small variations in the mutation spectra observed in both studies may be caused by limited sample size and may also reflect differences in the groups of patients selected for genetic testing.
The occurrence of BRCA2 mutations was higher in Moravia (33%; 29/88 of mutation positive patients) than in our study (20.5%; 9/44 of mutation positive patients) and the spectrum of genetic alterations was completely different [26]. The c.7910_7914del5 (8138_8142del5) and c.8537_8538delAG (8765_8766deAG) alterations found in Brno in 7 (24.1%) of the 29 families with detected mutations in BRCA2, were not present in our group of patients. On the contrary, the mutation c.5763dupT, the only recurrent alteration of the BRCA2 gene identified in our study, was not found in Moravia. A high frequency of unique BRCA2 mutations may be characteristic of the examined Prague area and Central Bohemia, although the examination of a larger group of families is required to obtain valid results.
The incidence of BRCA1/2 mutations in a group of Czech women with early onset non-familial BC was 12.9% (4/31; Table 3), whichsuggests that the age at diagnosis in patients with a negative family history is an important indicator for the presence of a pathogenic mutation and lends support to the screening of BRCA1/2 genes in patients with early onset disease. In contrast to our findings, only 2% of non-familial patients had pathogenic germline BRCA1 and BRCA2 mutations in a group of patients in Great Britain who were diagnosed with BC at the age of 30 years or younger [44]. A similarly low frequency of mutations was found in non-familial patients in Spain and Iran [45,46]. In our study, all patients carried the most common, easily detectable mutations of BRCA1 or BRCA2 that prevail in this region and can be associated with early onset cancer. A larger set of patients with early onset BC is currently under investigation to determine the incidence of mutations in this risk group.
Medullary carcinoma of the breast is not common (2% to 3%) in patients with BRCA2 mutations and those with no known germline gene alteration [47]. In BRCA1-related BC, the incidence of typical medullary carcinoma was 19% (6/32) in a French study [34], 8% (4/49) in a Dutch study [48], but 0% in a Swedish series of 40 BRCA1-associated tumors [49]. In our study, one medullary carcinoma was identified in a group of 22 patients (4.5%) with BRCA1-related BC (F-252), whereas no tumor of this histological type was found in seven women with BRCA2-associated tumors. We did, however, find three germline BRCA1 mutations in a set of 14 patients (21.4%) with medullary breast carcinoma selected for examination regardless of the family history. These women did not belong to high-risk families and did not fulfill common criteria for genetic testing. One patient (Table 1; F-80) with a medullary carcinoma at age 50 had a sister affected with BC at age 54 and the other two (Table 1; F-305 and F-390) with medullary carcinomas at ages 38 and 42, respectively, were without a family history of breast or ovarian cancer. Our results are in agreement with studies of Eisinger et al. [34] who tested 18 cases of medullary carcinoma for mutations in BRCA1 gene and found two (11%) harboring BRCA1 mutations. Interestingly, these cases did not belong to high-risk families either. Indication of cases with medullary carcinoma for BRCA1 testing, regardless of the family history, may be helpful in mutation screening. Examination of a larger group of patients with this histological type of carcinoma is required to determine the importance of this morphological parameter more exactly.
Conclusion
Mutational analysis of BRCA1/2 genes in 151 high-risk patients characterized the spectrum of gene alterations and demonstrated the dominant role of the BRCA1 c.5266dupC (5382insC) allele in hereditary breast and ovarian cancer. The pre-screening of high-risk patients for the presence of this allele, which accounted for more than 40% of all identified gene alterations, and pre-screening for the presence of other pathogenic, repeatedly occurring alleles (c.3700_3704del5 and c.181T>G), which contributed to about 16% of gene defects, could enable rapid detection of a high percentage of patients with hereditary cancer predisposition and, thus, reduce the cost of mutation analysis in the Czech population.
Abbreviations
BC = breast cancer; BIC = Breast Cancer Information Core; HBC = hereditary breast cancer; HBOC = hereditary breast and ovarian cancer; HOC = hereditary ovarian cancer; OC = ovarian cancer; PCR = polymerase chain reaction; PTT = protein truncation test.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
PP and MZ contributed equally to this work.
Acknowledgements
The work at this project was supported by the Internal Grant Agency of the Ministry of Health of the Czech Republic, Grant No. NC 7527-3; Grant Agency of the Czech Republic, Grant No. 310/01/1537, and Research Project of the Ministry of Education MSM 0021620808. MZ is recipient of 'lauréat du premier prix de médecine, 2001', organized by the French Embassy and French Institute in Prague and Laboratoires Fournier in France.
Figures and Tables
Table 1 Pathogenic germline BRCA1 mutations in breast and ovarian cancer patients from the Prague area
Family no. Mutation description Method of detection No. of cancers in a family and age at onset
Exon Traditional nomenclature Approved nomenclature Predicted effect Breast cancer (bilateral) Mean age at diagnosis Other cancers (age at onset)
F-24 2 c.187_188delAG c.68_69delAG p.Glu23fsX39 Sequencing 1 35 Colon (54)
F-111 5 c.300T>G c.181T>G p.Cys61Gly Sequencing 3 42 Stomach (51)
F-126 5 c.300T>G c.181T>G p.Cys61Gly Sequencing 2 45 -
F-252 5 c.300T>G c.181T>G p.Cys61Gly Sequencing 1 29 -
F-43 11 c.1135delA c.1016delA p.Lys339fsX340 PTT 2 41 Colon (50), lung (64)
F-361 11 c.1246delA c.1127delA p.Asn376fsX393 PTT 1 37 Ovarian (52, 54, 55)
F-21 11 c.1806C>T c.1687C>T p.Gln563X PTT 1 (1) 46 Ovarian (43), melanoma (53)
F-397 11 c.1866A>Ta c.1747A>Ta p.Lys583Xa PTT - - Ovarian (39, 43)
F-249 11 c.2382G>T c.2263G>T p.Glu755X PTT 4 53 Ovarian (41, 54)
F-61 11 c.2530_2531delAG c.2411_2412delAG p.Gln804fsX808 PTT 3 49 -
F-80 11 c.3450C>T c.3331C>T p.Gln1111X PTT 2 52 -
F-305 11 c.3819_3823del5 c.3700_3704del5 p.Val1234fsX1241 PTT 1 38 Leukemia (67), lung (65)
F-337 11 c.3819_3823del5 c.3700_3704del5 p.Val1234fsX1241 PTT 3 44 -
F-347 11 c.3819_3823del5 c.3700_3704del5 p.Val1234fsX1241 PTT 2 42 -
F-390 11 c.3819_3823del5 c.3700_3704del5 p.Val1234fsX1241 PTT 1 42 Lung (56), kidney (65)
F-164 11 c.3875_3878delGTCT c.3756_3759delGTCT p.Leu1252fsX1262 PTT 2 (1) 42 Ovarian (40, 43), stomach (?)
F-245 12 c.4284_4285delAG c.4165_4166delAG p.Ser1389fsX HDA 2 38 Ovarian (44, 50), kidney (75)
F-15 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 1 32 -
F-75 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 1 44 Ovarian (?, ?)
F-152 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 2 58 Ovarian (72)
F-185 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 3 51 Colon (?)
F-187 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 3 48 Ovarian (56)
F-194 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 4 43 Ovarian (52)
F-201 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 2 41 -
F-239 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 3 46 Ovarian (41), stomach (?)
F-243 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 3 (1) 42 -
F-261 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 2 29 -
F-265 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 3 (1) 52 Ovarian (?)
F-273 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 3 36 Ovarian (42), colon (51, 56)
F-331 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 1 (1) 31 Uterus (60), colon (64)
F-342 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 2 50 Ovarian (41), colon (?)
F-368 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 4 (2) 37 Ovarian (40), kidney 78)
F-370 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 2 49 Ovarian (55)
F-385 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 1 52 Ovarian (61), colon (83), melanoma (79)
F-387 20 c.5385dupC c.5266dupC p.Gln1756fsX1829 Sequencing 3 (1) 36 -
Position in cDNA is according to GenBank accession number U14680. aNovel mutations. HAD, heteroduplex analysis; PTT, protein truncation test.
Table 2 Pathogenic germline BRCA2 mutations in breast and ovarian cancer patients from the Prague area
Family no. Mutation description Method of detection No. of cancers in a family and age at onset
Exon Traditional nomenclature Approved nomenclature Predicted effect Breast cancer (bilateral) Mean age at diagnosis Other cancers (age at onset)
F-263 11 c.3036_3039delACAA c.2808_2811delACAA p.Lys936fsX958 PTT 3 (1) 41 -
F-279 11 c.3303G>T; 3304A>T c.3075G>T; 3076A>T p.Lys1025Asn; p.Lys1026X PTT 1 50 Ovarian (47, 50)
F-348 11 c.4167delCa c.3939delCa p.Tyr1313fsX PTT - - Ovarian (46, 58), lung (?)
F-237 11 c.4203_4206dupTGCT c.3975_3978dupTGCT p.Ala1327fsX1331 PTT 2 38 Stomach (?)
F-298 11 c.5466dupT c.5238dupT pAsn1747fsX PTT 1 41 Ovarian (59)
F-304 11 c.5873C>A c.5645C>A p.Ser1882X PTT 3 61 Ovarian (49), colon (67)
F-49 11 c.5910C>G c.5682C>G p.Tyr1894X PTT 2 (1) 45 Lung (?)
F-67 11 c. 5991dupTa c.5763dupTa p.Ala1922fsX1923a PTT 1 48 Ovarian (49), esophagus (?)
F-327 11 c. 5991dupTa c.5763dupTa p.Ala1922fsX1923a PTT 1 32 -
Position in cDNA is according to GenBank accession number U43746. aNovel mutations. PTT = protein truncation test.
Table 3 Frequencies of BRCA1 and BRCA2 mutations in relation to the classification of patients and families
Classification Number of patients/families No. of mutations in BRCA1 (%) No. of mutations in BRCA2 (%) No. of BRCA1 + BRCA2 mutations (%)
Cases without a family history of breast and ovarian cancer
Breast cancer before 36 31 3 (9.7) 1 (3.2) 4 (12.9)
Bilateral breast cancer before 51 3 1 (33.3) - 1 (33.3)
Breast and ovarian cancer 7 1 (14.3) - 1 (14.3)
Medullary breast carcinoma 14 3 (21.4) - 3 (21.4)
Breast cancer families (HBC)
2 breast cancer cases 25 5 (20.0) 2 (8.0) 7 (28.0)
≥3 breast cancer cases 35 6 (17.1) 1 (2.9) 7 (20.0)
Total 60 11 (18.3) 3 (5.0) 14 (23.3)
Breast and ovarian cancer families (HBOC)
1 breast cancer and 1 ovarian cancer 5 1 (20.0) 2 (40.0) 3 (60.0)
≥3 breast and ovarian cancer cases 27 14 (51.9) 2 (7.4) 16 (59.3)
Total 32 15 (46.9) 4 (12.5) 19 (59.4)
Ovarian cancer families (HOC)
≥2 ovarian cancer cases 4 1 (25.0) 1 (25.0) 2 (50.0)
HBC, hereditary breast cancer; HBOC, hereditary breast and ovarian cancer; HOC, hereditary ovarian cancer.
==== Refs
Hall JM Lee MK Newman B Morrow JE Anderson LA Huey B King MC Linkage of early-onset familial breast cancer to chromosome 17q21 Science 1990 250 1684 1690 2270482
Miki Y Swensen J Shattuck-Eidens D Futreal PA Harshman K Tavtigian S Liu Q Cochran C Bennett LM Ding W A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1 Science 1994 266 66 71 7545954
Wooster R Neuhausen SL Mangion J Quirk Y Ford D Collins N Nguyen K Seal S Tran T Averill D Localization of a breast cancer susceptibility gene, BRCA2, to chromosome 13q12-13 Science 1994 265 2088 2090 8091231
Wooster R Bignell G Lancaster J Swift S Seal S Mangion J Collins N Gregory S Gumbs C Micklem G Identification of the breast cancer susceptibility gene BRCA2 Nature 1995 378 789 792 8524414 10.1038/378789a0
Tavtigian SV Simard J Rommens J Couch F Shattuck-Eidens D Neuhausen S Merajver S Thorlacius S Offit K Stoppa-Lyonnet D The complete BRCA2 gene and mutations in chromosome 13q-linked kindreds Nat Genet 1996 12 333 337 8589730 10.1038/ng0396-333
Greene MH Genetics of breast cancer Mayo Clin Proc 1997 72 54 65 9005288
Couch FJ Weber BL Vogelstein B, Kinzler KW Breast cancer The genetic basis of human cancer 1998 1 New York: McGraw-Hill 537 563
Thompson D Easton DF Breast Cancer Linkage Consortium Cancer incidence in BRCA1 mutation carriers J Natl Cancer Inst 2002 94 1358 1365 12237281
Thorlacius S Olafsdottir G Tryggvadottir L Neuhausen S Jonasson JG Tavtigian SV Tulinius H Ogmundsdottir HM Eyfjord JE A single BRCA2 mutation in male and female breast cancer families from Iceland with varied cancer phenotypes Nat Genet 1996 13 117 119 8673089 10.1038/ng0596-117
Ford D Easton DF Stratton M Narod S Goldgar D Devilee P Bishop DT Weber B Lenoir G Chang-Claude J Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families Am J Hum Genet 1998 62 676 689 9497246 10.1086/301749
Breast Cancer Information Core (BIC)
Santarosa M Viel A Dolcetti R Crivellari D Magri MD Pizzichetta MA Tibiletti MG Gallo A Tumolo S Del Tin L Boiocchi M Low incidence of BRCA1 mutations among Italian families with breast and ovarian cancer Int J Cancer 1998 78 581 586 9808526 10.1002/(SICI)1097-0215(19981123)78:5<581::AID-IJC9>3.0.CO;2-G
Santarosa M Dolcetti R Magri MD Crivellari D Tibiletti MG Gallo A Tumolo S Della Puppa L Furlan D Boiocchi M Viel A BRCA1 and BRCA2 genes: Role in hereditary breast and ovarian cancer in Italy Int J Cancer 1999 83 5 9 10449599 10.1002/(SICI)1097-0215(19990924)83:1<5::AID-IJC2>3.0.CO;2-U
Roa BB Boyd AA Volcik K Richards CS Ashkenazi Jewish population frequencies for common mutations in BRCA1 and BRCA2 Nat Genet 1996 14 185 187 8841191 10.1038/ng1096-185
Peelen T van Vliet M Petrij-Bosch A Mieremet R Szabo C van den Ouweland AMW Hogervorst F Brohet R Ligtenberg MJL Teugels E A high proportion of novel mutations in BRCA1 with strong founder effects among Dutch and Belgian hereditary breast and ovarian cancer families Am J Hum Genet 1997 60 1041 1049 9150151
Hakansson S Johannsson O Johansson U Sellberg G Loman N Gerdes A-M Holmberg E Dahl N Pandis N Kristoffersson U Moderate frequency of BRCA1 and BRCA2 germ-line mutations in Scandinavian familial breast cancer Am J Hum Genet 1997 60 1068 1078 9150154
Stoppa-Lyonnet D Laurent-Puig P Essioux L Pages S Ithier G Ligot L Fourquet A Salmon RJ Clough KB Pouillart P BRCA1 sequence variations in 160 individuals referred to a breast/ovarian family cancer clinic Am J Hum Genet 1977 60 1021 1030
Diez O Osorio A Duran M Martinez-Ferrandis JI de la Hoya M Salazar R Vega A Campos B Rodriguez-Lopez R Velasco E Analysis of BRCA1 and BRCA2 genes in Spanish breast/ovarian cancer patients: a high proportion of mutations unique to Spain and evidence of founder effects Hum Mutat 2003 22 301 312 12955716 10.1002/humu.10260
Szabo CI King MC Population genetics of BRCA1 and BRCA2 Am J Hum Genet 1997 60 1013 1020 9150148
Gayther SA Harrington P Russell P Kharkevich G Garkavtseva RF Ponder BA Frequently occurring germline mutations of the BRCA1 gene in ovarian cancer families from Russia Am J Hum Genet 1997 60 1239 1242 9150173
Wagner TMU Moslinger RA Muhr D Langbauer G Hirtenlehner K Concin H Doeller W Haid A Lang AH Mayer P BRCA1-related breast cancer in Austrian breast and ovarian cancer families: Specific BRCA1 mutations and pathological characteristics Int J Cancer 1998 77 354 360 9663595 10.1002/(SICI)1097-0215(19980729)77:3<354::AID-IJC8>3.0.CO;2-N
Grzybowska E Zientek H Jasinska A Rusin M Kozlowski P Sobczak K Sikorska A Kwiatkowska E Gorniak L Kalinowska E High frequency of recurrent mutations in BRCA1 and BRCA2 genes in Polish families with breast and ovarian cancer Hum Mutat 2000 16 482 490 11102977 10.1002/1098-1004(200012)16:6<482::AID-HUMU5>3.0.CO;2-O
Gorski B Byrski T Huzarski T Jakubowska A Menkiszak J Gronwald J Pluzanska A Bebenek M Fischer-Maliszewska L Grzybowska E Founder mutations in the BRCA1 gene in Polish families with breast-ovarian cancer Am J Hum Genet 2000 66 1963 1968 10788334 10.1086/302922
van der Looij M Szabo C Besznyak I Liszka G Csokay B Pulay T Toth J Devilee P King MC Olah E Prevalence of founder BRCA1 and BRCA2 mutations among breast and ovarian cancer patients in Hungary Int J Cancer 2000 86 737 740 10797299 10.1002/(SICI)1097-0215(20000601)86:5<737::AID-IJC21>3.0.CO;2-1
Perkowska M Brozek I Wysocka B Haraldsson K Sandberg T Jahansson U Sellberg G Borg A Limon J BRCA1 and BRCA2 mutation analysis in breast-ovarian cancer families from Northeastern Poland Hum Mutat 2003 21 553 554 12673801 10.1002/humu.9139
Foretova L Machackova E Navratilova M Pavlu H Hruba M Lukesova M Valik D BRCA1 and BRCA2 mutations in women with familial or early-onset breast/ovarian cancer in the Czech Republic Hum Mutat 2004 23 397 398 15024741 10.1002/humu.9226
Pohlreich P Stribrna J Kleibl Z Zikan M Kalbacova R Petruzelka L Konopasek B Mutations of the BRCA1 gene in hereditary breast and ovarian cancer in the Czech Republic Med Princ Pract 2003 12 23 29 12566964 10.1159/000068163
Frank TS Deffenbaugh AM Reid JE Hulick M Ward BE Lingenfelter B Gumpper KL Scholl T Tavtigian SV Pruss DR Critchfield GC Clinical characteristics of individuals with germline mutations in BRCA1 and BRCA2: Analysis of 10,000 individuals J Clin Oncol 2002 20 1480 1490 11896095 10.1200/JCO.20.6.1480
den Dunnen JT Paalman MH Standardizing mutation nomenclature: Why bother? Hum Mutat 2003 22 181 182 12938082 10.1002/humu.10262
Hogervorst FBL Cornelis RS Bout M van Vliet M Oosterwijk JC Olmer R Bakker B Klijn JGM Vasen HFA Meijers-Heijboer H Rapid detection of BRCA1 mutations by the protein truncation test Nat Genet 1995 10 208 212 7663517
Krainer M Silva-Arrieta S Fitzgerald MG Shimada A Ishioka C Kanamaru R MacDonald DJ Unsal H Finkelstein DM Bowcock A Differential contributions of BRCA1 and BRCA2 to early-onset breast cancer N Engl J Med 1997 336 1416 1421 9145678 10.1056/NEJM199705153362003
Gayther SA Harrington P Russell P Kharkevich G Garkavtseva RF Ponder BA UKCCCR Familial Ovarian Cancer Study Group Rapid detection of regionally clustered germ-line BRCA1 mutations by multiplex heteroduplex analysis Am J Hum Genet 1996 58 451 456 8644703
Serova OM Mazoyer S Puget N Dubois V Tonin P Shugart YY Goldgar D Narod SA Lynch HT Lenoir GM Mutations in BRCA1 and BRCA2 in breast cancer families: Are there more breast cancer susceptibility genes? Am J Hum Genet 1997 60 486 495 9042907
Eisinger F Jacquemier J Charpin C Stoppa-Lyonnet D Bressac-de Paillerets B Peyrat J-P Longy M Guinebretiere J-M Sauvan R Noguchi T Mutations at BRCA1: The medullary breast carcinoma revisited Cancer Res 1998 58 1588 1592 9563465
Maquat LE When cells stop making sense: effects of nonsense codons on RNA metabolism in vertebrate cells RNA 1995 1 453 465 7489507
Puget N Torchard D Serova-Sinilnikova OM Lynch HT Feunteun J Lenoir GM Mazoyer S A 1-kb Alu-mediated germ-line deletion removing BRCA1 exon 17 Cancer Res 1997 57 828 831 9041180
King MC Marks JH Mandell JB The New York Breast Cancer Study Group Breast and ovarian cancer risks due to inherited mutations in BRCA1 and BRCA2 Science 2003 302 643 646 14576434 10.1126/science.1088759
Gayther SA Mangion J Russell P Seal S Barfoot R Ponder BA Stratton MR Easton D Variation of risks of breast and ovarian cancer associated with different germline mutations of the BRCA2 gene Nat Genet 1997 15 103 105 8988179 10.1038/ng0197-103
Thompson D Easton D Breast Cancer Linkage Consortium Variation in cancer risks, by mutation position, in BRCA2 mutation carriers Am J Hum Genet 2001 68 410 419 11170890 10.1086/318181
Gayther SA Warren W Mazoyer S Russel PA Harrington PA Chiano M Seal S Hamoudi R van Rensburg EJ Dunning AM Germline mutations of the BRCA1 gene in breast and ovarian cancer families provide evidence for a genotype-phenotype correlation Nature Genet 1995 11 428 433 7493024 10.1038/ng1295-428
Ramus SJ Kote-Jarai Z Friedman LS van der Looij M Gayther SA Csokay B Ponder BAJ Olah E Analysis of BRCA1 and BRCA2 mutations in Hungarian families with breast or breast-ovarian cancer Am J Hum Genet 1997 60 1242 1246 9150174
Gorski B Jakubowska A Huzarski T Byrski T Gronwald J Grzybowska E Mackiewicz A Stawicka M Bebenek M Sorokin D A high proportion of founder BRCA1 mutations in Polish breast cancer families Int J Cancer 2004 110 683 686 15146557 10.1002/ijc.20162
Backe J Hofferbert S Skawran B Dork T Stuhrmann M Karstens JH Untch M Meindl A Burgemeister R Chang-Claude J Weber BHF Frequency of BRCA1 mutation 5382insC in German breast cancer patients Gynecol Oncol 1999 72 402 406 10053113 10.1006/gyno.1998.5270
Lalloo F Varley J Ellis D Moran A O'Dair L Pharoah P Evans DGR The Early Onset Breast Cancer Study Group Prediction of pathogenic mutations in patients with early-onset breast cancer by family history Lancet 2003 361 1101 1102 12672316 10.1016/S0140-6736(03)12856-5
Martinez-Ferrandis JI Vega A Chirivella I Marin-Garcia P Insa A Lluch A Carracedo A Chaves FJ Garcia-Conde J Cervantes A Armengod M-E Mutational analysis of BRCA1 and BRCA2 in mediterranean Spanish women with early-onset breast cancer: Identification of three novel pathogenic mutations Hum Mutat 2003 22 417 418 14517958 10.1002/humu.9188
Yassaee VR Zeinali S Harirchi I Jarvandi S Mohagheghi MA Hornby DP Dalton A Novel mutations in the BRCA1 and BRCA2 genes in Iranian women with early-onset breast cancer Breast Cancer Res 2002 4 R6 12100744 10.1186/bcr443
Shousha S Medullary carcinoma of the breast and BRCA1 mutation Histopathology 2000 37 182 185 10931243 10.1046/j.1365-2559.2000.00880.x
Verhoog LC Brekelmans CT Seynaeve C van den Bosch LM Dahmen G van Geel AN Tilanus-Linthorst MM Bartels CC Wagner A van den Ouweland A Survival and tumour characteristics of breast-cancer patients with germline mutations of BRCA1 Lancet 1998 351 316 321 9652611 10.1016/S0140-6736(97)07065-7
Johannsson OT Ranstam J Borg A Olsson H Survival of BRCA1 breast and ovarian cancer patients: a population-based study from southern Sweden J Clin Oncol 1998 16 397 404 9469321
|
16168118
|
PMC1242140
|
CC BY
|
2021-01-04 16:54:51
|
no
|
Breast Cancer Res. 2005 Jul 19; 7(5):R728-R736
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1282
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12831616812010.1186/bcr1283Research ArticleGenotypes and haplotypes of the methyl-CpG-binding domain 2 modify breast cancer risk dependent upon menopausal status Zhu Yong [email protected] Heather N [email protected] Yawei [email protected] Theodore R [email protected] Tongzhang [email protected] Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut, USA2005 19 7 2005 7 5 R745 R752 9 9 2004 10 11 2004 7 6 2005 24 6 2005 Copyright © 2005 Zhu et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
MBD2, the gene encoding methyl-CpG-binding domain (MBD)2, is a major methylation related gene and functions as a transcriptional repressor that can specifically bind to the methylated regions of other genes. MBD2 may also mediate gene activation because of its potential DNA demethylase activity. The present case-control study investigated associations between two single nucleotide polymorphisms (SNPs) in the MBD2 gene and breast cancer risk.
Methods
DNA samples from 393 Caucasian patients with breast cancer (cases) and 436 matched control individuals, collected in a recently completed breast cancer case–control study conducted in Connecticut, were included in the study. Because no coding SNPs were found in the MBD2 gene, one SNP in the noncoding exon (rs1259938) and another in the intron 3 (rs609791) were genotyped. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated to estimate cancer risk associated with the variant genotypes and the reconstructed haplotypes.
Results
The variant genotypes at both SNP loci were significantly associated with reduced risk among premenopausal women (OR = 0.41 for rs1259938; OR = 0.54 for rs609791). Further haplotype analyses showed that the two rare haplotypes (A-C and A-G) were significantly associated with reduced breast cancer risk (OR = 0.40, 95% CI = 0.20–0.83 for A-C; OR = 0.47, 95% CI = 0.26–0.84 for A-G) in premenopausal women. No significant associations were detected in the postmenopausal women and the whole population.
Conclusion
Our results demonstrate a role for the MBD2 gene in breast carcinogenesis in premenopausal women. These findings suggest that genetic variations in methylation related genes may potentially serve as a biomarker in risk estimates for breast cancer.
==== Body
Introduction
It has recently been recognized that cancer is a manifestation of both abnormal genetic and epigenetic events [1]. Dysregulated epigenetic controls, which usually are represented by abnormal DNA methylation patterns such as global hypomethylation and region specific hypermethylation, are a hallmark of most cancers. Although the precise mechanisms underlying methylation alterations are far from being fully understood, the overall methylation process is mainly regulated by several groups of regulatory proteins [2-4].
The methyl-CpG-binding domain (MBD) proteins are among these protein families that bind specifically to a methylated gene and mediate transcriptional repression via effects on chromatin structure. Thus far, five MBD genes have been identified in mammalian cells that encode putative MBDs, namely MeCP2, MBD1, MBD2, MBD3, and MBD4 [5-7]. Human MBD genes are considered housekeeping genes because they are widely expressed in somatic tissues [6]. Given the epigenetic role of MBD proteins in regulating gene expression, MBDs may be involved in cancer development by affecting the expression of cancer related genes. In fact, there is growing evidence that aberrant expression of MBD proteins is associated with human cancers [8,9].
The MBD2 gene is mapped to the conserved region within human chromosome 18q21 [10]. Genomic sequence analysis determined that the MBD2 gene contains six exons and one noncoding exon spanning more than 50 kb in the genome. The MBD2 gene encodes two potential forms of protein MBD2 that correspond to the initiation of translation starting at either the first (MBD2a; 43.5 kDa) or second (MBD2b; 29.1 kDa) methionine codons. The signal functions of MBD2 are to bind specifically to methylated gene promoters and recruit histone deacetylases and chromatin remodeling proteins. The altered chromatin structure resulting from the binding of these factors may be resistant to the transcriptional machinery and, as a result, repress gene expression [11,12]. A recent finding also suggests that MBD2 has potential DNA demethylase activity [13], implying that it might mediate gene activation in addition to transcriptional repression. However, two subsequent studies could not demonstrate any demethylase activity of MBD2 [14,15], and this inconsistency in the functions of MBD2 remains to be resolved.
Although our understanding of the exact function of MBD2 in epigenetics is still in its early stages, several studies in human cancer research have demonstrated that the MBD2 protein plays a role in tumorigenesis. For example, a recent study [16] showed that breast carcinomas exhibit alterations in MBD2 expression. One interesting finding from that study was that breast carcinomas can be divided into two groups, with one expressing very high levels of MBD2 and the other expressing a much lower level. MBD2 has also been reported to be involved in the repression of GSTP1 transcription in breast cancer cells [17]. Moreover, a significant reduction in MBD2 mRNA expression was found in human colorectal and gastric cancerous tissues [18] and peripheral blood lymphocytes [19] in bladder cancer patients, implying a protective role for MBD2 in tumorigenesis. MBD2 protein expression and its demethylase activity were detected in normal human prostate tissue but not in cancerous tissue [20]. These differences between types of cancers in the abundance of MBD2 levels may reflect different roles for MBD2 either in transcriptional repression or in the demethylation process.
Given that there is a potential role for MBD2 in tumorigenesis, we hypothesized that genetic polymorphisms in the MBD2 gene may modify an individual's susceptibility to human cancers. In this molecular epidemiologic study, we genotyped two single nucleotide polymorphisms (SNPs; rs1259938 and rs609791) in the MBD2 gene to investigate whether genetic variations in the MBD2 gene are associated with breast cancer risk and whether the potential associations are modified by menopausal status.
Materials and methods
Study population
This study was built upon a recently completed breast cancer case–control study that was undertaken in Connecticut, USA. Detailed information regarding the study population is provided elsewhere [21]. Briefly, a total of 475 histologically confirmed incident breast cancer cases (ICD-O, 174.0–174.9) and 502 randomly selected control individuals were identified from the Tolland and New Haven County area of Connecticut between 1 January 1994 and 31 December 1997. All of the cases and controls were in the age range 30–80 years and had no previous diagnosis of cancer with the exception of nonmelanoma skin cancer.
For New Haven County, eligible cases were identified from the major hospital of the county (Yale–New Haven Hospital) through the computer database system at the Department of Surgical Pathology. Controls were also randomly selected from the computer database system from among women who were histologically confirmed to be without breast cancer. The participation rates were 77% for cases and 71% for controls in New Haven County. For Tolland County, because there was no major county hospital in this county, newly diagnosed breast cancer cases were identified from area hospital records by the Rapid Case Ascertainment system at the Yale Comprehensive Cancer Center. Controls from Tolland Country were recruited through random digit dialing methods for those under age 65 years and randomly selected from Health Care Financing Administration files for those aged 65 years and over. The participation rates were 74% for cases and 64% for controls in Tolland County.
The study pathologist reviewed all of the pathologic diagnoses for breast cancer patients and benign breast disease controls. Breast carcinoma were classified as carcinoma in situ, invasive ductal, or lobular carcinoma, and were staged according to the American Joint Committee on Cancer staging system [22].
Data collection
Informed consent was obtained from all study participants before collection of epidemiologic data through personal interview. The 45-min in-person interview, completed by all study participants, was administered by trained interviewers following institutional guidelines for human subjects. Data on smoking habits, alcohol consumption, and hormone replacement therapy of case and control individuals was obtained. Other information, including menstrual and reproductive factors (age at menarche, age at first pregnancy, age at menopause, parity, lifetime lactation history), family breast cancer history, lifetime occupational history, body mass index, hair dye use, and residence history, was also collected. Dietary information was obtained using a scannable semiquantitative food frequency questionnaire developed by the Fred Hutchinson Cancer Research Center, designed to optimize estimation of fat intake. Menopausal status was assessed at the time of diagnosis. Women with hysterectomy or bilateral oophorectomy were considered to be postmenopausal women, whereas very few women with dubious menopausal status were considered to represent missing data. At the completion of the interview, blood was drawn for DNA isolation and subsequent molecular analysis. The status of all samples – case or control – was concealed before they were handed to laboratory personnel.
Single nucleotide polymorphism selection
A SNP search using the National Center for Biotechnology Information SNP database [23] showed no non-synonymous SNPs in the coding region of the MBD2 gene. Therefore, two noncoding SNPs were chosen for genotyping. One (rs1259938) is located in the noncoding exon and another (rs609791) is located in intron 3 of the MBD2 gene. The noncoding exon is generally found at the 3'-untranslated region of a gene and this is now widely acknowledged. There is increasing evidence indicating that the 3'-untranslated region of a gene plays a vital biologic role in many post-transcriptional regulatory pathways that control mRNA localization, stability, and translation efficiency [24].
Genotyping methods
The restrictional fragment length polymorphism PCR assay was used to determine the genotypes of SNP rs1259938. The genomic DNA used for the assay was extracted from peripheral blood lymphocytes. The PCR primers used for amplifying this polymorphism were as follows: forward 5'-CCTTGCCTGTGACTTGGACT-3' and reverse 5'-TCGCGAGTTTCAACAGAAAA-3'. Standard PCR was performed in a 25 μl volume with annealing temperature at 58°C and followed by an overnight digestion with XbaI (New England BioLabs, Beverly, MA, USA) at 37°C. The products were separated for 45 min at 220 V on a 4% agarose gel stained with ethidium bromide. Following electrophoresis, the homozygous G/G alleles were represented by a DNA band with size at 319 bp, whereas the homozygous A/A alleles were represented by DNA bands with sizes at 103 bp and 216 bp, and the heterozygotes displayed a combination of both alleles (103 bp, 216 bp and 319 bp).
The TaqMan Assay was used to determine the genotypes of SNP rs609791. Assays-on-Demand primers and probes (C_3079439_10; Applied Biosystems, Inc., Foster City, CA, USA) were mixed with PCR reagents following the manufacturer's instructions in the TaqMan assay. Plates were sealed and cycled at 95°C for 5 min, followed by 45 cycles at 92°C for 15 seconds, and 60°C for 1 min in a Stratagene Real-Time Mx3000 thermocycler (Stratagene Corp., La Jolla, CA, USA).
Each genotyping plate contained positive and negative controls. Approximately 5% of the samples were duplicated to ensure quality control in genotyping and two reviewers separately performed genotype scoring to confirm results.
Statistical analysis
Because more than 90% of the study participants were Caucasians, with about 6% being black, 1% Asian, and 2% other races, we restricted our analysis to Caucasians only (393 cases and 436 controls). Pearson's χ2 test was used to evaluate differences in the distribution of selected characteristics between cases and controls. Genotype frequencies at both SNP loci in the control population were first checked for compliance with Hardy–Weinberg equilibrium using STATA statistical software (StataCorp, LP, College Station, TX, USA). Haplotype estimation was calculated using the PHASE program, which reconstructs haplotypes from population genotyping data [25]. The best haplotypes estimated by the PHASE were assigned to each study participant. STATA was also used to calculate both crude and adjusted odds ratios (ORs). ORs with 95% confidence intervals (CIs) were reported to illustrate relative cancer risk associated with genotypes and haplotypes.
For a SNP genotype, study participants with homozygous common allele were used as the reference group in OR calculation. For haplotype analysis, the most common haplotype (G-C) was used as the reference group in risk estimation. Logistic regression was used to control for confounding by age (as a continuous variable), body mass index (<25 kg/m2, 25–29.99 kg/m2, >29.99 kg/m2), family history of breast cancer in first-degree relatives, family income (tertiles based on distribution of controls), lifetime months of breastfeeding (never, 1–5, 6–15, >15 months) and study site (New Haven County, Tolland County). Control of other variables (such as age at menarche, age at menopause, number of live births) did not change the ORs significantly, and these variables were not included in the final model.
Results
This study, which included 393 breast cancer cases and 436 controls, was composed entirely of Caucasians. Table 1 presents the distribution of selected baseline characteristics for cases and controls. There were significantly more postmenopausal women in the case population (77.6%) than among controls (66.5%), indicating an increased risk for breast cancer associated with menopausal status. In addition, data showed that more controls had higher family income (31%) compared to the incidence of high family income in breast cancer cases (23.5%). No other baseline factors exhibited a material difference between cases and controls.
The genotype distributions of both SNP loci for cases and controls (Table 2) were in Hardy–Weinberg equilibrium (χ2 = 0.35, P = 0.56 for rs1259938; χ2 = 0.31, P = 0.57 for rs609791).
Among all women, we found no overall associations between genotypes at these two loci and breast cancer risk after adjustment for age, menopausal status, family history of breast cancer in first-degree relatives, family income, body mass index, lifetime months of breastfeeding, and study site. Among premenopausal women, a reduced breast cancer risk was significantly associated with variant genotypes (homozygous minor allele + heterozygote) at both SNP loci. Specifically, women with G/A and A/A at rs1259938 had 59% reduced breast cancer risk (OR = 0.41, 95% CI = 0.23–0.72) and women with C/G and G/G at rs609791 had 46% reduced breast cancer risk (OR = 0.54, 95% CI = 0.30–0.96). However, no significant associations were detected among postmenopausal women.
These two SNP loci in the MBD2 gene may generate four possible haplotypes, and their frequency distributions among cases and controls are shown in Table 3. G-C was the most common haplotype, with a frequency of 66.90% in our control group. The frequencies of the other three haplotypes were 5.56% (G-G), 11.70% (A-C), and 15.84% (A-G) in controls. Among all female participants, none of these haplotypes was associated with breast cancer risk. However, in premenopausal women the two rare haplotypes halved breast cancer risk, with ORs of 0.40 (95% CI = 0.20–0.83, P = 0.013) for A-C and 0.47 (95% CI = 0.26–0.84, P = 0.011) for A-G. Similar associations were not observed in postmenopausal women.
Discussion
It is becoming clear that carcinogenesis is a stepwise process of accumulation of both genetic and epigenetic abnormalities that can lead to cellular dysfunction. A large body of evidence has demonstrated that the epigenetic process is involved in breast carcinogenesis by influencing several broad gene categories, including cell cycle regulation, cell growth, steroid receptors, tumor susceptibility, carcinogen detoxification, cell adhesion, and inhibitors of matrix metalloproteinase genes. For example, methylation of p16 promoter and exon 1 regions are observed in both human breast cancer cell lines and 20–30% of primary breast cancers [26,27]. Methylation of the promoter region of GSTP1, a member of the glutathione S-transferases, which are a supergene family involved in the detoxification of carcinogens, is associated with gene inactivation in about 30% of primary breast carcinomas [28]. DNA methylation has also been found to be an alternative mechanism of inactivation of BRCA1 [29,30], a gene that accounts for one half of inherited breast carcinomas [31]. Moreover, three members of the steroid hormone superfamily, including estrogen receptor, progesterone receptor and retinoic acid receptor, have long been linked to mammary carcinogenesis [32] and recent studies have shown that epigenetic alterations appear to play a role in silencing estrogen receptors and retinoic acid receptors in breast malignancy [33-35]. Given the increasing evidence for a role of the epigenetic process, especially DNA methylation, in breast carcinogenesis, it is speculated that some genetic variations in methylation related genes may affect the expression and function of these genes and consequently contribute to breast cancer development.
Findings from the present study show associations between genotypes and haplotypes of the MBD2 gene and breast cancer, which have not previously been examined. These results support a potential role for methylation related genes in breast tumorigenesis. Interestingly, MBD2 polymorphisms have different effects in women depending on menopausal status. Our results demonstrate significant associations between MBD2 genotypes and haplotypes and breast cancer risk in premenopausal women but not in postmenopausal women. In fact, menopausal effects on breast cancer risk have also been observed in a previous study investigating genetic polymorphisms in catechol-O-methyltransferase [36]. That study found that the low-activity allele of catechol-O-methyltransferase was associated with increased risk among premenopausal women (OR = 2.1, 95% CI = 1.4–4.3) but was inversely associated with postmenopausal risk (OR = 0.4, 95% CI = 0.2–0.7). Our findings support arguments from previous studies suggesting that different etiologies may be involved in breast carcinogenesis between premenopausal and postmenopausal women [36,37].
Although the mechanisms are not elucidated, menopausal effects on the role of MBD2 in breast cancer development may be related to changes in sex hormone levels. One of the phases of breast cancer pathogenesis is exposure of breast tissue to ovarian hormones that drive the kinetics of breast tissue stem cells, resulting in carcinogenesis [38]. Dividing cells are particularly susceptible to alterations in DNA synthesis, DNA repair, and DNA methylation.
These biologic and physiologic effects of sex hormones are controlled by hormone receptors, the expression of which is regulated by the methylation status of their promoter regions [33-35].
On the other hand, steroid hormones may influence the epigenetic blue print of methylation of certain genes and consequently activate or inactivate gene expression [39]. Even though MBD2 might be involved in the epigenetic regulation of steroid hormone receptor gene expression, MBD2 itself could also be affected by steroid hormones. It is possible that MBD2 plays different roles in breast carcinogenesis when hormone levels dramatically change. Our findings support this speculation in that we found a significantly protective role of MBD2 variants in premenopausal women but no significant associations in postmenopausal women.
There are limitations to the present study. Only Caucasian women were included in the study, and so hypotheses must be further examined in multi-ethnic groups. In addition, the sample sizes of our study limit the analyses to explore other potential risk factors. Traditional risk factors such as parity and family history of cancer did not differ between cases and controls, which could also be due to the sample sizes. Inaccurate recall may affect our assessment of family history of cancer as well.
Conclusion
These findings imply a potential link between DNA methylation processes and hormonal expression. Although large molecular epidemiologic studies are warranted to further examine associations between MBD2 polymorphisms and breast cancer in multi-ethnic groups, this study suggests that genetic variations in methylation related genes may serve as a promising biomarker in risk estimate of breast cancer.
Abbreviations
bp = base pairs; CI = confidence interval; MBD = methyl-CpG-binding domain; OR = odds ratio; PCR = polymerase chain reaction; SNP = single nucleotide polymorphism.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
YZ designed the study and drafted the manuscript. HNB carried out the genotyping analysis. YZ and TRH performed the statistical analysis. TZ is the principal investigator for the parent case–control study and participated in the design of the present project and manuscript preparation. All authors read and approved the final manuscript.
Acknowledgements
This work was supported by the funds from Yale University. The work was also supported by NIH grants CA62986, CA81810, CA110937, and CA108369. We also thank Derek Leaderer and Amy Pimentel for laboratory assistance and Carly Guss for editorial help.
Figures and Tables
Table 1 Distributions of selected characteristics by case–control status in Caucasians
Variable Cases (n = 393) Controls (n = 436) P
Age at menarche (years)
<11 23 (5.9) 29 (6.7)
11–12 152 (39.2) 168 (38.6)
≥ 12 213 (54.9) 238 (54.7) 0.907
Menopausal status
Premenopausal 88 (22.4) 146 (33.5)
Postmenopausal 305 (77.6) 290 (66.5) 0.000
Age at menopause (years)
<44 76 (25.8) 91 (32.3)
44–49 105 (35.6) 87 (30.9)
≥ 49 114 (38.6) 104 (36.8) 0.202
Lifetime lactation (months)
0 242 (61.6) 260 (59.6)
1–5 50 (12.7) 61 (14.0)
6–15 56 (14.3) 59 (13.5)
≥ 15 45 (11.4) 56 (12.9) 0.853
Family history of breast cancer in first-degree relatives
Yes 94 (24.0) 98 (22.5)
No 299 (76.0) 338 (77.5) 0.623
Body mass index (kg/m2)
<25.0 72 (18.3) 77 (17.7)
25.0–29.9 100 (25.5) 110 (25.2)
≥ 30.0 221 (56.2) 249 (57.1 0.960
Fat intake
Low 120 (31.0) 141 (33.2)
Medium 132 (34.1) 143 (33.6)
High 135 (34.9) 141 (33.2) 0.786
Cigarette smoking
Never 169 (43.0) 196 (45.0)
Ever 224 (57.0) 239 (55.0) 0.534
Alcohol consumption
Never 66 (19.2) 64 (16.9)
Ever 277 (80.8) 315 (83.1) 0.411
Annual income
Low 126 (39.0) 125 (33.4)
Medium 121 (37.5) 133 (35.6)
High 76 (23.5) 116 (31.0) 0.074
Study site
New Haven County 266 (67.7) 282 (64.7)
Tolland County 127 (32.3) 154 (35.3) 0.361
Live births
0 46 (11.95) 66 (15.46)
1–2 200 (51.95) 202 (47.31
≥ 2 139 (36.10) 159 (37.24) 0.252
Note that missing data for each characteristic were excluded from the analyses. Values are expressed as n (%).
Table 2 MBD2 genotypes, menopausal status and breast cancer risk in Caucasians
MBD2 genotype Cases (n = 393) Controls (n = 436) Crude OR (95% CI) ORa (95% CI)
SNP 1 (rs1259938)
All women
G/G 210 (53.4) 226 (51.8) 1.00 1.00
G/A 158 (40.2) 179 (41.1) 0.95 (0.71–1.27) 0.94 (0.70–1.26)
A/A 25 (6.4) 31 (7.1) 0.86 (0.49–1.52) 0.86 (0.49–1.54)
G/A + A/A 183 (46.6) 210 (48.2) 0.94 (0.71–1.23) 0.94 (0.71–1.24)
Premenopausal women
G/G 59 (67.1) 69 (47.3) 1.00 1.00
G/A 25 (28.4) 65 (44.5) 0.43 (0.24–0.77) 0.40 (0.22–0.74)
A/A 4 (4.6) 12 (8.2) 0.39 (0.12–1.26) 0.39 (0.12–1.29)
G/A + A/A 29 (33.0) 77 (52.7) 0.42 (0.24–0.73) 0.41 (0.23–0.72)
Postmenopausal women
G/G 151 (49.5) 157 (54.1) 1.00 1.00
G/A 133 (43.6) 114 (39.3) 1.24 (0.88–1.74) 1.24 (0.88–1.75)
A/A 21 (6.9) 19 (6.6) 1.16 (0.59–2.27) 1.15 (0.58–2.29)
G/A + A/A 154 (50.5) 134 (45.9) 1.23 (0.90–1.70) 1.24 (0.90–1.72)
SNP 2 (rs609791)
All women
C/C 243 (63.9) 260 (61.6) 1.00 1.00
C/G 120 (31.6) 140 (33.2) 0.92 (0.68–1.24) 0.92 (0.68–1.24)
G/G 17 (4.5) 22 (5.2) 0.83 (0.43–1.59) 0.85 (0.44–1.64)
C/G + G/G 237 (36.1) 162 (38.4) 0.90 (0.68–1.21) 0.91 (0.68–1.22)
Premenopausal women
C/C 60 (70.6) 80 (56.3) 1.00 1.00
C/G 20 (23.5) 55 (38.7) 0.48 (0.26–0.89) 0.47 (0.26–0.88)
G/G 5 (5.9) 7 (5.0) 0.95 (0.29–3.15) 1.02 (0.30–3.49)
C/G + G/G 25 (29.4) 62 (43.7) 0.54 (0.30–0.95) 0.54 (0.30–0.96)
Postmenopausal women
C/C 183 (62.0) 180 (64.3) 1.00 1.00
C/G 100 (33.9) 85 (30.4) 1.16 (0.81–1.65) 1.15 (0.80–1.64)
G/G 12 (4.1) 15 (5.3) 0.79 (0.36–1.73) 0.80 (0.36–1.76)
C/G + G/G 112 (38.0) 100 (35.7) 1.10 (0.78–1.55) 1.10 (0.78–1.55)
Values are expressed as n (%). aAdjusted for age (as a continuous variable), body mass index (<25 kg/m2, 25–29.99 kg/m2, >29.99 kg/m2), family history of breast cancer in first-degree relatives, menopausal status, family income (tertiles based on distribution of controls), lifetime months of breast feeding (never, 1–5, 6–15, ≥ 15 months) and study site. CI, confidence interval; OR, odds ratio.
Table 3 MBD2 Haplotypes, menopausal status and breast cancer risk in Caucasians
MBD2 Haplotypes Frequency in cases Frequency in controls Crude OR (95% CI) ORa (95% CI)
All women
GC 69.22% 66.90% Reference Reference
GG 4.81% 5.56% 0.84 (0.53–1.31) 0.82 (0.52–1.29)
AC 11.17% 11.70% 0.92 (0.68–1.26) 0.92 (0.67–1.25)
AG 14.81% 15.84% 0.90 (0.69–1.19) 0.90 (0.68–1.19)
Premenopausal women
GC 77.65% 63.29% Reference Reference
GG 5.29% 5.59% 0.77 (0.33–1.80) 0.77 (0.33–1.81)
AC 6.47% 12.59% 0.42 (0.21–0.85) 0.40 (0.20–0.83)
AG 10.59% 18.53% 0.47 (0.26–0.83) 0.47 (0.26–0.84)
Postmenopausal women
GC 66.83% 68.75% Reference Reference
GG 4.67% 5.54% 0.87 (0.51–1.47) 0.84 (0.49–1.44)
AC 12.50% 11.25% 1.14 (0.80–1.64) 1.14 (0.79–1.65)
AG 16.00% 14.46% 1.14 (0.82–1.58) 1.13 (0.81–1.57)
Values are expressed as n (%). aAdjusted for age (as a continuous variable), body mass index (<25 kg/m2, 25–29.99 kg/m2, >29.99 kg/m2), family history of breast cancer in first-degree relatives, menopausal status, family income (tertiles based on distribution of controls), lifetime months of breast feeding (never, 1–5, 6–15, ≥15 months) and study site. CI, confidence interval; OR, odds ratio.
==== Refs
Jones PA Epigenetics in carcinogenesis and cancer prevention Ann NY Acad Sci 2003 983 213 219 12724226
Razin A CpG methylation, chromatin structure and gene silencing-a three-way connection EMBO J 1998 17 4905 4908 9724627 10.1093/emboj/17.17.4905
Davey C Pennings S Allan J CpG methylation remodels chromatin structure in vitro J Mol Biol 1997 267 276 288 9096225 10.1006/jmbi.1997.0899
Antequera F Boyes J Bird A High levels of de novo methylation and altered chromatin structure at CpG islands in cell lines Cell 1990 62 503 514 1974172 10.1016/0092-8674(90)90015-7
Cross SH Meehan RR Nan X Bird A A component of the transcriptional repressor MeCP1 shares a motif with DNA methyltransferase and HRX proteins Nat Genet 1997 16 256 259 9207790 10.1038/ng0797-256
Hendrich B Bird A Identification and characterization of a family of mammalian methyl-CpG binding proteins Mol Cell Biol 1998 18 6538 6547 9774669
Nan X Ng HH Johnson CA Laherty CD Turner BM Eisenman RN Bird A Transcriptional repression by the methyl-CpG-binding protein MeCP2 involves a histone deacetylase complex Nature 1998 393 386 389 9620804 10.1038/30764
Patra SK Patra A Zhao H Carroll P Dahiya R Methyl-CpG-DNA binding proteins in human prostate cancer: expression of CXXC sequence containing MBD1 and repression of MBD2 and MeCP2 Biochem Biophys Res Commun 2003 302 759 766 12646234 10.1016/S0006-291X(03)00253-5
Schlegel J Guneysu S Mennel HD Expression of the genes of methyl-binding domain proteins in human gliomas Oncol Rep 2002 9 393 395 11836615
Hendrich B Abbott C McQueen H Chambers D Cross S Bird A Genomic structure and chromosomal mapping of the murine and human Mbd1, Mbd2, Mbd3, and Mbd4 genes Mamm Genome 1999 10 906 912 10441743 10.1007/s003359901112
Snape A MBDs mediate methylation, deacetylation and transcriptional repression Trends Genet 2000 16 20 10647297 10.1016/S0168-9525(99)01925-3
Wade PA Methyl CpG-binding proteins and transcriptional repression Bioessays 2001 23 1131 1137 11746232 10.1002/bies.10008
Bhattacharya SK Ramchandani S Cervoni N Szyf M A mammalian protein with specific demethylase activity for mCpG DNA Nature 1999 397 579 583 10050851 10.1038/17533
Ng HH Zhang Y Hendrich B Johnson CA Turner BM Erdjument-Bromage H Tempst P Reinberg D Bird A MBD2 is a transcriptional repressor belonging to the MeCP1 histone deacetylase complex Nat Genet 1999 23 58 61 10471499
Wade PA Gegonne A Jones PL Ballestar E Aubry F Wolffe AP Mi-2 complex couples DNA methylation to chromatin remodelling and histone deacetylation Nat Genet 1999 23 62 66 10471500
Billard LM Magdinier F Lenoir GM Frappart L Dante R MeCP2 and MBD2 expression during normal and pathological growth of the human mammary gland Oncogene 2002 21 2704 2712 11965543 10.1038/sj.onc.1205357
Lin X Nelson WG Methyl-CpG-binding domain protein-2 mediates transcriptional repression associated with hypermethylated GSTP1 CpG islands in MCF-7 breast cancer cells Cancer Res 2003 63 498 504 12543808
Kanai Y Ushijima S Nakanishi Y Hirohashi S Reduced mRNA expression of the DNA demethylase, MBD2, in human colorectal and stomach cancers Biochem Biophys Res Commun 1999 264 962 966 10544038 10.1006/bbrc.1999.1613
Zhu Y Spitz MR Zhang H Grossman HB Frazier ML Wu X Methyl-CpG-binding domain 2: a protective role in bladder carcinoma Cancer 2004 100 1853 1858 15112265 10.1002/cncr.20199
Patra SK Patra A Zhao H Dahiya R DNA methyltransferase and demethylase in human prostate cancer Mol Carcinog 2002 33 163 171 11870882 10.1002/mc.10033
Zheng T Holford TR Mayne ST Tessari J Ward B Carter D Owens PH Boyle P Dubrow R Archibeque-Engle S Risk of female breast cancer associated with serum polychlorinated biphenyls and 1,1-dichloro-2,2'-bis(p-chlorophenyl)ethylene Cancer Epidemiol Biomarkers Prev 2000 9 167 174 10698477
Singletary SE Allred C Ashley P Bassett LW Berry D Bland KI Borgen PI Clark GM Edge SB Hayes DF Staging system for breast cancer: revisions for the 6th edition of the AJCC Cancer Staging Manual Surg Clin North Am 2003 83 803 819 12875597 10.1016/S0039-6109(03)00034-3
Single Nucleotide Polymorphism Database
Pesole G Mignone F Gissi C Grillo G Licciulli F Liuni S Structural and functional features of eukaryotic mRNA untranslated regions Gene 2001 276 73 81 11591473 10.1016/S0378-1119(01)00674-6
Stephens M Smith NJ Donnelly P A new statistical method for haplotype reconstruction from population data Am J Hum Genet 2001 68 978 989 11254454 10.1086/319501
Woodcock DM Linsenmeyer ME Doherty JP Warren WD DNA methylation in the promoter region of the p16 (CDKN2/MTS-1/INK4A) gene in human breast tumours Br J Cancer 1999 79 251 256 9888465
Herman JG Merlo A Mao L Lapidus RG Issa JP Davidson NE Sidransky D Baylin SB Inactivation of the CDKN2/p16/MTS1 gene is frequently associated with aberrant DNA methylation in all common human cancers Cancer Res 1995 55 4525 4530 7553621
Esteller M Corn PG Urena JM Gabrielson E Baylin SB Herman JG Inactivation of glutathione S-transferase P1 gene by promoter hypermethylation in human neoplasia Cancer Res 1998 58 4515 4518 9788592
Magdinier F Billard LM Wittmann G Frappart L Benchaib M Lenoir GM Guerin JF Dante R Regional methylation of the 5' end CpG island of BRCA1 is associated with reduced gene expression in human somatic cells FASEB J 2000 14 1585 1594 10928993 10.1096/fj.14.11.1585
Dobrovic A Simpfendorfer D Methylation of the BRCA1 gene in sporadic breast cancer Cancer Res 1997 57 3347 3350 9269993
Friedman LS Ostermeyer EA Lynch ED Szabo CI Anderson LA Dowd P Lee MK Rowell SE Boyd J King MC The search for BRCA1 Cancer Res 1994 54 6374 6382 7987831
Fishman J Osborne MP Telang NT The role of estrogen in mammary carcinogenesis Ann N Y Acad Sci 1995 768 91 100 8526389
Ottaviano YL Issa JP Parl FF Smith HS Baylin SB Davidson NE Methylation of the estrogen receptor gene CpG island marks loss of estrogen receptor expression in human breast cancer cells Cancer Res 1994 54 2552 2555 8168078
Archey WB McEachern KA Robson M Offit K Vaziri SA Casey G Borg A Arrick BA Increased CpG methylation of the estrogen receptor gene in BRCA1-linked estrogen receptor-negative breast cancers Oncogene 2002 21 7034 7041 12370825 10.1038/sj.onc.1205844
Widschwendter M Berger J Hermann M Muller HM Amberger A Zeschnigk M Widschwendter A Abendstein B Zeimet AG Daxenbichler G Methylation and silencing of the retinoic acid receptor-beta2 gene in breast cancer J Natl Cancer Inst 2000 92 826 832 10814678 10.1093/jnci/92.10.826
Thompson PA Shields PG Freudenheim JL Stone A Vena JE Marshall JR Graham S Laughlin R Nemoto T Kadlubar FF Genetic polymorphisms in catechol-O-methyltransferase, menopausal status, and breast cancer risk Cancer Res 1998 58 2107 2110 9605753
Lee HP Gourley L Duffy SW Esteve J Lee J Day NE Risk factors for breast cancer by age and menopausal status: a case-control study in Singapore Cancer Causes Control 1992 3 313 322 1617118 10.1007/BF00146884
Henderson BE Ross R Bernstein L Estrogens as a cause of human cancer: the Richard and Hinda Rosenthal Foundation award lecture Cancer Res 1988 48 246 253 2825969
Jost JP Saluz HP Steroid hormone dependent changes in DNA methylation and its significance for the activation or silencing of specific genes EXS 1993 64 425 451 8380354
|
16168120
|
PMC1242141
|
CC BY
|
2021-01-04 16:54:51
|
no
|
Breast Cancer Res. 2005 Jul 19; 7(5):R745-R752
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1283
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12851616812110.1186/bcr1285Research ArticlePhosphorylation of estrogen receptor α serine 167 is predictive of response to endocrine therapy and increases postrelapse survival in metastatic breast cancer Yamashita Hiroko [email protected] Mariko [email protected] Shunzo [email protected] Yoshiaki [email protected] Hiroshi [email protected] Zhenhuan [email protected] Maho [email protected] Keiko [email protected] Yoshitaka [email protected] Hirotaka [email protected] Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan2 Oncology and Endocrinology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan3 Josai Municipal Hospital of Nagoya, Nagoya, Japan2005 27 7 2005 7 5 R753 R764 29 1 2005 15 4 2005 12 6 2005 28 6 2005 Copyright © 2005 Yamashita et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Endocrine therapy is the most important treatment option for women with hormone-receptor-positive breast cancer. The potential mechanisms for endocrine resistance involve estrogen receptor (ER)-coregulatory proteins and crosstalk between ER and other growth factor signaling networks. However, the factors and pathways responsible for endocrine resistance are still poorly identified.
Methods
Using immunohistochemical techniques, we focused on the expression and phosphorylation of hormone receptors themselves and examined the phosphorylation of ER-α Ser118 and ER-α Ser167 and the expression of ER-α, ER-β1, ER-βcx/β2, progesterone receptor (PR), PRA, and PRB in the primary breast carcinomas of 75 patients with metastatic breast cancer who received first-line treatment with endocrine therapy after relapse.
Results
Phosphorylation of ER-α Ser118, but not Ser167, was positively associated with overexpression of HER2, and HER2-positive tumors showed resistance to endocrine therapy. The present study has shown for the first time that phosphorylation of ER-α Ser167, but not Ser118, and expression of PRA and PRB, as well as ER-α and PR in primary breast tumors are predictive of response to endocrine therapy, whereas expression of ER-β1 and ER-βcx/β2 did not affect response to the therapy. In addition, patients with either high phosphorylation of ER-α Ser167, or high expression of ER-α, PR, PRA, or PRB had a significantly longer survival after relapse.
Conclusion
These data suggest that phosphorylation of ER-α Ser167 is helpful in selecting patients who may benefit from endocrine therapy and is a prognostic marker in metastatic breast cancer.
==== Body
Introduction
The development and progression of breast cancer are influenced by steroid hormones, particularly estrogen, via their interaction with specific target receptors. Endocrine therapy has become the most important treatment option for women with estrogen receptor (ER)-positive breast cancer. Nevertheless, many breast cancer patients with tumors expressing high levels of ER are unresponsive to endocrine therapy, and all patients with advanced disease eventually develop resistance to the therapy. The potential mechanisms behind either this intrinsic or acquired endocrine resistance involve ER-coregulatory proteins and crosstalk between the ER pathway and other growth factor signaling networks [1,2]. An understanding of the molecular mechanisms that modulate the activity of the estrogen signaling network has enabled new ways of overcoming endocrine resistance to be developed.
ER-α is phosphorylated on multiple amino acid residues [3]. Serines 104, 106, 118, and 167 are all located within the activation function (AF)1 region of ER-α, and their phosphorylation provides the important mechanism that regulates AF1 activity [4,5]. In response to estradiol binding, human ER-α is phosphorylated mainly on Ser118 and to a lesser extent on Ser104 and Ser106 [4]. Although some authors have also reported that Ser167 is a major estradiol-induced phosphorylation site [5,6], this response to estradiol has not been universally observed [4,7]. Interestingly, in response to the activation of the mitogen-activated protein kinase (MAPK) pathway, phosphorylation occurs on Ser118 and Ser167 [8,9]. However, the role of phosphorylation of Ser118 and Ser167 of ER-α in human breast cancer has not been investigated.
ER-β and its splicing isoforms are widely expressed in both normal and malignant breast tissue [10]. Although several groups have reported results regarding the possible function of ER-β, and its potential as a prognostic or predictive factor in breast cancer, the data remain inconclusive and are often contradictory [11,12]. ER-βcx (also called ER-β2), a splice variant of ER-β, is considered to be a dominant repressor of ER-α; it is identical to ER-β1 (wild-type ER-β) except that the last exon, 8, is replaced by 26 amino acid residues [13]. The role of ER-β and its isoforms, especially with respect to the response of breast cancer to endocrine therapy, has also not been elucidated.
Progesterone receptors (PRs) occur as two isoforms, PRA and PRB, transcribed from two distinct promoters on a single gene. PRA, but not PRB, lacks the 164 amino acid N-terminal residues that contain AF3, and this is the cause of their functional differences [14]. In the mammary gland, the overexpression of PRA relative to PRB results in extensive epithelial cell hyperplasia, excessive ductal branching, and a disorganized basement membrane, all features associated with neoplasia [15]. In contrast, the overexpression of PRB leads to premature arrest of ductal growth and inadequate lobuloalveolar differentiation [16]. However, little is known about the unique roles of the two PR isoforms in breast cancer.
In this study, we focused on the expression and phosphorylation of the hormone receptors themselves and, using immunohistochemistry (IHC), examined the phosphorylation of ER-α Ser118 and Ser167 and the expression of ER-α, ER-β1, ER-βcx/β2, PR, PRA, and PRB in primary breast tumor specimens from 75 patients with metastatic breast cancer who received first-line treatment with endocrine therapy on relapse. Our results show that patients with primary breast tumors in which there is either high phosphorylation of ER-α Ser167 or high expression of ER-α, PR, PRA, or PRB significantly responded to endocrine therapy and had a better survival after relapse.
Materials and methods
Cell culture and transfections
COS-7 cells (ATCC American Type Culture Collection, Manassas, VA, USA) were grown in DMEM containing 10% fetal calf serum, 2 mM L-glutamine, and penicillin–streptomycin (50 IU/ml and 50 mg/ml, respectively) at 37°C with 5% CO2 as described previously [17]. T47D cells (ATCC) were grown in RPMI-1640 supplemented with 10% fetal calf serum, 2 mM L-glutamine, and penicillin–streptomycin (50 IU/ml and 50 mg/ml, respectively), at 37°C with 5% CO2. Six microliters of FuGENE6 transfection reagent (Roche Molecular Biochemicals, Indianapolis, IN, USA), 3 μg of an expression vector for human ER-α cDNA (pSG5/puromycine hERα, full length, kindly provided by Pierre Chambon, Strasbourg, France) were used for transfection into COS-7 cells as described previously [18]. After transfection, cells were starved in serum-free DMEM without phenol red for 20 hours.
Immunoblotting
Cells were treated in the absence or presence of 17β-estradiol (E2) (10 nM, Sigma-Aldrich Co, St Louis, MO, USA) and/or epidermal growth factor (EGF) (100 ng/ml, human recombinant EGF, Sigma-Aldrich) for 30 min, pelleted by centrifugation and solubilized in lysis buffer containing 10 mM Tris-HCl, pH 7.6, 5 mM EDTA, 50 mM NaCl, 30 mM sodium pyrophosphate, 50 mM sodium fluoride, 1 mM sodium orthovanadate, 1% Triton X-100, 1 mM phenylmethylsulfonylfluoride (PMSF), 5 μg/ml aprotinin, 1 μg/ml pepstatin A, and 2 μg/ml leupeptin, as described previously [19]. Equal amounts of total protein from whole-cell lysates were prepared and used for SDS–PAGE. Immunoblotting was performed as described previously [18] using polyvinylidene difluoride (PVDF) membranes (Invitrogen, Carlsbad, CA, USA; Catalogue no. LC2002), and polyclonal antibody against ER-α (H-184, Santa Cruz Biotechnology, Santa Cruz, CA, USA) (1:200 dilution), polyclonal rabbit antiphospho-ER-α (Ser118) antibody (Cell Signaling Technology, Beverly, MA, USA) (1:500 dilution), polyclonal rabbit antiphospho-ER-α (Ser167) antibody (Cell Signaling) (1:500 dilution) as primary antibodies and horseradish-peroxidase-conjugated goat antibodies to rabbit IgG as secondary antibodies in conjunction with enhanced chemiluminescence substrate mixture (SuperSignal WestPico Chemiluminescent Substrate, Pierce, Rockford, IL, USA) in accordance with the manufacturer's instructions. Phospho-ER-α Ser118 antibody detects ER-α only when the receptor is phosphorylated at Ser118, and not at Ser106 or Ser167; and phospho-ER-α Ser167 antibody detects ER-α only when the receptor is phosphorylated at Ser167, and not at Ser106 or Ser118, by immunoblotting, as described by Chen and colleagues [20].
Generation of specific antibodies for ER-β proteins
To detect specific ER-β1 and ER-βcx/β2 proteins, rabbit polyclonal antibodies were generated against synthesized peptides of the C-terminal region of ER-β1 (CSPAEDSKSKEGSQNPQSQ) and ER-βcx/β2 (MKMETLLPEATMEQ), in accordance with the method of Ogawa and colleagues [13] and purified on affinity columns bound with each synthetic peptide as described previously [17]. To confirm the specificity of these polyclonal antibodies, immunoblot analysis was performed using COS-7 cells transfected with either expression plasmid encoding ER-β1 or ER-βcx/β2 (kindly donated by Masami Muramatsu, Saitama, Japan) as previously described [18]. Immunoblotting with specific anti-ER-β antibodies showed that the polyclonal antibody for ER-β1 detected a specific band at 60 kDa only in the lysates of COS-7 cells transfected with an ER-β1 expression plasmid, and not in those transfected with an ER-βcx/β2 expression plasmid, as described previously by Ogawa and colleagues [13]. Conversely, the polyclonal antibody for ER-βcx/β2 detected a specific band at 55 and 51 kDa only in the lysates of COS-7 cells transfected with an ER-βcx/β2 expression plasmid, and not in those transfected with an ER-β1 expression plasmid, as described previously by Ogawa and colleagues [13].
Patients and breast cancer tissues
Breast tumor specimens from 75 women with metastatic breast cancer who were treated at Nagoya City University Hospital between 1982 and 2002 were included in this study (Table 1). The study protocol was approved by the institutional review board and conformed with the guidelines of the 1975 Declaration of Helsinki. All patients had undergone surgical treatment for primary breast cancer (either mastectomy or lumpectomy) and all primary tumors were ER+ or PR+. After surgery, five patients (6.7%) received no additional therapy. Of the remaining 71 patients, 32 (42.7%) received systemic adjuvant therapy consisting of endocrine therapy (tamoxifen) alone, two (2.7%) received chemotherapy alone, and 36 (48%) received combined endocrine therapy and chemotherapy. Patients who were positive for axillary lymph nodes received either oral administration of 5-fluorouracil derivatives for 2 years or a combination of cyclophosphamide, methotrexate, and fluorouracil (CMF). Patients were observed for disease recurrence at least once every six months for the first 5 years after the surgery and once every year thereafter.
First-line endocrine therapy for metastatic breast cancer and response criteria
When the patients relapsed and were diagnosed with metastatic breast cancer, they started endocrine therapy (Table 1). Patients were assessed monthly for clinical response, which was defined according to World Health Organization criteria as complete response, partial response, no change, or progressive disease. The presence of progressive disease indicated treatment failure; all other clinical responses were considered to show efficacy of treatment.
Immunohistochemical analysis
One 4-μm section of each submitted paraffin block was stained first with hematoxylin and eosin to verify that an adequate number of invasive carcinoma cells were present and that the fixation quality was adequate for immunohistochemical (IHC) analysis. Serial sections (4 μm) were prepared from selected blocks and float-mounted on adhesive-coated glass slides, for ER-α, ER-β, and PR staining as described previously [21]. Primary antibodies included monoclonal mouse antihuman ER-α antibody (1D5, DAKO, Glostrup, Denmark) at 1:100 dilution for ER-α; polyclonal rabbit antiphospho-ER-α (Ser118) antibody (Cell Signaling) at 1:25 dilution for phosphorylated ER-α Ser118; polyclonal rabbit antiphospho-ER-α (Ser167) antibody (Cell Signaling) at 1:50 dilution for phosphorylated ER-α Ser167; polyclonal rabbit anti-ER-β1 antibody at 1:10000 dilution for ER-β1; polyclonal rabbit anti-ER-βcx/β2 antibody at 1:2000 dilution for ER-βcx/β2; monoclonal mouse antihuman PR antibody (636, DAKO) at 1:100 dilution for PR; monoclonal mouse antihuman PR antibody (Ab-7, Neo Markers, Fremont, CA) at 1:100 dilution for PRA; and monoclonal mouse antihuman PR antibody (Ab-2, Neo Markers) at 1:100 dilution for PRB. With respect to the PRA and PRB antibodies, it has been reported that whereas AB-7 can recognize high-PRA and low-PRB forms, this antibody recognizes PRA only in 10% formalin-fixed and paraffin-embedded tissue sections, and AB-2 recognizes exclusively PRB in these same media [22]. The DAKO Envision system (DAKO EnVision labelled polymer, peroxidase) was used as the detection system as described previously [21]. HER2 immunostaining was done and evaluated using a method similar to the HercepTest (DAKO) [21].
Immunohistochemical scoring
Immunostained slides were scored after the entire slide had been evaluated by light microscopy. The expression and phosphorylation of hormone receptors were scored by assigning proportion and intensity scores, in accorance with the procedure of Allred and colleagues [23]. In brief, a proportion score represented the estimated proportion of tumor cells staining positive, as follows: 0 (none); 1 (<1/100); 2 (1/100 to 1/10); 3 (>1/10 to 1/3); 4 (>1/3 to 2/3); and 5 (>2/3). Any brown nuclear staining in invasive breast epithelium counted towards the proportion score. An intensity score represented the average intensity of the positive cells, as follows: 0 (none); 1 (weak); 2 (intermediate); and 3 (strong). The proportion and intensity scores were then added to obtain a total score, which could range from 0 to 8.
Statistical analysis
The Mann–Whitney U test or the Kruskal–Wallis test was used to compare the IHC scores of hormone receptors with clinicopathological characteristics. The Mann–Whitney U test and the unpaired t-test were used to compare the IHC scores of hormone receptors with response to endocrine therapy. The Spearman rank correlation test was used to study relations between expression and phosphorylation of hormone receptors and disease-free interval. To examine the change of expression and phosphorylation status between the primary and recurrent tumors, the one-sample Wilcoxon signed rank test was used. Estimation of overall survival was performed using the Kaplan–Meier method, and differences between survival curves were assessed with the log-rank test. Cox's proportional hazards model was used for univariate and multivariate analyses of prognostic values.
Results
Phosphorylation of ER-α Ser118 and ER-α Ser167 is induced in response to EGF
To test the ability of site-specific antiphosphoserine antibodies for ER-α Ser118 and ER-α Ser167, we examined the phosphorylation status of these two serines in transfected COS-7 cells by immunoblotting. Cells were grown in serum- and estrogen-deprived conditions and treated with vehicle (medium) (Fig. 1a, lane 1), E2 (lane 2), EGF (lane 3), or E2 and EGF (lane 4). Immunoblotting of replicate samples with antiphohphoserine antibodies showed that ER-α was constitutively phosphorylated on Ser118 (Fig. 1a, lane 1, top panel), but not on Ser167 (Fig. 1a, lane 1, second panel). ER-α became inducibly phosphorylated on both residues Ser118 and Ser167 in response to EGF (Fig. 1a, lanes 3 and 4, top and second panels). On the other hand, phosphorylation of Ser118 and Ser167 was not affected with E2 treatment in our analysis (Fig. 1a, lane 1 vs 2, lane 3 vs 4, top and second panels). Expression of ER-α was observed equally in each condition (Fig. 1a, bottom panel).
To further validate the ability of site-specific antibodies for phospho-ER-α Ser118 and Ser167 in ER+ and PR+ breast cancer cells, phosphorylation of ER-α Ser118 and Ser167 was analyzed in T47D cells. Cells were grown in serum- and estrogen-deprived conditions and treated with vehicle (medium) (Fig. 1b, lane 1), E2 for 10 min (lane 2) and 30 min (lane 3), EGF for 10 min (lane 4) and 30 min (lane 5), or E2 and EGF for 10 min (lane 6) and 30 min (lane 7). ER-α was inducibly phosphorylated on Ser167 in response to EGF (Fig. 1b, lanes 1, 4, 5, 6, and 7, second panel). Phosphorylation of Ser167 was increased with EGF treatment for 30 min compared with that for 10 min (lane 5 vs 4, lane 7 vs 6). On the other hand, ER-α was not phosphorylated on Ser167 in response to E2 (Fig. 1b, lanes 2 and 3, second panel). In addition, ER-α was not phosphorylated on Ser118 in response to either E2 or EGF (Fig. 1b, lanes 2 to 7, top panel) in T47D cells. Expression of ER-α was observed equally under both conditions (Fig. 1b, bottom panel). We concluded from immunoblotting that, in response to EGF in COS-7 and T47D cells, ER-α Ser167 was inducibly rather than constitutively phosphorylated, and that the phosphorylation of ER-α Ser118 was constitutive and further induced by EGF in COS-7 cells, but that Ser118 phosphorylation was not observed after the stimulation of T47D cells by either E2 or EGF.
Immunohistochemical staining for phosphorylation of ER-α Ser118 and ER-α Ser167, and expression of ER-β 1, ER-βcx/β2, PRA, and PRB in human breast cancer
To investigate the phosphorylation of ER-α Ser118 and ER-α Ser167 in human breast cancer specimens, IHC analysis was performed using the same site-specific antiphosphoserine antibodies served in the immunoblotting. IHC for phosphorylated ER-α Ser118 (Fig. 2a) and ER-α Ser167 (Fig. 2d) showed presence of nuclear staining in some but not all cells of normal breast epithelium (Fig. 2). Cancer cell nuclei of invasive carcinoma tissues were positively stained with specific antibodies for phosphorylated ER-α Ser118 (Fig. 2c) and ER-α Ser167 (Fig. 2f). Specific detection of expression of ER-β1 (Fig. 3a), ER-βcx/β2 (Fig. 3b), PRA (Fig. 3c), and PRB (Fig. 3d) showed positive nuclear staining in carcinoma cells.
Correlation between expression and phosphorylation of ER-α, ER-β, and PR and clinicopathological factors in primary breast tumors
We examined the phosphorylation of ER-α Ser118 and Ser167, and expression of ER-α, ER-β1, ER-βcx/β2, PR, PRA, and PRB in 75 primary invasive breast carcinomas by IHC. The IHC scores for ER-α, ER-β, and PR were compared among patient subgroups, according to clinicopathological factors. Phosphorylation of ER-α Ser118, but not of ER-α Ser167, was positively correlated with expression levels of HER2 (P = 0.038), whereas expression of ER-α (1D5) tended to be inversely correlated with HER2 overexpression. PR (636) expression was significantly associated with age (P = 0.0018). There was no difference between the expression and phosphorylation of hormone receptors and other clinicopathological factors.
Correlation between expression and phosphorylation of ER-α, ER-β, and PR in primary breast tumors
Links between the IHC scores for the expression and phosphorylation of hormone receptors were analyzed using the Spearman's rank correlation test (Table 2). Phosphorylation of ER-α Ser118 was strongly and positively associated with phosphorylation of ER-α Ser167 (P < 0.0001) and with expression of ER-β1 (P < 0.0001), ER-βcx/β2 (P < 0.0001), and PRA (P < 0.0001). Phosphorylation of ER-α Ser167 was also positively correlated with expression of ER-β1 (P = 0.0003), ER-βcx/β2 (P < 0.0001), and PRA (P = 0.0007), whereas no association was found between phosphorylation of ER-α Ser118/Ser167 and expression of ER-α (1D5). A significant correlation was observed between expression levels of ER-β1 and ER-βcx/β2 (P < 0.0001). Expression of ER-βcx/β2, but not ER-β1, was positively correlated with expression of PRA (P = 0.0011) and PRB (P = 0.0052). Strong associations were found between expression of PR (636), PRA, and PRB (P < 0.0001, respectively). Expression of ER-α (1D5) was significantly correlated with expression of PR (636) (P = 0.0001) and PRA (P = 0.028), but not with PRB.
Phosphorylation of ER-α Ser167, but not Ser118, and expression of PRA and PRB in primary breast tumors are predictive of response to endocrine therapy in metastatic breast cancer
At relapse, all patients received endocrine therapy as first-line treatment for metastatic breast cancer; 35 (46.7%) patients responded. We analyzed whether the expression and phosphorylation levels of hormone receptors in the primary breast tumors affected the response to endocrine therapy when given in this circumstance (Table 3). Patients with primary breast tumors that had high phosphorylation of ER-α Ser167, or high expression of ER-α (1D5), PR (636), PRA, or PRB, significantly responded to endocrine therapy (Mann-Whitney U test, P = 0.033, P = 0.0045, P = 0.0008, P = 0.0001, and P = 0.013, respectively). Interestingly, these patients with primary breast tumors with high phosphorylation or expression of the above factors also had a longer disease-free interval (compare Tables 3 and 4). In contrast, phosphorylation of ER-α Ser118 or expression of ER-β1 or ER-βcx/β2 did not affect response to endocrine therapy. In the subgroup with HER2 overexpression (n = 12), phosphorylation of ER-α Ser118 (IHC scores ≥3) was observed in all cases except one, whereas phosphorylation of ER-α Ser167 (IHC scores ≥3) was found in only six. Furthermore, endocrine therapy was effective in only three (25%) patients in the HER2-positive group (data not shown).
Correlation between expression and phosphorylation of ER-α, ER-β, and PR in primary breast tumors and disease-free interval
We then examined whether the expression and phosphorylation levels of hormone receptors in primary breast tumors affected disease-free interval in relapsing breast cancer patients. Spearman correlation coefficients between the IHC scores of ER-α, ER-β, and PR and disease-free interval are shown in Table 4. The time to relapse after primary surgery was significantly longer in patients with primary breast tumors with high phosphorylation levels of ER-α Ser167 or with high expression levels of ER-α (1D5), PR (636), PRA, or PRB (P = 0.0076, P = 0.035, P = 0.018, P = 0.0061, and P = 0.023, respectively). On the other hand, no significant relation was observed between either phosphorylation of ER-α Ser118 or expression of ER-β1 or ER-βcx/β2 and disease-free interval.
Comparison of IHC scores of ER-α, ER-β, and PR in primary and secondary tumors
Biopsy specimens were obtained from the secondary tumors of 10 patients after relapse. Six were local skin or subcutaneous tumors, two were axillar or supraclavicular lymph nodes, and two were distant (lung) tumors. Phosphorylation of ER-α Ser118 was much higher in secondary than in primary tumors (P = 0.0098) (Table 5). There was also a tendency for phosphorylation of ER-α Ser167 to increase in secondary tumors, although this did not reach significance. Furthermore, the IHC scores of ER-β1, ER-βcx/β2, PRA, and PRB were all significantly higher in secondary than in primary tumors (P = 0.041, P = 0.049, P = 0.018, and P = 0.027, respectively), while expression levels of ER-α (1D5) and PR (636) were lower in secondary than in primary tumors.
Patients with high phosphorylation of ER-α Ser167 and high expression of PRA and PRB in primary breast tumors had a significantly longer survival after relapse
Finally, we analyzed whether the expression and phosphorylation levels of hormone receptors in the primary breast tumors affected the survival after relapse. The median follow-up period was 77 months (range, 4 to 234 months). High expression of ER-α (1D5) (IHC score ≥3) significantly increased postrelapse survival (P = 0.0005) (Fig. 4a). Patients with high phosphorylation of ER-α Ser167 (IHC score ≥2) had a significant longer postrelapse survival (P = 0.03) (Fig. 4b). Similarly, high expression of PR (IHC score ≥3), PRA (IHC score ≥4), and PRB (IHC score ≥3) significantly increased postrelapse survival (P = 0.0008, P = 0.009, and P = 0.01, respectively) (Fig. 4c,d,e). Univariate analysis (Table 6) revealed significant associations between postrelapse survival and phosphorylation of ER-α Ser167 (P = 0,035), as well as expression of ER-α (1D5) (P = 0,0009), PR (636) (P = 0,0012), PRA (P = 0,03), and PRB (P = 0,013). On the other hand, phosphorylation of ER-α Ser118 and expression of ER-β1 and ER-βcx/β2 did not affect postrelapse survival. The status of phosphorylation of ER-α Ser167, expression of ER-α (1D5), and PRB were selected for the multivariate analysis because these three factors were significant in the univariate analysis and no significant association was found between IHC scores of these factors. Patients with primary tumors with high expression of ER-α (1D5) had significantly increased overall survival (P = 0.0076), whereas phosphorylation of ER-α Ser167 or expression of PRB were insignificant in the multivariate analysis (Table 5).
Discussion
Using IHC techniques, we investigated the phosphorylation of ER-α Ser118 and ER-α Ser167, and expression of ER-α, ER-β1, ER-βcx/β2, PR, PRA, and PRB, in primary breast tumor specimens from 75 patients with metastatic breast cancer who, on relapse, received endocrine therapy as first-line treatment. Our results indicate that patients with primary breast tumors with high phosphorylation of ER-α Ser167, or high expression of ER-α, PR, PRA, or PRB, significantly respond to endocrine therapy and had a better survival after relapse.
ER-α is phosphorylated on multiple amino acid residues [3]. In general, phosphorylation of serine residues in the AF1 domain of ER-α appears to influence the recruitment of coactivators, resulting in enhanced ER-mediated transcription. In this study, we measured the phosphorylation of ER-α Ser118 and Ser167 as well as the expression of ER-α in breast cancer by IHC using site-specific anti-ER-α-phosphoserine antibodies. Our results showed that phosphorylation of ER-α Ser118, but not of ER-α Ser167, was significantly correlated with expression levels of HER2. It has been reported that ER-α was significantly phosphorylated on Ser118 in response to either estradiol binding or the activation of the mitogen-activated protein kinase (MAPK) pathway, while Ser167 is phosphorylated by AKT, p90 ribosomal S6 kinase (RSK), and casein kinase II as well as MAPK [5,7,9,24]. Murphy and colleagues recently reported that in 45 human breast tumor biopsies phosphorylation of ER-α Ser118 correlated with active MAPK [25]. Because MAPK is located downstream of HER2, it is possible that phosphorylation of ER-α Ser118 is in part caused by HER2-MAPK signaling in breast cancer. On the other hand, phosphorylation of ER-α Ser167 seems to be led by different mechanisms.
Our results also showed that while phosphorylation of both ER-α Ser118 and Ser167 was strongly and positively correlated with expression of ER-β 1 and ER-βcx/β2, there was no observed association between expression of ER-α and ER-β proteins. Both antibodies for ER-β1 and ER-βcx/β2, generated in this study, were specific against their respective C-terminal amino acid residues, and positive nuclear staining was observed in normal breast epithelial cells, noninvasive ductal carcinoma, and invasive carcinoma. Saunders and colleagues also found that there was no quantitative relation between IHC scores for ER-α and ER-β [26]. However, using IHC, it was reported that ER-β expression was positively correlated with ER-α and PR [27]. Specific detection of ER-β1 from other isoforms also indicated a positive correlation between expression of ER-β1 and ER-α [28]. However, no studies have been reported concerning the relation between phosphorylation of ER-α and expression of ER-β proteins.
In our analysis, ER-α expression was positively correlated with PRA but not with PRB. In addition, phosphorylation of both ER-α Ser118 and Ser167 was strongly and positively associated with expression of PRA but not with PRB. This suggests that PRA is preferentially induced following the transcription of ER-α after the phosphorylation of Ser118 and/or Ser167. Two previous studies have reported investigations into the expression of PRA and PRB in breast cancer. The first, an analysis of 202 PR-positive breast cancers by immunoblotting, showed that while there was no significant difference in levels of PRA and PRB in most of the PR-positive tumors, nevertheless expression levels of PRA were higher in 59% of tumors and at least four times as high in 25% [29]. In the second study, of 32 PR-positive breast cancers, it was reported that excess PRB correlated with the absence of HER2, thereby indicating a good prognosis, whereas excess PRA correlated with a poorly differentiated phenotype and higher tumor grade [30]. The normally equal expression of PRA and PRB is disrupted early in carcinogenesis. PRA is usually the predominant isoform in tumors of the breast, and it appears, therefore, that disrupted progesterone signaling may play a role in the development or progression of these cancers [14,29,31].
The most important results to come out of this study concern the correlation between clinical response and the phosphorylation and expression of the receptors. We identified that patients with primary breast tumors with high phosphorylation of ER-α Ser167, or high expression of PRA or PRB, significantly responded to endocrine therapy, whereas phosphorylation of ER-α Ser118 and expression of ER-β1 and ER-βcx/β2 did not influence response. Phosphorylation of both ER-α Ser118 and ER-α Ser167 occurs in response to either estradiol binding or activation of growth factor signaling pathways. It is well established that peptide growth factor signaling pathways can activate ER-α, in the absence of its ligand, through phosphorylation of ER-α by MAPK [8,32]. In addition, the induction of ER-α by MAPK also enhances ER signaling and promotes tumor growth in the presence of estradiol, and such tumors have been shown to be responsive still to antiestrogen therapy [33]. Furthermore, Clark and colleagues reported that, independently of MAPK, p90 ribosomal S6 kinase 2 (Rsk2) specifically activates ER-α by phosphorylation of Ser167 and by docking to the hormone-binding domain of ER-α, and that the antiestrogen 4-hydroxytamoxifen blocks Rsk2-mediated activation of ER-α [7]. Since our results showed that phosphorylation of ER-α Ser167, but not ER-α Ser118, was predictive of response to endocrine therapy, they suggest that, in breast cancer, phosphorylation of ER-α Ser118 occurs frequently without estradiol, whereas phosphorylation of ER-α Ser167 may occur frequently in response to estradiol binding.
It has been reported that HER2-induced MAPK and ER-α activation leads to tamoxifen resistance [34]. Data from these clinical trials demonstrated that the antiproliferative response to endocrine therapy was impeded in ER-α-positive/HER2-positive primary breast cancers [35]. In contrast, a Southwest Oncology Group study reported that overexpression of HER2 was not associated with tamoxifen unresponsiveness or a more aggressive phenotype of ER-α-positive metastatic breast cancer [36]. In our analysis, HER2-positive tumors showed high phosphorylation levels of ER-α Ser118 and were resistant to endocrine therapy.
Finally, our results showed that expression of ER-β1 and ER-βcx/β2 does correlate with response to endocrine therapy. No significant differences in the expression of ER-β1, ER-β2, and ER-β5 mRNAs between tamoxifen-sensitive and -resistant groups, has been reported [37]. Taken together, these data suggest that the expression of ER-β proteins is not predictive of response to endocrine therapy in breast cancer. However, a significant correlation between a PR-negative phenotype and the presence of ER-βcx/β2 in ER-α-rich tumor foci and expression of ER-βcx/β2 with low PR expression has been shown to correlate with a poor response to tamoxifen [38].
In our analysis, the expression of PRA and PRB as well as PR was strongly predictive of response to endocrine therapy. In contrast, it has been reported, in a study of T47D human breast tumor xenografts, that tamoxifen treatment preferentially inhibited the growth of PRA tumors, whereas PRB tumors were unaffected [39]. Another study reported that, although estrogen induced PR expression in all breast cancer cell lines studied, the expression ratio of PRA/PRB induced by estrogen was dependent on the cell line, and that these results suggested that the PRA and PRB promoters were differentially regulated by estrogen in different breast cancer cells [40]. Further studies are obviously needed to resolve these apparent discrepancies and in order to identify the functional importance of altered PR isoform expression and how this might affect the response of breast tumors to endocrine therapy.
Conclusion
The present study has shown for the first time that patients with primary breast tumors with either high phosphorylation of ER-α Ser167 or high expression of PRA or PRB respond significantly to endocrine therapy and have a better survival after relapse. Our data suggest that phosphorylation of ER-α Ser167 is helpful in selecting patients who may benefit from endocrine therapy and is a prognostic marker in metastatic breast cancer.
Abbreviations
AF = action function; DMEM = Dulbecco's modified essential medium; E2 = 17β-estradiol; EGF = epidermal growth factor; ER = estrogen receptor; IHC = immunohistochemistry/immunohistochemical; MAPK = mitogen-activated protein kinase; PR = progesterone receptor.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
HY conceived of the study and participated in its design, coordination, and manuscript writing. MN carried out immunostaining experiments. SK, YF, and HI participated in its design and coordination and helped to draft the manuscript. YA, HS, MH, and KM provided tissue samples. ZZ assessed the immunostaining. All authors read and approved the final manuscript.
Acknowledgements
This work was supported by Grants-in Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture in Japan 16591267.
Figures and Tables
Figure 1 Immunoblot analysis of phosphorylated estrogen receptor (ER)-α Ser118 and ER-α Ser167. (a) Transfected COS-7 cells were grown in serum- and estrogen-deprived conditions and treated with vehicle (medium) (lane 1), 17β-estradiol (E2) (lane 2), epidermal growth factor (EGF) (lane 3), or E2 and EGF (lane 4) for 30 min. Equal amounts of total protein from whole cell lysates were blotted for either anti-ER-α-phosphoserine (α-pS118 and α-pS167) and anti-ER-α (α-ER-α) antibodies. (b) T47D cells were grown in serum- and estrogen-deprived conditions and treated with vehicle (medium) (lane 1), 17β-estradiol (E2) for 10 min (lane 2) and 30 min (lane 3), EGF for 10 min (lane 4) and 30 min (lane 5), or E2 and EGF for 10 min (lane 6) and 30 min (lane 7). Equal amounts of total protein from whole cell lysates were blotted for either anti-ER-α-phosphoserine (α-pS118 and α-pS167) and anti-ER-α (α-ER-α) antibodies.
Figure 2 Representative immunohistochemical staining of estrogen receptor (ER)-α Ser118, and ER-α Ser167 in normal breast epithelium and invasive ductal carcinoma. Phosphorylation of ER-α Ser118 in (a) normal breast epithelium and (b,c) invasive ductal carcinoma: negative (b) and positive (c) nuclear staining was observed in carcinoma cells. Phosphorylation of ER-α Ser167 in (d) normal breast epithelium and (e,f) invasive ductal carcinoma: negative (e) and positive (f) nuclear staining was observed in carcinoma cells (magnification, 400×).
Figure 3 Representative immunohistochemical staining of estrogen receptor (ER)-β1 (a,b), ER-βcx/β2 (c,d), progesterone receptor (PR)A (e,f), and PRB (g,h) in invasive ductal carcinoma. Negative (a,c,e,g) and positive (b,d,f,h) nuclear staining was observed in these cells.
Figure 4 Postrelapse survival of patients analyzed according to expression of (a) estrogen receptor (ER)-α, (b) phosphorylation of ER-α Ser167, and (c) expression of progesterone receptor (PR), (d) PRA, and (e) PRB. Higher levels of expression or phosphorylation of these factors were associated with a better survival.
Table 1 Clinicopathological characteristics of patients with metastatic breast cancer, their primary tumors, and treatment
Characteristic Number of patients
Total number of patients 75
Age at diagnosis (years)
≤50 35
>50 40
Range 29 to 77
Tumor size (cm)
<2.0 20
≥2.0 55
Number of positive lymph nodes
0 21
1–3 21
>3 33
Histological grade
1 12
2 43
3 20
HER2
Negative 63
Positive 12
Adjuvant therapy
None 5
Endocrine therapy 32
Chemotherapy 2
Combined 36
Disease-free interval (months)
Mean ± standard deviation 39.9 ± 26.4
Median 38
Range 1 to 123
First-line endocrine therapy
Tamoxifen 56
Aromatase inhibitors 11
LH-RH agonist 3
LH-RH agonist + tamoxifen 4
Fulvestrant 1
LH-RH agonist, luteinizing hormone–releasing hormone agonist.
Table 2 Correlations between immunohistochemistry scores for expression and phosphorylation of hormone receptors in primary breast tumors
Receptor ER-α (1D5) ER-α Ser118 ER-α Ser167 ER-β1 ER-βcx/β2 PR (636) PRA
ER-α Ser118 +0.224a
0.055b
ER-α Ser167 +0.140 +0.556
0.23 <0.0001*
ER-β1 +0.039 +0.589 +0.417
0.74 <0.0001* 0.0003*
ER-βcx/β2 +0.009 +0.640 +0.518 +0.693
0.94 <0.0001* <0.0001* <0.0001*
PR (636) +0.446 +0.032 +0.011 +0.004 -0.015
0.0001* 0.78 0.92 0.97 0.90
PRA +0.256 +0.463 +0.394 +0.203 +0.378 +0.500
0.028* <0.0001* 0.0007* 0.081 0.0011* <0.0001*
PRB +0.187 +0.211 +0.149 +0.206 +0.325 +0.553 +0.526
0.11 0.070 0.20 0.077 0.0052* <0.0001* <0.0001*
aSpearman correlation coefficient; bP, Spearman rank correlation test. * P <0.05.
ER, estrogen receptor; PR(A,B), progresterone receptor (A,B).
Table 3 Correlation between immunohistochemistry scores for hormone receptors and response to endocrine therapy in breast cancer
Receptor Responders (n = 35)a Nonresponders (n = 40)a Pb Pc
ER-α (1D5) 5.8 ± 2.3 (7; 0–8) 4.1 ± 2.9 (5; 0–8) 0.0045* 0.0046*
ER-α Ser118 4.3 ± 2.6 (5; 0–8) 4.2 ± 2.4 (4; 0–8) 0.90 0.96
ER-α Ser167 2.5 ± 2.0 (2; 0–6) 1.6 ± 1.7 (2; 0–5) 0.033* 0.025*
ER-β1 4.2 ± 2.2 (4; 0–8) 4.5 ± 2.3 (5; 0–8) 0.43 0.57
ER-βcx/β2 3.1 ± 2.4 (3; 0–8) 3.0 ± 2.5 (2; 0–8) 0.72 0.80
PR (636) 5.5 ± 2.5 (6; 0–8) 3.6 ± 2.7 (4; 0–8) 0.0008* 0.0014*
PRA 4.6 ± 2.0 (5; 0–8) 2.4 ± 2.5 (2; 0–8) 0.0001* <0.0001*
PRB 4.0 ± 2.1 (4; 0–8) 2.7 ± 2.4 (3; 0–8) 0.013* 0.015*
aValues are means ± standard deviations (medians; ranges). bMann–Whitney U test. cUnpaired t-test. *P <0.05. ER, estrogen receptor; PR(A,B), progresterone receptor (A,B).
Table 4 Correlation between immunohistochemistry scores for hormone receptors and disease-free interval
Receptor Spearman correlation coefficient Spearman rank correlation test (P)
ER-α (1D5) +0.246 0.035*
ER-α Ser118 +0.083 0.47
ER-α Ser167 +0.310 0.0076*
ER-β1 +0.039 0.74
ER-βcx/β2 +0.101 0.39
PR (636) +0.276 0.018*
PRA +0.319 0.0061*
PRB +0.263 0.023*
*P <0.05. ER, estrogen receptor; PR, progresterone receptor.
Table 5 Comparison of immunohistochemistry scores for hormone receptors in primary and secondary tumors
Receptor Primary tumorsa Secondary tumorsa P
ER-α (1D5) 4.9 ± 2.7 3.1 ± 3.2 0.049*
ER-α Ser118 4.9 ± 2.8 7.3 ± 1.6 0.0098*
ER-α Ser167 3.7 ± 1.6 5.5 ± 2.2 0.11
ER-β1 5.5 ± 2.2 7.0 ± 2.0 0.041*
ER-βcx/β2 4.1 ± 2.6 6.1 ± 2.6 0.049*
PR (636) 4.1 ± 3.2 2.9 ± 3.5 0.057
PRA 3.5 ± 2.6 6.1 ± 2.6 0.018*
PRB 3.7 ± 2.4 5.7 ± 2.6 0.027*
aMean ± SD. n = 10. *P <0.05. ER, estrogen receptor; PR(A,B), progresterone receptor (A,B).
Table 6 Univariate and multivariate analysis of factors predicting postrelapse survival
Univariate Multivariate
Factor P RR 95% CI P RR 95% CI
Age 0.67 0.882 0.494–1.576
Tumor size 0.21 1.483 0.797–2.759
Lymph node status 0.40 0.746 0.376–1.479
Histological grade 0.20 0.542 0.213–1.382
HER2 0.0033* 0.340 0.166–0.699
ER-α (1D5) 0.0009* 2.924 1.550–5.515 0.0076* 2.465 1.271–4.781
ER-α Ser118 0.36 1.308 0.732–2.340
ER-α Ser167 0.035* 1.890 1.045–3.420 0.19 1.536 0.841–2.900
ER-β1 0.37 1.355 0.697–2.635
ER-βcx/β2 0.10 1.627 0.905–2.926
PR (636) 0.0012* 2.796 1.498–5.220
PRA 0.03* 2.182 1.202–3.960
PRB 0.013* 2.157 1.177–3.951 0.24 1.499 0.763–2.945
CI, confidence interval; ER, estrogen receptor; PR, progesterone receptor; RR, relative risk. *P < 0.05.
==== Refs
Johnston SR Head J Pancholi S Detre S Martin LA Smith IE Dowsett M Integration of signal transduction inhibitors with endocrine therapy: an approach to overcoming hormone resistance in breast cancer Clin Cancer Res 2003 9 524S 532S 12538510
Schiff R Massarweh S Shou J Osborne CK Breast cancer endocrine resistance: how growth factor signaling and estrogen receptor coregulators modulate response Clin Cancer Res 2003 9 447S 4454S 12538499
Lannigan DA Estrogen receptor phosphorylation Steroids 2003 68 1 9 12475718 10.1016/S0039-128X(02)00110-1
Le Goff P Montano MM Schodin DJ Katzenellenbogen BS Phosphorylation of the human estrogen receptor. Identification of hormone-regulated sites and examination of their influence on transcriptional activity J Biol Chem 1994 269 4458 4466 8308015
Arnold SF Obourn JD Jaffe H Notides AC Serine 167 is the major estradiol-induced phosphorylation site on the human estrogen receptor Mol Endocrinol 1994 8 1208 1214 7838153 10.1210/me.8.9.1208
Castano E Vorojeikina DP Notides AC Phosphorylation of serine-167 on the human oestrogen receptor is important for oestrogen response element binding and transcriptional activation Biochem J 1997 326 149 157 9337862
Clark DE Poteet-Smith CE Smith JA Lannigan DA Rsk2 allosterically activates estrogen receptor alpha by docking to the hormone-binding domain EMBO J 2001 20 3484 3494 11432835 10.1093/emboj/20.13.3484
Kato S Endoh H Masuhiro Y Kitamoto T Uchiyama S Sasaki H Masushige S Gotoh Y Nishida E Kawashima H Activation of the estrogen receptor through phosphorylation by mitogen-activated protein kinase Science 1995 270 1491 1494 7491495
Joel PB Smith J Sturgill TW Fisher TL Blenis J Lannigan DA pp90rsk1 regulates estrogen receptor-mediated transcription through phosphorylation of Ser-167 Mol Cell Biol 1998 18 1978 1984 9528769
Leygue E Dotzlaw H Watson PH Murphy LC Expression of estrogen receptor beta1, beta2, and beta5 messenger RNAs in human breast tissue Cancer Res 1999 59 1175 1179 10096542
Palmieri C Cheng GJ Saji S Zelada-Hedman M Warri A Weihua Z Van Noorden S Wahlstrom T Coombes RC Warner M Estrogen receptor beta in breast cancer Endocr Relat Cancer 2002 9 1 13 11914179 10.1677/erc.0.0090001
Speirs V Oestrogen receptor beta in breast cancer: good, bad or still too early to tell? J Pathol 2002 197 143 147 12015736 10.1002/path.1072
Ogawa S Inoue S Watanabe T Orimo A Hosoi T Ouchi Y Muramatsu M Molecular cloning and characterization of human estrogen receptor betacx: a potential inhibitor of estrogen action in human Nucleic Acids Res 1998 26 3505 3512 9671811 10.1093/nar/26.15.3505
Graham JD Clarke CL Expression and transcriptional activity of progesterone receptor A and progesterone receptor B in mammalian cells Breast Cancer Res 2002 4 187 190 12223122 10.1186/bcr450
Shyamala G Yang X Silberstein G Barcellos-Hoff MH Dale E Transgenic mice carrying an imbalance in the native ratio of A to B forms of progesterone receptor exhibit developmental abnormalities in mammary glands Proc Natl Acad Sci USA 1998 95 696 701 9435255 10.1073/pnas.95.2.696
Shyamala G Yang X Cardiff RD Dale E Impact of progesterone receptor on cell-fate decisions during mammary gland development Proc Natl Acad Sci USA 2000 97 3044 3049 10737785 10.1073/pnas.97.7.3044
Yamashita H Xu J Erwin RA Farrar WL Kirken RA Rui H Differential control of the phosphorylation state of proline-juxtaposed serine residues Ser725 of Stat5a and Ser730 of Stat5b in prolactin-sensitive cells J Biol Chem 1998 273 30218 30224 9804779 10.1074/jbc.273.46.30218
Yamashita H Iwase H Toyama T Fujii Y Naturally occurring dominant-negative Stat5 suppresses transcriptional activity of estrogen receptors and induces apoptosis in T47D breast cancer cells Oncogene 2003 22 1638 1652 12642867 10.1038/sj.onc.1206277
Yamashita H Nevalainen MT Xu J LeBaron MJ Wagner KU Erwin RA Harmon JM Hennighausen L Kirken RA Rui H Role of serine phosphorylation of Stat5a in prolactin-stimulated beta-casein gene expression Mol Cell Endocrinol 2001 183 151 163 11604235 10.1016/S0303-7207(01)00546-9
Chen D Washbrook E Sarwar N Bates GJ Pace PE Thirunuvakkarasu V Taylor J Epstein RJ Fuller-Pace FV Egly JM Phosphorylation of human estrogen receptor alpha at serine 118 by two distinct signal transduction pathways revealed by phosphorylation-specific antisera Oncogene 2002 21 4921 4931 12118371 10.1038/sj.onc.1205420
Yamashita H Nishio M Toyama T Sugiura H Zhang Z Kobayashi S Iwase H Coexistence of HER2 over-expression and p53 protein accumulation is a strong prognostic molecular marker in breast cancer Breast Cancer Res 2004 6 R24 R30 14680497 10.1186/bcr738
Clarke CL Zaino RJ Feil PD Miller JV Steck ME Ohlsson-Wilhelm BM Satyaswaroop PG Monoclonal antibodies to human progesterone receptor: characterization by biochemical and immunohistochemical techniques Endocrinology 1987 121 1123 1132 3304978
Allred DC Harvey JM Berardo M Clark GM Prognostic and predictive factors in breast cancer by immunohistochemical analysis Mod Pathol 1998 11 155 168 9504686
Campbell RA Bhat-Nakshatri P Patel NM Constantinidou D Ali S Nakshatri H Phosphatidylinositol 3-kinase/AKT-mediated activation of estrogen receptor alpha: a new model for anti-estrogen resistance J Biol Chem 2001 276 9817 9824 11139588 10.1074/jbc.M010840200
Murphy L Cherlet T Adeyinka A Niu Y Snell L Watson P Phospho-serine-118 estrogen receptor-alpha detection in human breast tumors in vivo Clin Cancer Res 2004 10 1354 1359 14977836
Saunders PT Millar MR Williams K Macpherson S Bayne C O'Sullivan C Anderson TJ Groome NP Miller WR Expression of oestrogen receptor beta (ERbeta1) protein in human breast cancer biopsies Br J Cancer 2002 86 250 256 11870515 10.1038/sj.bjc.6600035
Fuqua SA Schiff R Parra I Moore JT Mohsin SK Osborne CK Clark GM Allred DC Estrogen receptor beta protein in human breast cancer: correlation with clinical tumor parameters Cancer Res 2003 63 2434 2439 12750263
Omoto Y Inoue S Ogawa S Toyama T Yamashita H Muramatsu M Kobayashi S Iwase H Clinical value of the wild-type estrogen receptor beta expression in breast cancer Cancer Lett 2001 163 207 212 11165756 10.1016/S0304-3835(00)00680-7
Graham JD Yeates C Balleine RL Harvey SS Milliken JS Bilous AM Clarke CL Characterization of progesterone receptor A and B expression in human breast cancer Cancer Res 1995 55 5063 5068 7585552
Bamberger AM Milde-Langosch K Schulte HM Loning T Progesterone receptor isoforms, PR-B and PR-A, in breast cancer: correlations with clinicopathologic tumor parameters and expression of AP-1 factors Horm Res 2000 54 32 37 11182633 10.1159/000063434
Mote PA Bartow S Tran N Clarke CL Loss of co-ordinate expression of progesterone receptors A and B is an early event in breast carcinogenesis Breast Cancer Res Treat 2002 72 163 172 12038707 10.1023/A:1014820500738
Bunone G Briand PA Miksicek RJ Picard D Activation of the unliganded estrogen receptor by EGF involves the MAP kinase pathway and direct phosphorylation EMBO J 1996 15 2174 2183 8641283
Atanaskova N Keshamouni VG Krueger JS Schwartz JA Miller F Reddy KB MAP kinase/estrogen receptor cross-talk enhances estrogen-mediated signaling and tumor growth but does not confer tamoxifen resistance Oncogene 2002 21 4000 4008 12037682 10.1038/sj.onc.1205506
Kurokawa H Lenferink AE Simpson JF Pisacane PI Sliwkowski MX Forbes JT Arteaga CL Inhibition of HER2/neu (erbB-2) and mitogen-activated protein kinases enhances tamoxifen action against HER2-overexpressing, tamoxifen-resistant breast cancer cells Cancer Res 2000 60 5887 5894 11059787
Dowsett M Harper-Wynne C Boeddinghaus I Salter J Hills M Dixon M Ebbs S Gui G Sacks N Smith I HER-2 amplification impedes the antiproliferative effects of hormone therapy in estrogen receptor-positive primary breast cancer Cancer Res 2001 61 8452 8458 11731427
Elledge RM Green S Ciocca D Pugh R Allred DC Clark GM Hill J Ravdin P O'Sullivan J Martino S HER-2 expression and response to tamoxifen in estrogen receptor-positive breast cancer: a Southwest Oncology Group Study Clin Cancer Res 1998 4 7 12 9516946
Murphy LC Leygue E Niu Y Snell L Ho SM Watson PH Relationship of coregulator and oestrogen receptor isoform expression to de novo tamoxifen resistance in human breast cancer Br J Cancer 2002 87 1411 1416 12454770 10.1038/sj.bjc.6600654
Saji S Omoto Y Shimizu C Warner M Hayashi Y Horiguchi S Watanabe T Hayashi S Gustafsson JA Toi M Expression of estrogen receptor (ER) (beta)cx protein in ER(alpha)-positive breast cancer: specific correlation with progesterone receptor Cancer Res 2002 62 4849 4853 12208729
Sartorius CA Shen T Horwitz KB Progesterone receptors A and B differentially affect the growth of estrogen-dependent human breast tumor xenografts Breast Cancer Res Treat 2003 79 287 299 12846413 10.1023/A:1024031731269
Vienonen A Syvala H Miettinen S Tuohimaa P Ylikomi T Expression of progesterone receptor isoforms A and B is differentially regulated by estrogen in different breast cancer cell lines J Steroid Biochem Mol Biol 2002 80 307 313 11948015 10.1016/S0960-0760(02)00027-4
|
16168121
|
PMC1242143
|
CC BY
|
2021-01-04 16:04:33
|
no
|
Breast Cancer Res. 2005 Jul 27; 7(5):R753-R764
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1285
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12901616812210.1186/bcr1290Research ArticleProteomic identification of heat shock protein 90 as a candidate target for p53 mutation reactivation by PRIMA-1 in breast cancer cells Rehman Abdur [email protected] Manpreet S [email protected] Xiaoting [email protected] James E [email protected] Yves [email protected] Sayed S [email protected] Department of Pharmaceutical Sciences, Washington State University, Pullman, WA, USA2 Pharmacology and Toxicology Graduate Program, Washington State University, Pullman, WA, USA3 Department of Chemistry, Washington State University, Pullman, WA, USA4 Laboratory of Molecular Pharmacology, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA2005 27 7 2005 7 5 R765 R774 27 7 2004 7 10 2004 2 5 2005 29 6 2005 Copyright © 2005 Rehman et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
A loss of p53 function resulting from mutation is prevalent in human cancers. Thus, restoration of p53 function to mutant p53 using small compounds has been extensively studied for cancer therapy. We previously reported that PRIMA-1 (for 'p53 reactivation and induction of massive apoptosis') restored the transcriptional activity of p53 target genes in breast cancer cells with a p53 mutation. By using functional proteomics approach, we sought to identify molecular targets that are involved in the restoration of normal function to mutant p53.
Methods
PRIMA-1 treated cell lysates were subjected to immunoprecipitation with DO-1 primary antibody against p53 protein, and proteins bound to p53 were separated on a denaturing gel. Bands expressed differentially between control and PRIMA-1-treated cells were then identified by matrix-assisted laser desorption ionization-time-of-flight spectrometry. Protein expression in whole cell lysates and nuclear extracts were confirmed by Western blotting. The effect of combined treatment of PRIMA-1 and adriamycin in breast cancer cells was determined with a cytotoxicity assay in vitro.
Results
PRIMA-1 treated cells distinctly expressed a protein band of 90 kDa that was identified as heat shock protein 90 (Hsp90) by the analysis of the 90 kDa band tryptic digest. Immunoblotting with isoform-specific antibodies against Hsp90 identified this band as the α isoform of Hsp90 (Hsp90α). Co-immunoprecipitation with anti-Hsp90α antibody followed by immunoblotting with DO-1 confirmed that p53 and Hsp90α were interacting proteins. PRIMA-1 treatment also resulted in the translocation of Hsp90α to the nucleus by 8 hours. Treatment of cells with PRIMA-1 alone or in combination with adriamycin, a DNA-targeted agent, resulted in increased sensitivity of tumor cells.
Conclusion
The studies demonstrate that PRIMA-1 restores the p53-Hsp90α interaction, enhances the translocation of the p53-Hsp90α complex and reactivates p53 transcriptional activity. Our preliminary evidence also suggests that PRIMA-1 could be considered in combination therapy with DNA-targeted agents for the treatment of breast cancer, especially for tumors with aberrant p53 function.
==== Body
Introduction
Many clinical studies have shown that mutations in p53 are a strong predictor of relapse and are associated with resistance to several therapeutic regimens [1,2]. Studies in our laboratories and others, for example, showed that mutations in p53 in human tumor cells were correlated with decreased sensitivity to DNA-damaging agents [3-6]. An improved understanding of the relationship between p53 and chemosensitivity might therefore lay the groundwork for new cancer therapies. To understand this relationship better, we recently used a pharmacogenomic approach with complementary DNA microarrays to characterize gene expression profiles of cells containing wild-type p53 (p53+/+) and those containing an isogenic p53 knockout counterpart (p53-/-) after treatment with topotecan, a specific topoisomerase I inhibitor and a DNA-targeted agent [7]. About 10% of the transcripts detected were differentially expressed in the p53+/+ cells in response to topotecan, whereas only 1% of the transcripts changed in the p53-/- cells [7]. These data clearly showed the broad effect of p53 on the transcriptional response to DNA damage, which can lead to growth arrest or apoptosis.
Given that p53 is the most commonly mutated gene in human cancers and that more than 50% of breast tumors are defective in p53 [8-10], extensive research efforts are centered on restoring normal function to mutant p53 to promote tumor suppression. This effort includes the use of modifying peptides [11,12], antisense oligonucleotides [13] and small molecules [14,15]. Unfortunately, the problem of in vivo delivery and lack of selectivity to tumor cells has limited the practical application of most of these efforts. Recently, PRIMA-1 has emerged from an in vitro screen of small molecules that reactivate the transcriptional activity of mutant p53 [16]. PRIMA-1 has the capability of restoring the transcriptional transactivation function to mutant p53 in vitro and in vivo with subsequent tumor regression. PRIMA-1 also has the ability to trigger apoptosis in tumor cells as a function of its mutant p53 reactivation response [17]. We reported recently [18] that PRIMA-1 (in a effect dependent on both dose and time) restored the transcriptional activity of p53 target genes such as p21Waf1/cip1 in breast cancer cells possessing a p53 mutation. However, the exact molecular mechanisms for mutant p53 reactivation by PRIMA-1 are not yet determined. A direct interaction between PRIMA-1 and p53 has not yet been demonstrated. It is possible that PRIMA-1 affects cellular chaperones, resulting in the refolding of mutant p53. Alternatively, PRIMA-1 may block complex formation between mutant p53 and p73, leading to the release of active p73, which triggers proapoptotic target genes [15]. To identify the possible molecular candidates of mutant p53 reactivation by PRIMA-1 in breast tumor cells, in this study we used tools available for a functional proteomics approach.
Our study indicates that the restoration of the transcriptional transactivation of p53 target genes such as p21Waf1/cip1 is dependent on p53 and that the α isoform of heat shock protein 90 (Hsp90α) is associated with mutant p53 reactivation by PRIMA-1. As a result of refolding of mutant p53 by Hsp90, we show that both p53 and Hsp90α are translocated to the nucleus of the tumor cells for the activation of p53 target genes. The use of PRIMA-1 to reactivate mutant p53 may therefore be considered further as an approach for adjuvant chemotherapy in the treatment of breast tumors, especially in cancers with aberrant p53 function.
Materials and methods
Drugs and materials
PRIMA-1 (NSC-281668) was obtained from the Drug Synthesis and Chemistry Branch, National Cancer Institute (Bethesda, MD); pifithrin-α (PFTα) and doxorubicin (adriamycin hydrochloride) were purchased from Biomol Research Inc. (Plymouth Meeting, PA). Primary antibodies against p53 and p21 were purchased from Santa Cruz Biotechnology (Santa Cruz, CA); Hsp 90 primary antibodies were from Stressgen (Victoria, BC, Canada). The goat anti-mouse, anti-rat and anti-rabbit secondary antibodies labeled with IRDye™ 38 were purchased from LI-COR, Inc. Biosciences (Lincoln, NE) or with Alex680 from Molecular Probes, Inc. (Eugene, OR). All other chemicals were of reagent grade.
Cell and culture conditions
The human MCF-7 breast carcinoma cells (p53+/+), MDA-MB-231 and GI-101A (p53 mutant) were routinely maintained in monolayer cultures in RPMI-1640 medium (Invitrogen, Inc., Carlsbad, CA) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT) as reported previously [3]. As we reported previously [18], p53 protein in GI-101A cells contains mutations at Y236C, A278P and R72P, whereas p53 protein in MDA-231 cells contains mutations at A278P, R280K and M385T. Exponentially growing cultures at 80% confluence were used in all experiments. For cytotoxicity assays, cells from exponentially growing cultures were plated in 24-well tissue culture plates in RPMI-1640 medium plus 10% fetal calf serum at about 104 cells per well, and the IC50 doses of PRIMA-1 in MDA-231 and GI-101A cells were determined as reported previously [18]. For drug combination studies, we used 100 μM PRIMA-1 and 0.2 μM adriamycin. Both cell lines were treated with the following drug sequence: A3, cells were treated with adriamycin for 3 hours; A24, cells were treated with adriamycin for 24 hours; P24, cells were treated with PRIMA-1 for 24 hours; AP24, cells were treated with both adriamycin and PRIMA-1; and A3P24, cells were treated with adriamycin for 3 hours followed by the removal of adriamycin-containing medium, washing with fresh medium and then incubation with PRIMA-1 for 24 hours. Each treatment was performed in quadruplicate wells in two independent experiments. The wells were washed twice with prewarmed medium, followed by the addition of 2 ml of drug-free medium. The plates were then incubated for a further 3 days. The surviving fraction of cells was determined with the crystal violet assay, as reported previously [19]. The precision of this method with quadruplicate determinations is 10% (SD). The data were analyzed with Student's two-tailed t-test; P < 0.01 was considered statistically significant.
Immunoblotting and immunoprecipitation
Cells were incubated with 0 or 100 μM PRIMA-1 for 0, 2, 4 or 8 hours, then washed twice with PBS (pH 7.4) and harvested. Harvested cells were either used to prepare whole cell lysates or for subcellular fractionation studies. Cell lysates were prepared as described previously [7]. For immunoblotting, 20 μg protein samples were separated by SDS-PAGE (4 to 20% polyacrylamide gradient gel) and transferred on nitrocellulose (Millipore, Bedford, MA). The membrane was developed in accordance with a protocol provided by LI-COR, Inc. Biosciences using anti-β-actin, anti-p53 and anti-p21 primary antibodies (Santa Cruz Biotechnology) or anti-Hsp90α and anti-Hsp90β primary antibodies (Stressgen). The goat anti-mouse, anti-rat and anti-rabbit secondary antibodies labeled with IRDye™ 38 (ex/em: 774/800 nm; LI-COR) or with Alexafluor® 680 (ex/em: 680/707 nm; Molecular Probes) were used. The reactive bands were revealed and detected with the Odyssey™ Infrared Imaging System (LI-COR, Inc.).
For immunoprecipitation studies, 107 MDA-MB-231 or GI-101A cell cultures were grown in T-150 tissue culture flasks. Cultures were treated with 100 μM PRIMA-1 for 4 hours. Cleared cell lysates (total 1 mg of protein) were immunoprecipitated either with 5 μg of DO-1 anti-p53 monoclonal antibody or 5 μg of anti-Hsp90α monoclonal antibody in a co-immunoprecipitation assay in accordance with the protocol provided by eBiosciences, Inc. (San Diego, CA) using Protein A or G-plus agarose beads (Santa Cruz Biotechnology). After SDS-PAGE (4 to 20% polyacrylamide), gels were either stained with Coomassie blue (Bio-Rad, Inc., Hercules, CA) or subjected to immunoblotting.
In-gel enzymatic digestion and mass spectrometry
The protein bands were excised from the one-dimensional Coomassie blue-stained polyacrylamide gel shown in Fig. 1. After reduction and alkylation the protein bands were dehydrated in acetonitrile and dried. The bands were digested in the gel with an excess of sequencing-grade trypsin (Promega, Madison, WI). The digestion was performed overnight at 37°C. The resulting tryptic peptides were extracted from the gels, then desalted and concentrated with C18 ziptip (Millipore) before spotting for matrix-assisted laser desorption ionization-time-of-flight (MALDI-TOF) analysis.
An OmniFlex MALDI-TOF mass spectrometer (Bruker Daltonics, Billerica, MA) was used for peptide mass fingerprinting. Desalted peptide solution (1 μl) was mixed with 1 μl of matrix solution (α-cyano-4-hydroxycinnamic acid in 0.1% trifluoroacetic acid and 50% acetonitrile) and was then spotted directly on the MALDI target. All data used for protein identification and calibration were acquired in reflectron mode and averaged for 100 shots. These studies employed external mass calibration and used angiotensin I, corticotropin (ACTH) clip (1 to 17) and ACTH clip (18 to 39) peptides as calibrants. After spectral acquisition, the peak m/z values were extracted and used to search the Swiss Protein Database with the ProFound program [20] search engine. The database search was performed with an implementation of the following search parameters: taxonomy (Homo sapiens), enzyme (trypsin), missing cleavage, mass tolerance (0.24 Da), modification (carbamidomethyl for cysteine and oxidation for methionine) and database (NCBRnr).
Subcellular fractionation and immunocytochemistry
To determine the nuclear translocation of p53 and Hsp90α, cells treated with PRIMA-1 were subjected to nuclear isolation with the FOCUS cytoplasmic and nuclear protein extraction kit (Geno Technology, Inc., St. Louis, MO) in accordance with the manufacturer's instructions. A fraction of the isolated nuclear pellet was fixed in 4% paraformaldehyde in 1 × PBS to determine the intactness of nuclei by staining with 4',6-diamidino-2-phenylindole (DAPI). Fixed nuclei were mounted on SuperFrost-plus slides (Fisher Scientific, Pittsburgh, PA) with Prolong Gold antifade reagent with DAPI (Molecular Probes, Invitrogen Detection Technologies, Carlsbad, CA). Our nuclear preparation contained 99% intact nuclei. To determine the expression of p53, and Hsp90s in intact nuclear fractions, the nuclear pellet was lysed into 1 × SDS sample buffer and subjected to denaturing gel electrophoresis followed by immunoblotting as described above.
To confirm translocation of proteins into the nucleus, nuclear fractions were subjected to immunostaining for p53 and Hsp90α. Immunocytochemistry was performed in accordance with a previously described protocol, with modifications [21]. In brief, fixed nuclear fractions were suspended in antibody buffer (0.1 M PBS, pH 7.4, containing 0.1% Triton X-100) and incubated overnight at 4°C with mouse monoclonal anti-p53 (DO-1) and rat monoclonal anti-Hsp90α antibody at 50:1 dilution. In addition, all appropriate negative controls without the primary antibodies were run in parallel. Unbound antibodies were removed by centrifuging nuclear preparations at 800 g for 3 min at room temperature (22–24°C) followed by three subsequent 5 min washes with shaking (300 r.p.m.) in antibody incubating buffer at room temperature. Next, the bound primary antibody was detected by using species-specific secondary antibody conjugated with different fluorescent dyes at 100:1 dilution in antibody incubation buffer at room temperature for 1 hour. The mouse IgG secondary antibody conjugated with Oregon green was used to detect mouse monoclonal antibody against p53 (DO-1) and the rat IgG secondary antibody conjugated with Texas red was used to detect rat monoclonal antibody against Hsp90α (Molecular Probes, Invitrogen Detection Technologies). The unbound secondary antibody was removed by using three washes as described earlier for primary antibody. After three washes, nuclear fractions were diluted in antibody incubating buffer without Triton X-100 and spread onto a SuperFrost-plus slides. The air-dried slides were then coverslipped with medium containing the nuclear stain DAPI, as before. The immunofluorescence for DAPI, Texas Red and Oregon green was detected with an Axioplan 2 epifluorescent microscope (Zeiss, Thornwood, NY) equipped with appropriate filters. Individual nuclei were visualized at × 20 magnification.
Results and discussion
Inhibition of transcriptional reactivation function of p53 with PFTα
We previously reported that PRIMA-1 restored the transcriptional activity of p53 target genes, such as p21Waf1/cip1, in breast cancer cells [18]. The IC50 values for PRIMA-1 in MDA-231, GI-101A and MCF-7 cells were determined as 141, 51 and 122 μM, respectively. We selected p21Waf1/cip1 as a marker for measuring the extent of the transcriptional reactivation of p53 by PRIMA-1 at both mRNA and protein levels, because the promoter of the p21 gene exhibits high occupancy to wild-type p53 protein on its p53 binding sites, in vivo; it is therefore considered a benchmark for p53-dependent genes [22]. However, p21Waf1/cip1 can also be transcriptionally regulated by p53-independent mechanisms [23,24]. To determine whether the expression of p21Waf1/cip1 is dependent on the restored transcriptional function of p53, cells were treated with PRIMA-1 in the presence and in the absence of PFTα. PFTα is a small molecule that was isolated for its ability to block p53-dependent transcriptional activation [25]. As shown in Fig. 2, treatment of GI-101A cells (mut p53) with 100 μM PRIMA-1 induced the expression of p21Waf1/cip1 at 4 hours. However, treatment of these cells with PRIMA-1 in the presence of 20 μM PFTα resulted in an inhibition of p21Waf1/cip1 expression. No change in the level of p53 protein was observed under these conditions. In contrast, no change in the expression of p21Waf1/cip1 protein was observed when MCF-7 cells (wild-type p53) were treated with PRIMA-1 in the presence of PFTα, confirming the specificity of action of PFTα as an inhibitor of p53-dependent transcriptional function. The lack of inhibition of p21 expression in MCF-7 cells after treatment with PFTα suggests that there is p53-independent expression of p21 in these cells or that MCF-7 cells is not sensitive to the dosage of PFTα used in our studies. Furthermore, the data also show that mutant p53 reactivation by PRIMA-1 results in the transcriptional activation of p53 target genes such as p21Waf1/cip1. However, the exact molecular mechanisms by which this activation occurred are not yet determined. Identification of the molecular targets that are involved in mutation reactivation of p53 by PRIMA-1 is essential for understanding the molecular mechanisms for p53 mutation reactivation and for devising therapeutic strategies aimed at enhancing the use of PRIMA-1 in cancer therapy. It is conceivably possible, for example, that PRIMA-1 affects cellular chaperones resulting in the refolding of mutant p53. In an attempt to identify possible molecular targets involved in mutation reactivation of p53 by PRIMA-1, we used a functional proteomics approach in which cell lysates were co-immunoprecipitated with DO-1 primary antibody directed against p53 after treatment with PRIMA-1 followed by protein partner identification with MALDI-TOF mass spectrometry.
Identification of Hsp90 as a candidate target for p53 mutation reactivation
Figure 1 shows a Coomassie blue-stained gel of proteins co-immunoprecipitated with DO-1 antibody from MDA-MB-231 cells (mut p53) after treatment with 100 μM PRIMA-1 for 4 hours. We chose to resolve proteins by one-dimensional electrophoresis because we were able to observe clearly and reproducibly the separation of protein mixtures, especially that of proteins smaller than 100 kDa. Single bands of polyacrylamide gel slices from SDS-PAGE that are differentially expressed after treatment with PRIMA-1 were excised and subjected to in-gel digestion by trypsin. After digestion, a small portion of the supernatant was removed and analyzed by high-accuracy peptide mass mapping with MALDI. The peptide masses obtained by MALDI analysis were used to search protein databases. We routinely observe peaks that correspond to keratin, presumably resulting from contamination during the handling of gel slices. Such peaks were excluded from our analysis and were occasionally used for internal calibration. Figure 1 focuses on a single band (90 kDa) that is marked by an arrow because it is one of the bands that is distinctly expressed in MDA-MB-231 cells after treatment with PRIMA-1 and co-immunoprecipitated with DO-1 monoclonal antibody. The tryptic digests of the excised bands were subjected to mass fingerprinting with the MALDI-TOF mass spectrum as reported previously [26]. Figure 3 illustrates the MALDI mass spectrum acquired from the peptide mixture resulting from in-gel digestion of the 90 kDa bands shown in Fig. 1. On average, six peptide masses were identified in this spectrum that agreed with the expected peptide masses within a mass tolerance of ± 0.24 Da. The peak m/z values were used to search the SwissProt Database with the ProFound program. This search resulted in the identification of Hsp90 (Klenow fragment) with a probability score of 1.0 and Z score of 1.8, a strong identification of this protein. The sequence coverage of the matched protein candidate (Klenow fragment) was 27%. These results were repeated with multiple co-immunoprecipitation experiments, which all resulted in the identification of Hsp90α as a candidate protein that is differentially expressed after treatment of cells with PRIMA-1.
Because the data collected by peptide mass fingerprinting are sometimes insufficient for the reliable identification of a protein, we further validated and confirmed the identity of Hsp90 protein by immunoblotting analyses. Cells were treated with 100 μM PRIMA-1 for 2, 4 or 8 hours, and protein samples from the whole cell extracts (WCE) and nuclear extracts (NE) were Western blotted with antibodies directed against both the α and β forms of Hsp90 protein. In mammalian cells there are at least two Hsp90 isoforms, Hsp90α and Hsp90β, which are encoded by separate genes. The amino acid sequences of human and yeast Hsp90α are 85% and 90% homologous to Hsp90β, respectively [27]. All known members of the Hsp90 protein family are highly conserved, especially in the amino-terminal and carboxy-terminal regions that contain independent chaperone sites with different substrate specificity [28,29]. To confirm the interaction of p53 with Hsp90α, we performed a co-immunoprecipitation assay with monoclonal Hsp90α antibody on control cells and PRIMA-1-treated cells and subjected the immunoprecipitated proteins to immunoblotting with monoclonal antibody against p53 protein (DO-1). As shown in Fig. 4, both MDA-231 and GI-101A cells exhibited interaction of p53 with Hsp90α protein. The data indicate that p53 interacts with Hsp90α in both breast cancer cells. However, the p53 and Hsp90α protein-protein interaction is different in both cell lines. For example, GI-101A cells show no noticeable changes in protein-protein interaction after treatment with PRIMA-1 (Fig. 4b), whereas more Hsp90α is bound to p53 in MDA-231 cells after treatment with PRIMA-1 (Fig. 4a). This may reflect the phenotype of cells as well as differences in p53 mutations and polymorphism [16].
Nuclear translocation of Hsp90α after treatment with PRIMA-1
The Western blots in Fig. 5a show that both isoforms of Hsp90 are expressed in both cell lines, although in different amounts. The α-isoform is expressed in greater amounts than the β-isoform in samples obtained from both WCE and NE. In addition, Hsp90α was detected in the NE of protein samples obtained from both cell lines. The level of Hsp90α in the NE was increased after treatment of cells with PRIMA-1 for 8 hours. In contrast, the Hsp90β isoform was not detected in the NE of protein samples obtained from either cell line, suggesting that only the α isoform of Hsp90 is translocated to the nucleus after treatment with 100 μM PRIMA-1. The exact mechanism(s) for this differential translocation are not clear yet. However, the intracellular localization of Hsp90 has been investigated under normal and heat-stressed conditions. After heat stress, an increased amount of Hsp90α was detected in the nucleus of cells compared to unstressed cells[30,31]. For nuclear localization to occur, the nuclear localization signal (NLS) is usually required. For Hsp90α, as well as for many other proteins, the nuclear import is recognized by an NLS receptor (a heterodimeric complex) that is composed of importin α and β subunits [32-34]. It has been shown that, under in vitro conditions, Hsp90α interacts with importin α [31].
It is also possible that the nuclear transport of Hsp90α after treatment with PRIMA-1 constitutes part of a selective delivery of restored conformation of p53 to the nucleus. The nuclear transport of the wild-type p53 is also known to be dependent on NLS systems [34]. We therefore investigated whether the observed nuclear accumulation of Hsp90α after treatment with PRIMA-1 was correlated with any changes in p53 protein expression and/or localization. Figure 5b shows the nuclear accumulation of p53 after treatment of cells with 100 μM PRIMA-1 for 2, 4 or 8 hours. The ratio of the integrated absorbance of p53 bands to that of the actin band was used as an index of protein expression in both WCE and NE. It is clearly shown in Fig. 5b that the levels of p53 protein in the NE are much higher after the treatment of both cell lines with PRIMA-1, especially at 8 hours, indicating that the observed increase of Hsp90α protein levels in the NE (Fig. 5a) is correlated with that of p53 protein nuclear accumulation. We observed no change in the level of p53 protein expression or localization when MCF-7 cells were treated with PRIMA-1 (data not shown). These data therefore suggest that the Hsp90 protein is a binding partner to p53 after its conformational restoration by PRIMA-1 and that the α-isoform is involved in the nuclear accumulation of the active form of p53 protein. The enhanced nuclear translocation of Hsp90α was also investigated with immunocytochemistry. Figure 6 shows that Hsp90α is selectively localized in the nucleus of MDA-231 breast cancer cells treated with PRIMA-1 for 8 hours.
Many studies showed that the Hsp90 protein has a major role in the stability of p53 protein and in its translocation to the nucleus. King and colleagues [35] reported that the wild-type p53 protein forms a complex with Hsp90 in the presence of Hop and that Hsp90 may assist p53 import into the nucleus. However, the type of Hsp90 isoform that is actually involved in the nuclear translocation of p53 was not mentioned in that study. Chen and Wang [36] recently reported that the stabilization of p53 conformation after heat shock is associated with its binding to Hsp90 protein and that the phosphorylation of p53 is dependent on the Ataxia-telangiectasia-mutated protein kinase (ATM)-mediated activation of human checkpoint 2 (ChK2) kinase. Although these studies investigated the role of Hsp90 in regulating p53 stability under stressful conditions (heat shock), the present studies confirmed the association between Hsp90 and p53 under non-stressful conditions, because there was no increase in Hsp90 expression in WCE in PRIMA-1-treated cells (Fig. 5b), although a small decrease in WCE of PRIMA-1-treated samples at 8 hours was concomitant with an increase in nuclear translocation of Hsp90α in both cell lines.
Recently, Walerych and colleagues [37] reported that Hsp90α interacts with wild-type p53 protein and that this interaction facilitates the binding of Hsp90α to the p21 promoter in ATP-dependent manner. Murphy and colleagues [38] showed that PFTα inhibits the p53 signaling after interaction with Hsp90α. Thus, both reports [37,38] suggest that the interaction of p53 and Hsp90α enhances the transcriptional transactivation of genes containing p53-binding sites in their promoter region. Our findings can be interpreted as the restoration of the p53-Hsp90α interaction by PRIMA-1, enhancing the nuclear translocation of p53-Hsp90α and reactivating the transcriptional activity of p53.
Sensitization of breast cancer cells to DNA targeted agents with PRIMA-1
We next examined whether reactivation of the p53 mutation by PRIMA-1 would enhance the cytotoxicity of DNA-damaging agents such as adriamycin in vitro. Because many studies have shown that p53 mutations are associated with a decreased sensitivity of tumor cells to many chemotherapeutic agents [3-6,39], it seems reasonable that PRIMA-1, which restores p53 transcriptional activity and enhances its nuclear translocation, would increase the sensitivity of these cells to DNA-targeted agents. Figure 7 shows that simultaneous treatment of human breast cancer MDA-231 and GI-101A cells in vitro with 100 μM PRIMA-1 and 0.2 μM adriamycin for 24 hours, or sequential treatment with adriamycin for 3 hours followed by PRIMA-1 for 24 hours, significantly enhances the sensitivity of tumor cells to the drug combination compared with the use of each drug alone. On close examination of the percentage of cell survival after drug treatment of MDA-231 cells, it is clearly shown that adriamycin treatment for 3 hours (A3) had 95% cell survival, compared with only 30% in the presence of PRIMA-1 (A3P24). Treatment of these cells for 24 hours with PRIMA-1 alone produced 75% cell survival. These data suggest that PRIMA-1 enhances adriamycin efficacy and that the drug combination, in this case, is synergistic or supra-additive. Preliminary evidence from our laboratory indicates that the drug combination is also synergistic on a panel of breast cancer cell lines (Tao Wang and S.S.D., unpublished work).
Conclusion
This study illustrates the use of a functional proteomics approach to identify target molecules that are associated with the reactivation of a p53 mutation by PRIMA-1. Our approach has identified Hsp90α as a partner protein that is associated, in part, with the restoration of p53 transcriptional transactivation function by PRIMA-1. We also showed that the α isoform of Hsp90 protein is associated with the nuclear translocation of p53 protein after treatment of breast tumor cells with PRIMA-1. Our results that PFTα inhibits the induction of p21 expression after treatment with PRIMA-1, and the recent work of Murphy and colleagues [38] indicating that PFTα inhibits the signaling of p53-mediated gene transcription, strongly suggest that PRIMA-1 facilitates the interaction of Hsp90α and the restoration of mutant p53 to wild type followed by the translocation of both proteins into the nucleus. This may result in the transactivation of genes containing p53-binding sites as reported by Walerych and colleagues [37].
Abbreviations
DAPI = 4',6-diamidino-2-phenylindole; Hsp90α = heat shock protein 90, α isoform; Hsp90β = heat shock protein 90, β isoform; MALDI-TOF = matrix-assisted laser desorption ionization-time-of-flight; NE = nuclear extracts; NLS = nuclear localization signal; PBS = phosphate-buffered saline; PFTα = pifithrin-α; WCE = whole cell extracts.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AR performed cell culture studies, immunoblotting and immunoprecipitation, subcellular fractionation and immunocytochemistry, and participated in drafting the manuscript. MSC participated in immunoblotting studies. XT performed out in-gel enzymatic digestion of separated bands, and mass spectrometry. JEB participated in the protein identification and sequencing. YP participated in the design of the study and drafting of the manuscript. SSD conceived the study, participated in the design of its study and coordination and drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We thank Dr Heiko T Jansen (Washington State University, Pullman, WA) for help with the immunocytochemistry studies. This work is supported in part by grant no. BCTR0402398 to SSD from The Susan G Komen Foundation for Breast Cancer Research.
Figures and Tables
Figure 1 Coomassie blue-stained gel of proteins co-immunoprecipitated with DO-1 primary antibody from MDA-MB-231 cells. Cleared cell lysates were immunoprecipitated with DO-1 primary antibody directed against p53, washed, and resolved by SDS-PAGE (4 to 20% polyacrylamide). Two independent co-immunoprecipitated samples from untreated control (-) and cells treated for 4 hours with 100 μM PRIMA-1 (+) were loaded. The gels were stained with Coomassie blue. Molecular masses of protein size markers are indicated (MW). The arrowhead indicates the band of stained proteins excised for enzymatic digestion by trypsin and subsequent mass fingerprinting with matrix-assisted laser desorption ionization-time-of-flight mass spectrometry.
Figure 2 Inhibition of PRIMA-1 mediated transcriptional reactivation function of p53 with pifithrin-α (PFTα). MCF-7 (p53+/+) and GI-101A (mut p53) cells were treated with 100 μM PRIMA-1 for 2, 4 and 8 hours (lanes 1, 2 and 3, respectively). Cells were treated with 20 μM PFTα for 6 hours (lane 4) or with 20 μM PFTα for 2 hours followed by PRIMA-1 for 4 hours (lane 5). 20 μg of protein samples of cell lysates were separated by SDS-PAGE (4 to 20% polyacrylamide) and subjected to Western blot analysis with p53 and p21 primary antibodies. The reactive bands were revealed and detected with the Odyssey™ Infrared Imaging System. β-Actin was used as a loading control for protein samples.
Figure 3 Peptide mass fingerprinting of in-gel tryptic digest of the 90 kDa band generated with a matrix-assisted laser desorption ionization-time-of-flight mass spectrometer. Protein bands indicated by the arrowhead in Fig. 1 were subjected to in-gel digestion and analyzed by mass spectrometry. The tryptic peptides from this band showed the presence of six peptides corresponding to heat shock protein 90 (Hsp90) as one of the proteins that were found in the altered protein-protein interaction pattern of p53 with and without PRIMA-1 treatment.
Figure 4 Protein-protein interaction analysis of p53 and the α isoform of heat shock protein 90 (Hsp90α). (a) MDA-231 cell lysates from untreated cells (lanes 1 and 3) and cells treated for 4 hours with 100 μM PRIMA-1 (lanes 2 and 4) were immunoprecipitated (IP) with anti-Hsp90α monoclonal antibody and subjected to Western blotting (WB) with anti-p53 (DO-1) monoclonal antibody (lanes 1 and 2) in addition to reciprocal immunoprecipitation with DO-1 and Western blotting with anti-Hsp90α (lanes 3 and 4). (b) GI-101A cell lysates from untreated cells (lanes 1 and 3) and cells treated for 4 hours with 100 μM PRIMA-1 (lanes 2 and 4) were immunoprecipitated with anti-Hsp90α monoclonal antibody and subjected to Western blotting with anti-p53 (DO-1) monoclonal antibody (lanes 1 and 2) in addition to reciprocal immunoprecipitation with DO-1 and Western blotting with anti-Hsp90α (lanes 3 and 4).
Figure 5 Western blots of heat shock protein 90 (Hsp90) proteins from control and PRIMA-1-treated cells. Cells were treated with 100 μM PRIMA-1 for 2, 4 or 8 hours. 20 μg of protein samples of cell lysates from the whole cell extracts (WCE) and nuclear extracts (NE) of the control (C) and treated samples were separated by SDS-PAGE (4 to 20% polyacrylamide) and Western blotted with antibodies directed against both the α and β isoforms of Hsp90 (a) and against p53 (b). β-Actin was used as a loading control. The reactive bands were detected with the Odyssey™ Infrared Imaging System.
Figure 6 The nuclear localization of the α isoform of heat shock protein 90 (Hsp90α) is enhanced by treatment of cells with PRIMA-1. MDA-231 cells treated with PRIMA-1 were subjected to nuclear isolation. A fraction of the isolated nuclear pellet was fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100 and incubated overnight with anti-p53 and anti-Hsp90α monoclonal antibodies. Nuclear fractions were then immunostained with a secondary antibody conjugated with Oregon green to detect p53 (green) and a secondary antibody conjugated with Texas red to detect Hsp90α (red) before detection by fluorescence microscopy. Nuclei were stained with 4',6-diamidino-2-phenylindole (blue). Normal mouse immunoglobulin G was used as a negative control (data not shown). Arrows mark nuclear staining of Hsp90α. Scale bar, 5 μm.
Figure 7 Combination and sequential exposure of cells to PRIMA-1 and adriamycin. Exponentially growing MDA-MB-231 or GI-101A cells were seeded in 10% serum-supplemented RPMI-1640 medium at 104 cells per well. After 24 hours, cells were treated with 100 μM PRIMA-1 for 24 hours (P24), 0.2 μM adriamycin for 3 hours (A3) or 24 hours (A24), and a combination of PRIMA-1 plus adriamycin for 24 hours (AP24) or adriamycin for 3 hours followed by PRIMA-1 for 24 hours (A3P24). After drug treatment, the cells were reincubated in drug-free medium for a further 3 days and cell survival was determined with the crystal violet assay. Results are means ± SD for quadruplicate determinations (SD<10%) representative of two to four independent experiments; P<0.01.
==== Refs
Berns EM van Staveren IL Look MP Smid M Klijn JG Foekens JA Mutations in residues of TPp53 that directly contact DNA predict poor outcome in human primary breast cancer Br J Cancer 1998 77 1130 1136 9569050
Weinstein JN Myers TG O'Connor PM Friend SH Fornace AJ Kohn KW Fojo J Bates SE Rubinstein LV Anderson NL An information-intensive approach to the molecular pharmacology of cancer Science 1997 275 343 349 8994024 10.1126/science.275.5298.343
Jones CB Clements MK Wasi S Daoud SS Sensitivity to camptothecin of human breast carcinoma and normal endothelial cells Cancer Chemother Pharmacol 1997 40 475 483 9332461 10.1007/s002800050690
Goldwasser F Shimizu T Jackman J Hoki Y O'Connor PM Kohn KW Pommier Y Correlations between S and G2 arrest and the cytotoxicity of camptothecin in human colon carcinoma cells Cancer Res 1996 56 4430 4437 8813137
Fan S El-Deiry WS Bae I Freeman J Jondle D Bhatia K Fornace AJ Magrath I Kohn KW O'Connor PM p53 gene mutations are associated with decreased sensitivity of human lymphoma cells to DNA damaging agents Cancer Res 1994 54 5824 5830 7954409
Iwadate Y Fujimoto S Tagawa M Namba H Sueyoshi K Hirose M Sakiyama S Association of p53 gene mutation with decreased chemosensitivity in human malignant gliomas Int J Cancer 1996 69 236 240 8682593 10.1002/(SICI)1097-0215(19960621)69:3<236::AID-IJC14>3.0.CO;2-5
Daoud SS Munson PJ Reinhold W Young L Prabhu V Yu Q LaRose J Kohn KW Weinstein JN Pommier Y Impact of p53 knockout and topotecan treatment on gene expression profiles in human colon carcinoma cells: a pharmacogenomic study Cancer Res 2003 63 2782 2793 12782583
Lai H Lin L Nadji M Lai S Trapido E Meng L Mutations in the p53 tumor suppressor gene and early onset breast cancer Cancer Biol Ther 2002 1 31 36 12170762
Blaszyk H Hartmann A Cunningham JM Schaid D Wold LE Kovach JS Sommer SS A prospective trial of Midwest breast cancer patients: a p53 gene mutation is the most important predictor of adverse outcome Int J Cancer 2000 89 32 39 10719728 10.1002/(SICI)1097-0215(20000120)89:1<32::AID-IJC6>3.0.CO;2-G
Bull SB Ozcelik H Pinnaduwage D Blackstein ME Sutherland DA Pritchard KI Tzontcheva AT Sidlofsky S Hanna WM Qizilbash AH The combination of p53 mutation and neu/erbB2 amplification is associated with poor survival in node-negative breast cancer J Clin Oncol 2004 22 86 96 14701769 10.1200/JCO.2004.09.128
Hupp TR Sparks A Lane DP Small peptides activate the latent sequence-specific DNA binding function of p53 Cell 1995 83 237 245 7585941 10.1016/0092-8674(95)90165-5
Issaeva N Friedler A Bozko P Wiman KG Fersht AR Selivanova G Rescue of mutants of the tumor suppressor p53 in cancer cells by a designed peptide Proc Natl Acad Sci USA 2003 100 13303 13307 14595027 10.1073/pnas.1835733100
Chen L Lu W Agarwal S Zhou W Zhang R Chen J Ubiquitous induction of p53 in tumor cells by antisense inhibition of MDM2 expression Mol Med 1999 5 21 34 10072445
Foster BA Coffey HA Morin MJ Rastinejad F Pharmacological rescue of mutant p53 conformation and function Science 1999 286 2507 2510 10617466 10.1126/science.286.5449.2507
Bykov VJN Selivanova G Wiman KG Small molecules that reactivate mutant p53 Eur J Cancer 2003 39 1828 1834 12932659 10.1016/S0959-8049(03)00454-4
Bykov VJN Issaeva N Shilov A Hultcrantz M Pugacheva E Chumakov P Bergman J Wiman K Selivanova G Restoration of the tumor suppressor function to mutant p53 by a low-molecular-weight compound Nat Med 2002 8 282 288 11875500 10.1038/nm0302-282
Chipuk JE Maurer U Green DR Schuler M Pharmacologic activation of p53 elicits Bax-dependent apoptosis in the absence of transcription Cancer Cell 2003 4 371 381 14667504 10.1016/S1535-6108(03)00272-1
Rehman A Chahal MS Pommier Y Daoud SS PRIMA-1 restores the p21 transactivation by tumor suppressor p53 in breast cancer cells Proc AACR 2003 44 4592
Clements MK Jones CB Cumming M Daoud SS Antiangiogenic potential of camptothecin and topotecan Cancer Chemother Pharmacol 1999 44 411 416 10501915 10.1007/s002800050997
ProFound program
Jansen HT Cutter C Hardy S Lehman MN Goodman RL Seasonal plasticity within the gonadotropin-releasing hormone (GnRH) system of the ewe: changes in identified GnRH inputs and glial association Endocrinology 2003 144 3663 3676 12865349 10.1210/en.2002-0188
El-Deiry WS Tokino T Velculescu VE Levy DB Parson R Trent JM Lin D Mercer WE Kinzler KW Vogelstein B WAF1, a potential mediator of p53 tumor suppression Cell 1993 75 817 825 8242752 10.1016/0092-8674(93)90500-P
Akashi M Osawa Y Koeffler HP Hachiya M p21WAF1 expression by an activator of protein kinase C is regulated mainly at the post-transcriptional level in cells lacking p53: important role of RNA stabilization Biochem J 1999 337 607 616 9895308 10.1042/0264-6021:3370607
Zeng YX El-Deiry WS Regulation of p21WAF1/CIP1 gene expression by p53-independent pathways Oncogene 1996 12 1557 1564 8622872
Komarov PG Komarova EA Kondratov RV Christov-Tselkov K Coon JS Chernov MV Gudkov AV A chemical inhibitor of p53 that protects mice from the side effects of cancer therapy Science 1999 285 1733 1737 10481009 10.1126/science.285.5434.1733
Goodlett DR Bruce JE Anderson GA Rist B Pasa-Tolic L Fiehn O Smith RD Aebersold R Protein identification with a single accurate mass of a cysteine-containing peptide and constrained database searching Anal Chem 2000 72 1112 1118 10740847 10.1021/ac9913210
Nemoto T Sato N Oligomeric forms of the 90-kDa heat shock protein Biochem J 1998 330 989 995 9480920
Yamada S Ono T Mizuno A Nemoto TK A hydrophobic segment within the C-terminal domain is essential for both client binding and dimer formation of the HSP90-family molecular chaperone Eur J Biochem 2003 270 146 154 12492485 10.1046/j.1432-1033.2003.03375.x
Matsumoto S Tanaka E Nemoto TK Ono T Takagi T Imai J Kimura Y Yahara I Kobayakawa T Ayuse T Interaction between the N-terminal and middle regions is essential for the in vivo function of HSP90 molecular chaperone J Biol Chem 2002 277 34959 34966 12121981 10.1074/jbc.M203038200
Akner G Mossberg K Sundqvist KG Gustafsson JA Wikström AC Evidence for reversible, non-microtubule and non-microfilament-dependent nuclear translocation of Hsp90 after heat shock in human fibroblasts Eur J Cell Biol 1992 58 356 364 1425772
Langer T Rosmus S Fasold H Intracellular localization of the 90 kDA heat shock protein (HSP90α) determined by expression of a EGFP-HSP90α-fusion protein in unstressed and heat stressed 3T3 cells Cell Biol Int 2003 27 47 52 12713799 10.1016/S1065-6995(02)00256-1
Görlich D Kutay U Transport between the cell nucleus and the cytoplasm Annu Rev Cell Dev Biol 1999 15 607 660 10611974 10.1146/annurev.cellbio.15.1.607
Longshaw VM Chapple JP Balda MS Cheetham ME Blatch GL Nuclear translocation of the Hsp70/Hsp90 organizing protein mSTI1 is regulated by cell cycle kinases J Cell Sci 2004 117 701 710 14754904 10.1242/jcs.00905
Febbro M Henderson BR Regulation of tumor suppressors by nuclear-cytoplasmic shuttling Exp Cell Res 2003 282 59 69 12531692 10.1016/S0014-4827(02)00019-8
King FW Wawrzynow A Höhfeld J Zylicz M Co-chaperones Bag-1, Hop and Hsp40 regulate Hsc70 and Hsp90 interactions with wild type or mutant p53 EMBO J 2001 20 6297 6305 11707401 10.1093/emboj/20.22.6297
Chen J Wang C Phosphorylation and hsp90 binding mediate heat shock stabilization of p53 J Biol Chem 2003 278 2066 2071 12427754 10.1074/jbc.M206697200
Walerych D Kudla G Gutkowska M Wawrzynow B Muller L King FW Helwak A Boros J Zylicz A Zylicz M Hsp90 chaperones wild-type p53 tumor suppressor protein J Biol Chem 2004 279 48836 48845 15358769 10.1074/jbc.M407601200
Murphy PJ Galigniana MD Morishima Y Harrell JM Kwok RP Ljungman M Pratt W Pifithrin-α inhibits p53 signaling after interaction of the tumor suppressor protein with hsp90 and its nuclear translocation J Biol Chem 2004 279 30195 30201 15145929 10.1074/jbc.M403539200
Aas T Borresen AL Geisler S Smith-Sorensen B Johnsen H Varhaug JE Akslen LA Lonning PE Specific p53 mutations are associated with de novo resistance to doxorubicin in breast cancer patients Nat Med 1996 2 811 814 8673929 10.1038/nm0796-811
|
16168122
|
PMC1242148
|
CC BY
|
2021-01-04 16:54:51
|
no
|
Breast Cancer Res. 2005 Jul 27; 7(5):R765-R774
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1290
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12911616812310.1186/bcr1291Research ArticleStrong evidence that the common variant S384F in BRCA2 has no pathogenic relevance in hereditary breast cancer Wappenschmidt B [email protected] R 2Rhiem K [email protected] M 3Wardelmann E [email protected] A [email protected] N [email protected] P [email protected] RK [email protected] Department of Obstetrics and Gynaecology, University Hospital of Cologne, Cologne, Germany2 Institute of Medical Biometrics, Statistics and Epidemiology, University Hospital of Bonn, Bonn, Germany3 Data Management of the German Consortium for Hereditary Breast and Ovarian Cancer at the Institute of Medical Informatics, Statistics and Epidemiology, University of Leipzig, Leipzig, Germany4 Department of Pathology, University Hospital of Bonn, Bonn, Germany5 Department of Obstetrics and Gynaecology, Technical University Hospital, Munich, Germany6 Department of Obstetrics and Gynaecology, University Hospital of Kiel, Kiel, Germany2005 27 7 2005 7 5 R775 R779 23 11 2004 25 1 2005 26 5 2005 30 6 2005 Copyright © 2005 Wappenschmidt et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Unclassified variants (UVs) of unknown clinical significance are frequently detected in the BRCA2 gene. In this study, we have investigated the potential pathogenic relevance of the recurrent UV S384F (BRCA2, exon 10).
Methods
For co-segregation, four women from a large kindred (BN326) suffering from breast cancer were analysed. Moreover, paraffin-embedded tumours from two patients were analysed for loss of heterozygosity. Co-occurrence of the variant with a deleterious mutation was further determined in a large data set of 43,029 index cases. Nature and position of the UV and conservation among species were evaluated.
Results
We identified the unclassified variant S384F in three of the four breast cancer patients (the three were diagnosed at 41, 43 and 57 years of age). One woman with bilateral breast cancer (diagnosed at ages 32 and 50) did not carry the variant. Both tumours were heterozygous for the S384F variant, so loss of the wild-type allele could be excluded. Ser384 is not located in a region of functional importance and cross-species sequence comparison revealed incomplete conservation in the human, dog, rodent and chicken BRCA2 homologues. Overall, the variant was detected in 116 patients, five of which co-occurred with different deleterious mutations. The combined likelihood ratio of co-occurrence, co-segregation and loss of heterozygosity revealed a value of 1.4 × 10-8 in favour of neutrality of the variant.
Conclusion
Our data provide conclusive evidence that the S384F variant is not a disease causing mutation.
==== Body
Introduction
Inherited mutations in the BRCA1 and BRCA2 genes predispose to early breast and ovarian cancer. Besides clear pathogenic mutations (nonsense mutations or insertions and deletions leading to truncated proteins), many unclassified variants also exist, for example, missense mutations of unknown relevance that constitute about 30% of all mutations detectable in the BRCA1 or BRCA2 genes [1,2].
A lack of functional assays has hampered the conclusive validation of the consequences of these variants. This raises various problems for the clinical management, genetic counselling and personal life planning of people who carry such an unclassified BRCA1/2 variant. To characterise the potential pathogenic relevance of unclassified variants, co-segregation, co-occurrence, loss of heterozygosity (LOH) analysis, cross-species comparison, biochemical characterisation and occurrence in a healthy control cohort can be used [3,4]. We have investigated the common variant S384F (BRCA2, exon 10) in a large kindred (BN326) using these approaches.
Materials and methods
Co-segregation
A large kindred with the BRCA2 variant S384F [GenBank U43746] was recruited at the Centre for Familial Breast and Ovarian Cancer located at the universities of Cologne and Bonn (Fig. 1a). Blood samples were available from four women suffering from breast cancer (IDs 326.1, 326.2, 326.3 and 326.4). The study was permitted by the local ethics committee and written consent was obtained from all patients. DNA was extracted from peripheral leukocytes using a conventional phenol-chloroform protocol. Mutation analysis was performed as described before [2]. Segregation analysis in the BRCA2 gene was restricted to a 171 base pair fragment of exon 10 spanning the variant. For amplification, PCR was performed under the following conditions: 95°C for 30 s, 54°C for 30 s, 72°C for 1 minute, 35 cycles (Perkin Elmer, model 9600, Shelton, CT, USA); forward primer, 5'-gca aac gct gat gaa tgt g-3'; reverse primer 5'-ggc caa aga cgg tac aac t-3'. PCR products from leukocyte and tumour DNA were directly sequenced forward and reverse on a semiautomated sequencer (Applied Biosystems, model 377, Foster City, USA) using the ABI PRISM BigDye™ Terminator Cycle Sequencing Ready Reaction Kit version 1.1 (Applied Biosystems), according to the manufacturer's protocol.
Co-occurrence
Data on co-occurrence of the variant with deleterious mutations in the BRCA2 gene were provided by the German Consortium for Hereditary Breast and Ovarian Cancer (GCHBOC) and by Myriad Genetic Laboratories, where 3,029 and 40,000 patients suffering from familial breast and/or ovarian cancer had been tested.
Loss of heterozygosity analysis
Paraffin-embedded tumour samples from two affected women (ID 326.1 and ID 326.2, Fig. 1a) were available. After morphologic evaluation on 4 μm hematoxylin and eosin-stained sections, manual microdissection was performed to enrich tumour cells to >90%. DNA was extracted from two 10 μm sections using an all-tissue DNA-extraction kit according to the manufacturer's protocol (GEN-IAL, Troisdorf, Germany). LOH analysis was performed by direct sequencing as described above. In the case of LOH of the wild-type allele, the tumour is hemizygous for the variant.
Statistical analysis
A multifactorial model was used to calculate a combined likelihood ratio [3,4]. For analysis of co-segregation, the linkage program was applied [5]. For analysis of co-occurrence, we followed Goldgar's assumption that compound heterozygotes or homozygotes for deleterious mutations in the BRCA genes are extremely rare [3]. The probability of the observation of no LOH of the wild-type allele, in the case that the mutation is deleterious, was defined as 5% [6], whereas the observation of no LOH of the wild-type allele, in the case that the mutation is not deleterious, was defined as 80% [7].
Results
The variant S384F was first identified in patient 326.1, who suffered from uni-lateral breast cancer at 43 years of age. Further analysis showed that patients 326.2 and 326.4 (uni-lateral breast cancers at 41 and 57 years of age, respectively) were also carriers of the S384F variant. In contrast, a patient with bilateral breast cancer (ID326.3, diagnosed at 32 and 50 years of age) did not exhibit the variant. A complete BRCA1 and BRCA2 analysis was performed in this patient in order to exclude a second mutation. Carrier status and degree of kinship of all persons tested for the S384F variant are shown in Fig. 1a. Using the approach of Thompson et al. [4] and Lathrop et al. [5], we analysed the co-segregation of the variant and the disease. We calculated a likelihood ratio of LRcoseg = 3.236 that the variant is causal for the disease.
The variant was detected in 109 patients tested by Myriad Genetic Laboratories and in 9 patients tested by the GCHBOC and co-occurred five times with the different deleterious mutations Q258X, K2013X, S2219X, 3036del4 (Myriad) and 2046X (GCHBOC). It could not be detected in 200 healthy control women over the age of 60 years. Using the algorithm of Goldgar et al. [3] the likelihood ratio was LRcooc = 0.00000111 in favour of neutrality of the variant. Combining co-occurrence and co-segregation analysis revealed a likelhood ratio of LRcoseg+cooc = 0.0000036.
Tumour DNA from two patients (ID 326.1, 326.2) carrying the unclassified variant were heterozygous; loss of the wild-type allele could thus be excluded (Fig. 1b). Considering the different frequencies of occurrence of LOH in familial (95%) and sporadic breast carcinomas (20%) gave a likelihood ratio of LRLOH = 0.0039 for the variant to be neutral. If one assumes LOH as an independent event, the total likelihood ratio rises to LRtotal = 1.4 × 10-8 in favour of neutrality (Table 1).
Considering the position of the variant, codon 384 is not located in a region of functional importance. Cross-species sequence comparison revealed good alignment of the human, dog, rodent and chicken BRCA2 amino acid homologues (Fig. 2). Whereas human (U43746) and dog (NP001006654) BRCA2 homologues at positions 384 and 375 both encode a serine, however, mouse (U89652) and rat (U89653) BRCA2 homologues at positions 376 and 374 encode a cysteine. The chicken (NP989607) BRCA2 homologue at position 400 encodes for a valine. Biochemical evaluation revealed that the S384F variant results in a non-conservative substitution of a hydrophilic amino acid (Ser) by a hydrophobic one (Phe). Both are, however, uncharged.
Discussion
The germline variant S384F was previously classified by us as a variant of unknown significance. Segregation analysis revealed incomplete segregation in a large kindred. While three women affected by breast cancer carried the variant, one was lacking it. The latter was affected by bi-lateral breast cancer at 32 and 50 years of age so it is very unlikely that this patient suffered from sporadic breast cancer, indicating that the underlying mutation is not yet identified. Using the approach of Thompson et al. [4] we calculated a LR of 3.236 in favour of the variant being deleterious.
The variant could be identified in 116 of 43,029 patients with a history of familial breast cancer. In this cohort, the variant co-occurred five times with different deleterious mutations, indicating that the variant is located in trans. In accordance with Goldgar et al. [3], we assumed that biallelic deleterious BRCA2 mutations are extremely rare as they generally cause Fanconi anemia type D1 leading to early childhood death. Based on a proposed probability of p2 = 0.001 that an individual with the variant also carries a deleterious mutation in trans, Goldgar et al. suggested that a LR >1.0 is in favour of the variant being deleterious whereas a LR <0.01 argues in favour of the variant being neutral. For the S384F variant, we calculated a combined LR for co-occurrence and co-segragation of 0.0000036, leading to the conclusion that the variant is neutral.
Recently, Osorio and co-workers [6] evaluated the rate and significance of LOH at the BRCA loci in 47 tumour samples from high risk patients with familial breast cancer. Their results suggest that LOH of the wild-type allele is a common mechanism of inactivation of the BRCA genes (95%) and that LOH analysis can be used to clarify the relevance of variants of unknown significance. In contrast, LOH of the BRCA genes occurs in only 20% of sporadic breast carcinomas [7]. Based on these rates, we again calculated a LR in favour of the variant being neutral from the observation of the two patients (326.1 and 326.2) who were heterozygous.
Also, the position and the nature of the amino acid substitution provide evidence for neutrality of the variant: the serine residue at codon 384 is not located in a protein domain of critical function [8]; and, although the unclassified variant results in a non-conservative substitution of a hydrophilic amino acid (Ser) by a hydrophobic one (Phe), both amino acids are uncharged. Additionally, both cysteine in rodents and valine in chicken are large hydrophobic amino acids with uncharged residues suggesting that these substitutions are not functionally relevant. The degree of conservation of BRCA2 homologues between species, however, is incomplete.
Conclusion
To summarize, the combined use of co-segregation, co-occurrence and LOH analysis, as well as the position and nature of the amino acid substitution, provides strong evidence that the S384F variant is not the disease causing mutation in family BN326. We cannot, however, exclude that the variant may act as a modifying or low penetrance gene that may exhibit incomplete segregation and retention of the wild-type allele, as has been demonstrated for the CHEK2 gene [9,10].
Abbreviations
GCHBOC = German Consortium for Hereditary Breast and Ovarian Cancer; LOH = loss of heterozygosity; LR = likelihood ratio; PCR = polymerase chain reaction.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
BW and RKS are responsible for the design of the study and drafted the manuscript. BW carried out the molecular genetic studies. RKS and KR counselled the family members and collected the samples. AD and MB provided the data for co-occurrence and RF performed the statistical analysis. EW took responsibility for morphological evaluation and manual microdissection of the tumour samples. The 200 healthy individuals were collected by AM and NA at the University Hospitals Munich and Kiel. PM participated in preparing the manuscript.
Acknowledgements
This work was supported by a grant from the German Cancer Aid (Deutsche Krebshilfe) to RKS. We are much obliged to Amie Deffenbaugh from Myriad Genetic Laboratories and the German Consortium for Hereditary Breast and Ovarian Cancer who provided the data on co-occurrence. The technical assistance by Brigitte Kau, Brigitte Heim and Gabi Krebsbach is gratefully acknowledged. We thank all members of the family who took part in the study and gave written consent for publication of the results.
Figures and Tables
Figure 1 Segregation and loss of heterozygosity analysis of the S384F variant (BRCA2, exon 10) in kindred ID 326. (a) Patients ID 326.1, ID 326.2 and ID 326.4 were heterozygous for the variant. In contrast, a patient with bilateral breast cancer (ID 326.3) diagnosed at 32 and 50 years of age did not carry the variant. Filled symbols indicate individuals with breast cancer. UV+, patients heterozygous for the S384F variant; UV-, patients homozygous for the wild-type allele. (b) Arrows indicate patients from whom tumour samples were available. In the tumours of both ID 326.1 (left) and ID 326.2 (right), loss of the wild-type allele could be excluded.
Figure 2 Cross-species comparison of the BRCA2 homologue: Alignment of the predicted BRCA2 amino acid sequence of human (U43746), dog (NP001006654), mouse (U89652), rat (U89653) and chicken (NP989607). Amino acid sequences identical in humans, dog, rodents and chicken are in red. The asterisk indicates the position of the S384F variant.
Table 1 Likelihood ratios for the S384F variant
LR for S384F (BRCA2)
Co-segregation 3.236
Co-occurrence 0.00000111
Loss of heterozygosity 0.00391
Overall LR 1.4 × 10-8
An overall likelihood ratio (LR) of <0.01 is considered to prove that the variant is neutral [3].
==== Refs
Breast Cancer Information Core (BIC) Database
German Consortium for Hereditary Breast and Ovarian Cancer Comprehensive analysis of 989 patients with breast and ovarian cancer provides BRCA1 and BRCA2 mutation profiles and frequencies for the German population Int J Cancer 2002 97 472 480 11802209 10.1002/ijc.1626
Goldgar DE Easton DF Deffenbaugh A Monteiro ANA Tavtigian SV Couch FJ the Breast Cancer Information Score (BIC) Steering Committee Integrated evaluation of the DNA sequence variants of unknown clinical significance: approach to BRCA1 and BRCA2 Am J Hum Genet 2004 75 535 544 15290653 10.1086/424388
Thompson D Easton DF Goldgar DE A full-likelihood method for the evaluation of causality of sequence variants from familial data Am J Hum Genet 2003 73 652 655 12900794 10.1086/378100
Lathrop GM Lalouel J-M Julier C Ott J Strategies fur multilocus linkage analysis in humans Proc Natl Acad Sci USA 1984 81 3443 3446 6587361
Osorio A De la Hoya M Rodriguez-Lopez R Martinez-Ramirez A Cazorla A Granizo JJ Esteller M Rivas C Caldes T Benitez J Loss of heterozygosity analysis at the BRCA loci in tumour samples from patients with familial breast cancer Int J Cancer 2002 99 305 309 11979449 10.1002/ijc.10337
Schmutzler RK Fimmers R Bierhoff E Lohmar B Homann A Speiser A Kubista E Laeger K Krebs D Zeilinger R Association of allelic losses on human chromosomal arms 11q and 16q in sporadic breast cancer Int J Cancer 1996 69 307 311 8797873 10.1002/(SICI)1097-0215(19960822)69:4<307::AID-IJC12>3.0.CO;2-2
Wong AKC Pero R Ormonde PA Tavtigian SV Bartel PL RAD51 interacts with the evolutionary conserved BRC motifs in the human breast cancer susceptibility gene BRCA2 J Biol Chem 1997 272 31941 31944 9405383 10.1074/jbc.272.51.31941
Dufault MR Betz B Wappenschmidt B Hofmann V Bandick K Golla A Pietschmann A Nestle-Krämling C Rhiem K Hüttner C Limited relevance of the CHEK2 gene in hereditary breast cancer Int J Cancer 2004 110 320 325 15095295 10.1002/ijc.20073
Oldenburg RA Kroeze-Jansema K Kraan J Morreau H Klijn JGM Hoogerbrugge N Ligtenberg MJL van Asperen CJ Vasen HFA Meijers C The CHEK2* 1100delC variant acts as a breast cancer risk modifier in non-BRCA1/BRCA2 multiple case families Cancer Res 2003 63 8153 8157 14678969
|
16168123
|
PMC1242149
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jul 27; 7(5):R775-R779
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1291
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer ResBreast Cancer Research : BCR1465-54111465-542XBioMed Central bcr12921616812810.1186/bcr1292Research ArticleMetastasis of hormone-independent breast cancer to lung and bone is decreased by α-difluoromethylornithine treatment Richert Monica M [email protected] Pushkar A [email protected] Gail [email protected] Douglas J [email protected] Sharlene [email protected] Laurence M [email protected] Judith S [email protected] Andrea [email protected] Danny R [email protected] Department of Pathology, Comprehensive Cancer Center, Center for Metabolic Bone Disease, National Foundation for Cancer Research – Center for Metastasis Research, University of Alabama at Birmingham, Alabama, USA2 Department of Biochemistry and Molecular Biology, The Pennsylvania State University College of Medicine, Hershey, Pennsylvania, USA3 Department of Medicine, The Pennsylvania State University College of Medicine, Hershey, Pennsylvania, USA4 Department of Pathology, The Pennsylvania State University College of Medicine, Hershey, Pennsylvania, USA2005 9 8 2005 7 5 R819 R827 5 5 2005 21 6 2005 22 6 2005 1 7 2005 Copyright © 2005 Richert et al, licensee BioMed Central Ltd.2005Richert et al, licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is cited.Introduction
Polyamines affect proliferation, differentiation, migration and apoptosis of cells, indicating their potential as a target for cancer chemotherapy. Ornithine decarboxylase converts ornithine to putrescine and is the rate-limiting step in polyamine synthesis.
α-Difluoromethylornithine (DFMO) irreversibly inhibits ornithine decarboxylase and MDA-MB-435 human breast cancer metastasis to the lung without blocking orthotopic tumor growth. This study tested the effects of DFMO on orthotopic tumor growth and lung colonization of another breast cancer cell line (MDA-MB-231) and the effects on bone metastasis of MDA-MB-435 cells.
Methods
MDA-MB-231 cells were injected into the mammary fat pad of athymic mice. DFMO treatment (2% per orally) began at the day of tumor cell injection or 21 days post injection. Tumor growth was measured weekly. MDA-MB-231 cells were injected into the tail vein of athymic mice. DFMO treatment began 7 days prior to injection, or 7 or 14 days post injection. The number and incidence of lung metastases were determined. Green fluorescent protein-tagged MDA-MB-435 cells were injected into the left cardiac ventricle in order to assess the incidence and extent of metastasis to the femur. DFMO treatment began 7 days prior to injection.
Results
DFMO treatment delayed MDA-MB-231 orthotopic tumor growth to a greater extent than growth of MDA-MB-435 tumors. The most substantial effect on lung colonization by MDA-MB-231 cells occurred when DFMO treatment began 7 days before intravenous injection of tumor cells (incidence decreased 28% and number of metastases per lung decreased 35–40%). When DFMO treatment began 7 days post injection, the incidence and number of metastases decreased less than 10%. Surprisingly, treatment initiated 14 days after tumor cell inoculation resulted in a nearly 50% reduction in the number of lung metastases without diminishing the incidence. After intracardiac injection, DFMO treatment decreased the incidence of bone metastases (55% vs 87%) and the area occupied by the tumor (1.66 mm2 vs 4.51 mm2, P < 0.05).
Conclusion
Taken together, these data demonstrate that DFMO exerts an anti-metastatic effect in more than one hormone-independent breast cancer, for which no standard form of biologically-based treatment exists. Importantly, the data show that DFMO is effective against metastasis to multiple sites and that treatment is generally more effective when administered early.
==== Body
Introduction
Metastasis is the leading cause of death of breast cancer patients. As tumors progress from hormone-dependent to hormone-independent, their risk of metastasis increases and the treatment options decrease. There are currently no efficacious biologically-based therapies for the more aggressive hormone-independent breast cancers. It is therefore important to explore potential therapies that may be effective against both metastasis and growth of hormone-independent breast tumors.
Polyamines are organic compounds derived from decarboxylation of the amino acid ornithine. These compounds play a role in a wide variety of cellular functions including proliferation, differentiation, migration and apoptosis [1]. They are required for cell viability, as demonstrated by the lethality of ornithine decarboxylase (ODC) null mutations in mice [2]. ODC is the rate-limiting enzyme in polyamine synthesis: It converts ornithine to putrescine to begin the polyamine cascade. Overexpression of ODC in the immortalized, but non-tumorigenic, mammary epithelial cell line MCF10A results in a partially transformed phenotype as well as in the activation of ERK, which is often activated in breast cancers [3,4]. Previous studies demonstrated that increased ODC activity correlates with transition to hormone independence, increased anaplasia and metastasis. Increased ODC activity is an independent adverse prognostic factor for overall breast cancer survival in women with localized disease [5-7], suggesting that increased activity of the polyamine pathway may result in more aggressive, hormone-independent breast cancers.
α-Difluromethylornithine (DFMO) is an irreversible inhibitor of ODC. Treatment with DFMO results in decreased polyamine pools causing a cytostatic effect. Our laboratory previously demonstrated that DFMO treatment decreased lung metastasis from the estrogen receptor-negative/progesterone receptor-negative breast carcinoma cell line MDA-MB-435 xenografts, while having only a modest or absent effect on the growth of the orthotopic tumors [8,9]. Previous studies have shown that DFMO (0.5–3% in drinking water) was well tolerated in both rats and mice. The minimal toxicities of DFMO make it a suitable candidate for long-term treatments.
The studies described here were designed to address two questions. First, does DFMO reduce metastasis to sites other than lung? To address this question, we took advantage of a recently developed enhanced green fluorescent protein (GFP)-expressing variant of MDA-MB-435 [10,11]. Second, does DFMO affect orthotopic tumor growth and metastasis of another hormone-independent breast carcinoma? To test this question, we utilized the estrogen receptor-negative/progesterone receptor-negative MDA-MB-231 cell line. We present here that DFMO can decrease metastasis of MDA-MB-435 cells to the bone and can affect lung metastasis and orthotopic tumor growth of MDA-MB-231 cells.
Materials and methods
Cell lines
MDA-MB-231 and MDA-MB-435 human breast cancer cell lines were kindly provided by Dr Janet E Price at the University of Texas MD Anderson Cancer Center. They were cultured in DMEM/Ham's F12 medium supplemented with 5% fetal bovine serum, 1% non-essential amino acids, 1 mM sodium pyruvate, and were maintained at 37°C with 5% CO2 in a humidified atmosphere. The cells were passaged using 0.125% trypsin, 2 mM ethylenediamine tetraacetic acid in Ca2+/Mg2+-free Dulbecco's PBS.
Metastasis assays
All animals were maintained under the guidelines of the IACUC of the University of Alabama at Birmingham under registered protocols. Female athymic mice (Harlan Sprague-Dawley, Inc., Indianapolis, IN, USA) were used for all studies.
To study bone metastasis, 4-week-old to 6-week-old female mice were injected intracardially with 3 × 105 GFP-labeled MDA-MB-435 cells in 0.2 ml ice-cold Hanks Balanced Salt Solution (HBSS). Seven days prior to injection, two groups of 10 mice each were randomly selected. One group remained untreated, while the second group was provided DFMO as a 2% solution in the drinking water ad libitum until 6 weeks post injection. This dose had been previously demonstrated as efficacious. The mice were euthanized and the femurs dissected away from the soft tissue. Bone metastases were visualized by fluorescence microscopy [10,11]. Brightfield and fluorescent photographs were analyzed using Sigma Scan (SPSS Inc., Chicago, IL, USA) to determine the total area of bone, the area of bone occupied by fluorescing tumor cells and the percentage of bone occupied by fluorescing tumor cells.
MDA-MB-231 breast cancer cells (2 × 105 cells in 0.2 ml HBSS) were injected intravenously into the lateral tail vein of 3-week-old to 4-week-old athymic mice to evaluate lung colonization as metastases from an orthotopically growing tumor are relatively rare from this cell line. Each treatment group consisted of 20 mice. DFMO treatment (2% in drinking water) was administered using three schedules beginning: 7 days prior to tumor cell injection, 7 days after tumor cell injection, or 14 days after tumor cell injection. DFMO was administered until the experiment was terminated (4–6 weeks post injection). Control animals did not receive DFMO. At termination, the lungs were removed and fixed in Bouin's fixative diluted 1:5 with neutral-buffered formalin. The number of lungs with surface metastases were determined, as well as the number of surface metastases per lung by examination under a dissecting microscope, as described elsewhere [12].
Orthotopic tumor growth was measured by injecting MDA-MB-231 breast cancer cells (5 × 105 cells in 0.1 ml HBSS) into the second thoracic mammary fat pad of 5-week-old to 6-week-old female athymic nude mice, as described previously [12]. DFMO treatment (2% in drinking water) began at the time of mammary fat pad injection or at 21 days post injection. The control group consisted of 17 mice, the group started on DFMO at the time of injection consisted of 20 mice, while three mice began treatment 21 days post injection. Tumor growth was monitored weekly by measuring the tumor length and width with a caliper and was reported as the mean tumor diameter as previously described [12]. Since DFMO retarded tumor growth, DFMO-treated mice were euthanized when average local tumor diameters reached approximately 1.5 cm. In a subsequent experiment, all animals were euthanized 6 weeks post tumor cell injection. To examine growth parameters, mice were injected with 100 mg/kg BrdU approximately 2 hours prior to euthanasia. Tumors were removed and divided into two portions. One aliquot was fixed in 10% neutral-buffered formalin and processed for histological analysis. The other aliquot was frozen in liquid nitrogen and stored at -70°C for analysis of polyamine levels as described [9,13].
Polyamine levels
Athymic mice were injected subcutaneously (or in the mammary fat pad as already described) with 5 × 105 MDA-MB-231 cells in 0.1 ml HBSS. The comparison of orthotopic sites with ectopic sites was carried out in order to evaluate potential pharmacologic differences. The mice were separated into two animals per group. The control group was left untreated, while DFMO treatment began 7 days prior to tumor cell injection, on the day of injection, or 21 days post tumor cell injection. At 6 weeks post injection, tumors were removed and homogenized in 25 mM Tris-HCl buffer, pH 7.4, containing 0.1 mM ethylenediamine tetraacetic acid and 2.5 mM dithiothreitol. The homogenates were centrifuged for 30 min at 20,000 × g. An aliquot of the supernatant was used to measure ODC activity. The remaining aliquot was extracted with 0.6 N perchloric acid for 1 hour at 4°C prior to centrifugation at 15,000 × g for 15 min. The supernatant was used for polyamine determination. Polyamine levels were determined using high-pressure liquid chromatography as previously described [9,13].
Quantification of bromodeoxyuridine incorporation
Incorporation of BrdU was determined using immunohistochemical staining. Briefly, the sections were deparaffinized using xylenes and ethanol, and were then treated with 3% hydrogen peroxide for 10 min, 0.1% trypsin for 30 min and 2 N HCl for 30 min with PBS washes between each treatment. The sections were blocked with 5% normal goat serum in 1% BSA, 0.05% Tween 20 and 0.1% NaN3 in PBS followed by a 2-hour incubation at 37°C in a 1:1000 dilution of a monoclonal anti-BrdU antibody (B2531; Sigma, St. Louis, MO, USA). The sections were rinsed and then incubated for 1 hour at 37°C in a 1:5000 dilution of anti-mouse AlexaFluor Green. After the final washes, the sections were cover-slipped in Vectashield mounting media containing DAPI diluted 1:7 with non-DAPI-containing Vectashield mounting media. Five photographs (100 × magnification) from two sections of each tumor were counted for BrdU-labeled cells and the total number of labeled cells was graphed.
Quantification of TUNEL staining
Apoptotic cells were identified using the Apoptag plus staining kit from Chemicon (S7101; Temecula, CA, USA) according to the manufacturer's instructions. Five photographs (63 × magnification) from two sections of each tumor were counted for TUNEL-labeled cells, and the total number of labeled cells was graphed.
Semi-quantitative RT-PCR
Total RNA was isolated using Trizol (Invitrogen, Carlsbad, CA, USA) and 500 ng was used for RT-PCR with a SuperscriptIII/Platinum Taq One Step Kit (Invitrogen) as directed. Human glyceraldehyde-3-phosphate dehydrogenase primers (5'-GTGAAGGTCGGAGTCAACGGATT-3' and 5'-AGTGATGGCATGGACTGTGGTC-3') and hypoxanthine phosphoribosyl transferase primers (5'-CCAAAGATGGTCAAGGTCGC-3' and 5'-CTGCTGACAAAGATTCACTGG-3') were used to assess equal template loadings, and the linear range for each primer pair was determined independently. Human meprin α levels were determined with the following primers: 5'-ATCGGAGGCACGGCTGGCGTG-3' and 5'-GCCTGCCCTCATGGAGCTTACAG-3'. RT-PCR products were separated on a 1% TAE/agarose gel and quantified using a Stratagene Eagle Eye system with multiple integrations.
Statistics
Comparisons between treatment groups and control-treated mice were made using SigmaStat statistical software (SPSS Inc.). For multiple group comparisons, one-way analysis of variance was performed, followed by the Student-Neumann-Keuls post-test. Results were considered statistically different if P < 0.05.
Results and discussion
In normal tissues, ODC activity is increased by a variety of environmental and genetic factors associated with carcinogenesis, including ultraviolet light, asbestos and androgens. Increased ODC activity persists in and is associated with a wide variety of epithelial neoplasms including breast, skin, colon and prostate (reviewed in [1]). Together, these correlations suggest a role for ODC in tumor development and progression. Functional evidence supports this conclusion. In skin carcinogenesis models, ODC heterozygous knockout mice have significantly reduced tumor development [14].
Hormone-independent breast cancers are generally more aggressive and metastatic than hormone-responsive tumors. Unfortunately, there are currently no efficacious biologically-based treatments for these more aggressive forms of breast cancer whose survival rates are less than 26%. Polyamines and ODC are increased in breast cancer compared with normal breast tissue, and increased ODC activity directly correlates with a less differentiated and more metastatic tumor phenotype [5,6,15]. ODC activity is also an independent adverse prognostic factor for overall breast cancer survival [5,6,15]. DFMO, an irreversible inhibitor of ODC, reduces polyamine pools resulting in a cytostatic effect. Our previous studies demonstrated that treatment with DFMO reduces lung metastasis of the hormone-independent MDA-MB-435 breast cancer cell line with only a modest effect on the growth of orthotopic tumors [8,16]. This led to two questions: Can inhibition of ODC activity by DFMO decrease metastasis of MDA-MB-435 cells to another secondary site, namely bone? Can DFMO decrease orthotopic tumor growth and lung metastasis of another hormone-independent breast cancer cell line, MDA-MB-231? DFMO inhibition of tumor growth and metastasis of hormone-independent cells would strongly support targeting the polyamine pathway as a potential treatment.
Polyamine levels are decreased in tumors treated with DFMO
Polyamine levels were determined in orthotopically (mammary fat pad) and ectopically (subcutaneous) growing tumors in order to demonstrate that DFMO was inhibiting ODC activity. DFMO treatment beginning at either 7 days prior to injection (d-7), on the day of injection (d0) or 21 days following (d+21) injection of MDA-MB-231 cells into the mammary fat pad resulted in decreased putrescine and spermidine levels with only slight effects on spermine levels (Fig. 1). Putrescine levels decreased from an average of 0.200 nmol/mg tissue in the control samples to below the detection level in all of the DFMO-treated samples, while the spermidine levels decreased from a mean of 3.7905 nmol/mg in the untreated samples to 1.606 nmol/mg after treatment. As expected [17,18], the spermine levels decreased only slightly from an average of 2.931 nmol/mg in the untreated controls to 2.390 nmol/mg after treatment. This demonstrates that ODC was effectively inhibited by DFMO treatment.
Figure 1 α-Difluoromethylornithine (DFMO) (2% per orally in drinking water) decreases polyamine levels in mammary fat pad tumors of MDA-MB-231 cells. Polyamine levels were determined in control subcutaneous (s.c.) and mammary fat pad (mfp) tumors as well as in DFMO-treated tumors grown in the mammary fat pad. DFMO treatment began at either 7 days prior to tumor cell injections (DFMO-7), on the day of injection (DFMO 0) or 21 days post injection (DFMO+21). Putrescine (black), spermidine (diagonal line) and spermine (cross-hatched) levels are graphed as nanomoles per milligram of tumor.
Curiously, polyamine levels in MDA-MB-231 cells were significantly different when the cells were grown orthotopically versus those grown ectopically (Fig. 1). This result suggested that DFMO treatments might exert different effects on cells depending upon their location. The result is not entirely surprising since several previous experiments have shown that the biologic behavior of tumor cells can vary widely based upon the site of injection [12].
DFMO treatment decreases bone metastasis by MDA-MB-435 cells
We previously showed that MDA-MB-435 metastasis was decreased by 74% to the lung when mice were treated with DFMO [9] despite no apparent change in local tumor invasion. That result indicated that DFMO was affecting the later stages of metastasis (i.e. colonization). Breast cancers commonly metastasize to bone [19]. While bone metastases are not directly responsible for most breast cancer deaths, they result in profoundly decreased quality of life. While current therapies exist to decrease osteolysis and reduce sequella of bone metastases, there are few options to diminish bone tumor burden [19,20].
Taking into account the importance of breast cancer metastasis to bone and observations that polyamine levels varied depending upon tumor cell location in the body, we therefore asked whether DFMO had an effect on bone metastasis. To do this, we employed a model recently developed by ourselves using GFP-labeled MDA-MB-435 cells [10,11]. This model allows for relatively rapid assessment of the tumor burden in bone and is not dependent solely upon radiographic imaging and histology, which are less sensitive and more laborious, respectively.
As for all xenograft models, metastases in bone are not observed following orthotopic injection; tumor cells are therefore injected directly into the left ventricle of the heart. While metastases develop in other bones, this study involved complete analysis only on femurs because this site reflects changes elsewhere [10,11] and because it is a site commonly affected in women with breast cancer. DFMO treatment was begun 7 days prior to intracardiac injection of GFP-labeled MDA-MB-435 cells and lasted for the duration of the experiment. Mice with fluorescently-labeled femoral tumors were counted (Fig. 2a,b). DFMO treatment decreased the number of mice with bone metastases from 87.5% to 55.5% (Fig. 2a,b). The GFP label in each femur was then quantified by image analysis and the area of bone occupied by tumor cells was calculated (Fig. 2c,d). DFMO significantly decreased the area of bone occupied by the tumor from 4.51 to 1.69 mm2 (Fig. 2c; P < 0.05) and the percentage of bone occupied by the tumor decreased from 19.03% to 6.45% (Fig. 2d; P < 0.05). The intensity of fluorescence was not a variable for these analyses since we previously observed that intensity was dependent upon the depth of tumor cells to the bone surface [10]. As an internal control, however, bones were examined from both sides, and the areas occupied by tumor were found to be comparable (data not shown). Previous and ongoing independent studies have validated tumor location by histology and histomorphometry.
Figure 2 α-Difluoromethylornithine (DFMO) decreases the incidence and size of bone metastases from MDA-MB-435 cells. Green fluorescent protein-labeled MDA-MB-435 cells were injected intracardially into athymic mice. DFMO (2% per orally in drinking water ad libitum) treatment began 7 days prior to injection and continued for 6 weeks post injection. Bone metastases were visualized using a fluorescence microscope. (a) Photographs of femurs from control (n = 10) and DFMO-treated (n = 10) mice show the presence and size of the bone metastases. (b) Incidence of metastasis was determined by counting the number of mice with green fluorescence in the femur regardless of size of the mass. Image analysis was used to quantify fluorescence in each bone. (c) The area of tumor in bone was quantified by comparing the total area of the femur with the area containing green fluorescence. (d) The percentage of bone occupied by the tumor was also determined.
To the best of our knowledge, these are the first data to demonstrate efficacy of DFMO for metastases at a site other than the lung. The polyamine pathway therefore shows promise as a target for decreasing metastasis to multiple sites in breast cancer patients. Our studies have a modest limitation in that they only test efficacy of DFMO against the MDA-MB-435 cells to bone. We and other workers [21-23] have shown that the MDA-MB-231 cells will also colonize bone. But in the absence of GFP-tagged variants, the MDA-MB-231 bone metastasis assays are limited to radiologic or histologic detection methods that are less sensitive or more laborious. They were therefore not utilized here.
MDA-MB-231 orthotopic tumor growth is delayed by DFMO
This study, combined with previous data, indicates that the polyamine pathway can be modulated to affect primary tumor growth and metastasis of a hormone-independent breast cancer cell line (MDA-MB-435). To begin assessing whether DFMO treatment might be generally efficacious against other hormone-independent breast cancers, we examined orthotopic tumor growth and lung colonization of a second metastatic hormone-independent breast cancer cell line, MDA-MB-231.
Orthotopic tumor growth of MDA-MB-435 cells was previously shown to be mildly, but significantly, delayed by or not affected by treatment with DFMO [8,9]. Untreated MDA-MB-435 tumors reached a size of 100 mm2 2 weeks earlier than DFMO-treated tumors. We had also previously shown that MDA-MB-231 orthotopic tumors were delayed by DFMO treatment. We repeated this experiment with slight variations. DFMO treatment began either on the day of injection (d0) or 21 days post injection (d+21) of MDA-MB-231 cells into the second thoracic mammary fat pad and were continued until tumors reached a mean tumor diameter of approximately 1.2 cm (Experiment 1) or until 6 weeks post injection (Experiment 2). DFMO treatment delayed orthotopic tumor growth to ~1.2 cm from 42 days in control animals to 77 days (Fig. 3). At 42 days post injection, control tumors were 1.15 cm while DFMO-treated tumors were 0.75 cm when treatment began concurrent with tumor injection, and surprisingly smaller (0.62 cm) when treatment began on day 21 (Fig. 3).
Figure 3 α-Difluoromethylornithine (DFMO) delays orthotopic tumor growth of MDA-MB-231 cells. Athymic mice were injected into the second thoracic mammary fat pad with MDA-MB-231 cells. DFMO treatment (2% per orally) began either contemporaneously (▴, n = 20) or 21 days post injection (▼, n = 3). Mammary fat pad tumor length and width were measured once weekly. Control (●, n = 17) and 21 days post injection (DFMO+21) animals were sacrificed 6 weeks post injection. Day of injection (DFMO 0) animals were euthanized either 6 weeks (n = 10) or 11 weeks (n = 10) post injection.
DFMO treatment can thus delay orthotopic tumor growth of more than one breast cancer cell line, albeit with heterogeneous responses. MDA-MB-231 cells were more sensitive to DFMO at the primary site; that is, MDA-MB-435 tumors were mildly delayed (~2 weeks) or not delayed at all compared with a delay of ~5 weeks for MDA-MB-231. Postponement of DFMO treatment until 3 weeks post injection also inhibited MDA-MB-231 tumor growth, suggesting that DFMO could still be efficacious if administered at the time of diagnosis.
MDA-MB-231 lung metastasis is decreased by DFMO
Like MDA-MB-435 cells, DFMO reduced lung metastasis of MDA-MB-231 cells. The latter, however, had to be evaluated following injection of cells directly into the lateral tail vein because the parental cells are not effective at spontaneous metastasis. This experimental design afforded examination of scheduling since all tumor cells were administered as a bolus. While lung metastasis was reduced in all treatment groups, we were surprised by the pattern of suppression, even though it was reproducible. The incidence of metastases (i.e. proportion of mice developing lung metastases) decreased as expected. Earlier treatment resulted in fewer mice with metastases: 25%, 10% and 0% decreased for treatments beginning d-7, d0 and d+14 (Fig. 4a). However, the mice receiving DFMO beginning on d+14 showed diminishment of the number of metastases per lung (42%, mean = 12.5/lung vs 22.2/lung for controls), compared with no effect for d+7 (mean = 20.5/lung) and 32% for d-7 (mean = 15.1/lung) treatment schedules (Fig. 4b). Despite the trend toward decreased lung metastasis, and in contrast to MDA-MB-435 cells, the reductions in lung metastasis never reached statistical significance (P < 0.05) for MDA-MB-231 cells. Nonetheless, DFMO was clearly exerting reproducible effects on breast carcinoma growth in the mammary fat pad and in the lung.
Figure 4 α-Difluoromethylornithine (DFMO) treatment decreases lung metastasis of MDA-MB-231 cells. MDA-MB-231 cells were injected into the tail vein of athymic mice. DFMO treatment (2% per orally) began either 7 days prior to (DFMO-7), or 7 days (DFMO+7) or 14 days (DMFO+14) after, tumor cell injection. Mice were killed either 4 weeks (Experiment 1) or 6 weeks post injection (Experiment 2). Lungs were removed and fixed in Bouin's fixative diluted 1:5 in neutral buffered formalin. Each lung was examined for the presence of surface metastases. (a) The incidence of mice with lung metastases was graphed as a percentage of the total number of mice (n = 10 for each group). (b) The average number of metastases per lung was graphed as a percentage of the control (n = 10 for each group). Experiment 1 (black) and Experiment 2 (small hatched) are graphed separately as well as combined (large hatched).
DFMO treatment decreases proliferation in MDA-MB-231 orthotopic tumors
To begin addressing mechanisms for increased susceptibility of MDA-MB-435 cells to spontaneous metastasis suppression and of MDA-MB-231 cells to suppression of tumor growth, we performed additional studies. In vitro growth was not significantly affected (data not shown). To further assess growth and apoptosis, we examined MDA-MB-231 tumors growing orthotopically by BrdU and TUNEL staining to determine whether DFMO treatment altered proliferation or apoptosis, respectively. We found that DFMO treatment beginning d0 increased apoptosis (Fig. 5a) and slightly decreased proliferation (Fig. 5b) compared with control tumors. The decrease in BrdU labeling is consistent with, but less dramatic than, the 60% decrease of Ki67 labeling observed in DFMO-treated MDA-MB-435 tumors, while the increase in TUNEL staining is in contrast to the MDA-MB-435 tumors where no change in apoptosis was observed, even though cleaved caspase-3 levels were increased fourfold. Non-apoptotic necrosis was decreased by 60% in the MDA-MB-435 tumors, resulting in no change in overall tumor volume [24].
Figure 5 α-Difluoromethylornithine (DFMO) treatment affects both cell proliferation and cell death in MDA-MB-231 orthotopic tumors. (a) TUNEL analysis and (b) BrdU analysis of MDA-MB-231 orthotopic tumors from either control mice or mice treated with 2% DFMO in the drinking water beginning at the day of injection.
Meprin α expression is affected by DFMO in MDA-MB-435 cells
Since we previously showed that DFMO does not affect local tumor invasion of either MDA-MB-231 or MDA-MB-435 cells, we reasoned that DFMO was exerting its major effect on later stages of metastasis. This conclusion incorrectly led us to hypothesize that proteinases were not (or only modestly) affected by DFMO treatment. However, recent evidence suggested that proteinases, and their corresponding inhibitors, have roles in tumor progression at steps other than invasion [25], prompting a closer look.
Meprin α is a zinc-dependent endopeptidase that is normally expressed in kidney and intestinal epithelial cells. It is found as either an apical membrane-bound form or as a secreted protein. At different stages of development, meprin α is expressed in a variety of tissues. In vitro, matrix proteins such as fibronectin, laminin and collagen are meprin α substrates. In some epithelial carcinomas, meprin α secretion can occur both apically and basolaterally, and is activated by plasmin released by stromal fibroblasts. Together these data indicate that meprin α can contribute to cancer invasion and/or metastasis.
To assess the impact of DFMO treatment on meprin α, MDA-MB-435 and MDA-MB-231 cells were treated in vitro with 1 mM DFMO ± 2.5 mM putrescine (to demonstrate that the inhibition is specific for ODC) for 48 hours. Semi-quantitative RT-PCR showed that DFMO treatment had no effect on meprin α expression in MDA-MB-231 cells, while meprin α expression was decreased by greater than 50% in MDA-MB-435 cells (Fig. 6). Administration of putrescine partially reversed the effect of DFMO in MDA-MB-435 cells (Fig. 6b,d). Although not an exhaustive study, the meprin α results highlight the heterogeneity between different hormone-independent breast carcinomas with regard to their response to DFMO.
Figure 6 In vitro α-difluoromethylornithine (DFMO) treatment decreases meprin α expression in MDA-MB-435 cells, but not in MDA-MB-231 cells. End-point RT-PCR analysis of total RNA from (a) MDA-MB-231 cells or (b) MDA-MB-435 cells, either left untreated (C), treated with 1 mM DFMO for 48 hours (D) or treated with 1 mM DFMO plus 2.5 mM putrescine for 48 hours (D + P). Quantities of meprin α mRNA in (c) MDA-MB-231 cells or (d) MDA-MB-435 cells were normalized to glyceraldehyde-3-phosphate dehydrogenase levels (GADPH) – normalization to hypoxanthine phosphoribosyl transferase [HPRT] was also performed in (a). Meprin α expression levels are graphed as a percentage of control.
While there are likely to be other mechanisms involved, the difference in meprin α levels in DFMO-treated MDA-MB-231 and MDA-MB-435 cells provides at least a partial explanation for the different metastatic phenotypes of these cells in response to DFMO.
In support of our data, another inhibitor of polyamine pathway activity, SL11144, inhibited the growth of both MDA-MB-231 and MDA-MB-435 cells in vitro and of MDA-MB-231 orthotopic tumor growth in vivo [26]. In vitro cell death assays demonstrated a differential effect of SL11144 inhibition on the time line of apoptosis. Specifically, MDA-MB-435 cells exhibited significant DNA laddering at 12 hours after treatment, while MDA-MB-231 cells only showed the initial stages of laddering at 96 hours post-treatment. These results were explained by more potent activation of caspases and rapid cytochrome c release in MDA-MB-435 cells in response to the inhibitor compared with MDA-MB-231 cells. Similar to the studies reported here, SL11144 was effective against both cell lines, but the effect was different between the two lines.
Although heterogeneous responses were observed in the cell lines examined in the present study, the results highlight the potential utility of DFMO for the treatment of advanced hormone-independent breast cancers. Moreover, the finding that DFMO can significantly reduce bone metastasis warrants further studies into the utilization of polyamine pathway modulators in combination with other therapies for control of cancer metastasis.
Conclusion
DFMO, a potent inhibitor of polyamine metabolism, blocks primary tumor growth and/or metastasis to the lung and bone from hormone-independent breast carcinoma xenografts. While the effects are heterogeneous, the findings warrant follow-up into the utilization of polyamine pathway modulators in combination with other therapies for control of this particularly difficult-to-treat class of breast carcinomas.
Abbreviations
BSA = bovine serum albumin; DFMO = α-difluoromethylornithine; DMEM = Dubecco's modified Eagle's medium; GFP = green fluorescent protein; HBSS = Hanks Balanced Salt Solution; ODC = ornithine decarboxylase; PBS = phosphate-buffered saline; RT-PCR = reverse transcriptase-polymerase chain reaction; TUNEL = terminal deoxynucleotidyl transferase dUTP nick end labeling.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MMR provided Figs 1, 3, 4 and 5, and drafted the manuscript. PAP and DJD provided Fig. 2. GM and JSB helped conceive of the meprin part of the study and provided Fig. 6. SW assisted with multiple technical aspects of this study. LMD performed the high-pressure liquid chromatography to determine polyamine levels. AM, MMR and DRW conceived of the study and helped to draft the manuscript. All authors have read and approved the final version of this manuscript.
Acknowledgements
The authors would like to thank Dr Jim Griffith, Dr Michael Verderame, Dr Andra Frost and members of the Welch laboratory for helpful discussions and critical reading of this manuscript. We also thank Dr. Lalita Shevde-Samant for assistance with some of the early animal studies. This work was supported by The National Foundation for Cancer Research, DAMD-17-02-1-0541, CA-89019 and CA-87728 to DRW, by DK19691 to JSB and by CA-98237 to AM. This work was also supported by Pennsylvania State University College of Medicine Tobacco Settlement Fund Awards.
==== Refs
Gerner EW Meyskens FL Polyamines and cancer: old molecules, new understanding Nat Rev Cancer 2004 4 781 792 10.1038/nrc1454 15510159
Pendeville H Carpino N Marine JC Takahashi Y Muller M Martial JA Cleveland JL The ornithine decarboxylase gene is essential for cell survival during early murine development Molec Cell Biol 2001 21 6549 6558 10.1128/MCB.21.19.6549-6558.2001 11533243
Manni A Wechter R Verderame MF Mauger D Cooperativity between the polyamine pathway and HER-2neu in transformation of human mammary epithelial cells in culture: role of the MAPK pathway Int J Cancer 1998 76 563 570 10.1002/(SICI)1097-0215(19980518)76:4<563::AID-IJC20>3.0.CO;2-7 9590135
Manni A Wechter R Gilmour S Verderame MF Mauger D Demers LM Ornithine decarboxylase over-expression stimulates mitogen-activated protein kinase and anchorage-independent growth of human breast epithelial cells Int J Cancer 1997 70 175 182 10.1002/(SICI)1097-0215(19970117)70:2<175::AID-IJC7>3.0.CO;2-U 9009157
Glikman P Vegh I Pollina MA Mosto AH Levy CM Ornithine decarboxylase activity, prolactin blood levels, and estradiol and progesterone receptors in human breast cancer Cancer 1987 60 2237 2243 3440234
Canizares F Salinas J de las HM Diaz J Tovar I Martinez P Penafiel R Prognostic value of ornithine decarboxylase and polyamines in human breast cancer: correlation with clinicopathologic parameters Clin Cancer Res 1999 5 2035 2041 10473083
Manni A Mauger D Gimotty P Badger B Prognostic influence on survival of increased ornithine decarboxylase activity in human breast cancer Clin Cancer Res 1996 2 1901 1906 9816147
Manni A Washington S Craig L Cloud M Griffith JW Verderame MF Texter LJ Mauger D Demers LM Harms JF Effects of alpha-difluoromethylornithine on local recurrence and pulmonary metastasis from MDA-MB-435 breast cancer xenografts in nude mice Clin Exp Metastasis 2003 20 321 325 10.1023/A:1024055522067 12856719
Manni A Washington S Griffith JW Verderame MF Mauger D Demers LM Samant RS Welch DR Influence of polyamines on in vitro and in vivo features of aggressive and metastatic behavior by human breast cancer cells Clin Exp Metastasis 2002 19 95 105 10.1023/A:1014536909007 11964084
Harms JF Welch DR MDA-MB-435 human breast carcinoma metastasis to bone Clin Exp Metastasis 2003 20 327 334 10.1023/A:1024062911144 12856720
Harms JF Welch DR Samant RS Shevde LA Miele ME Babu GR Goldberg SF Gilman VR Sosnowski DM Campo DA A small molecule antagonist of the alpha-v, beta-3 integrin suppresses MDA-MB-435 skeletal metastasis Clin Exp Metastasis 2004 21 119 128 10.1023/B:CLIN.0000024763.69809.64 15168729
Welch DR Technical considerations for studying cancer metastasis in vivo Clin Exp Metastasis 1997 15 272 306 10.1023/A:1018477516367 9174129
Manni A Fischer S Franks M Washington S De AR Griffith J Demers L Verderame M Leiby B Mauger D S-adenosylmethionine decarboxylase overexpression reduces invasiveness and tumorigenicity in nude mice of MCF-7 breast cancer cells Int J Oncol 2001 19 317 323 11445845
Guo YJ Cleveland JL O'Brien TG Haploinsufficiency for Odc modifies mouse skin tumor susceptibility Cancer Res 2005 65 1146 1149 15734996
Manni A Grove R Kunselman S Aldaz M Involvement of the polyamine pathway in breast cancer progression Cancer Lett 1995 92 49 57 10.1016/0304-3835(95)03763-M 7757960
Manni A Washington S Mauger D Hackett DA Verderame MF Cellular mechanisms mediating the anti-invasive properties of the ornithine decarboxylase inhibitor α-difluoromethylornithine (DFMO) in human breast cancer cells Clin Exp Metastasis 2004 21 461 467 10.1007/s10585-004-2724-3 15672871
Gerner EW Mamont PS Restoration of the polyamine contents in rat hepatoma tissue-culture cells after inhibition of polyamine biosynthesis. Relationship with cell proliferation Eur J Biochem 1986 156 31 35 10.1111/j.1432-1033.1986.tb09544.x 3082634
Meyskens FL JrGerner EW Emerson S Pelot D Durbin T Doyle K Lagerberg W Effect of alpha-difluoromethylornithine on rectal mucosal levels of polyamines in a randomized, double-blinded trial for colon cancer prevention J Natl Cancer Inst 1998 90 1212 1218 10.1093/jnci/90.16.1212 9719082
Roodman GD Mechanisms of disease: mechanisms of bone metastasis N Engl J Med 2004 350 1655 1664 10.1056/NEJMra030831 15084698
Mundy GR Metastasis: metastasis to bone: causes, consequences and therapeutic opportunities Nat Rev Cancer 2002 2 584 593 10.1038/nrc867 12154351
Kang YB Siegel PM Shu WP Drobnjak M Kakonen SM Cordón-Cardo C Guise TA Massagué J A multigenic program mediating breast cancer metastasis to bone Cancer Cell 2003 3 537 549 10.1016/S1535-6108(03)00132-6 12842083
Guise TA Parathyroid hormone-related protein and bone metastases Cancer 1997 80 1572 1580 10.1002/(SICI)1097-0142(19971015)80:8+<1572::AID-CNCR7>3.0.CO;2-M 9362424
Yoneda T Sasaki A Mundy GR Osteolytic bone metastasis in breast cancer Breast Cancer Res Treat 1994 32 73 84 10.1007/BF00666208 7819589
Matters GL Manni A Bond JS Inhibitors of polyamine synthesis decrease the expression of the metalloproteases meprin alpha and MMP-7 in hormone-independent human breast cancer cells Clin Exptl Metastasis in press
Chambers AF Matrisian LM Changing views of the role of matrix metalloproteinases in metastasis J Natl Cancer Inst 1997 89 1260 1270 10.1093/jnci/89.17.1260 9293916
Huang Y Hager ER Phillips DL Dunn VR Hacker A Frydman B Kink JA Valasinas AL Reddy VK Marton LJ A novel polyamine analog inhibits growth and induces apoptosis in human breast cancer cells Clin Cancer Res 2003 9 2769 2777 12855657
|
16168128
|
PMC1242150
|
CC BY
|
2021-01-05 08:27:14
|
no
|
Breast Cancer Res. 2005 Aug 9; 7(5):R819-R827
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1292
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12931616813010.1186/bcr1293Research ArticleChanges in body weight and the risk of breast cancer in BRCA1 and BRCA2 mutation carriers Kotsopoulos Joanne [email protected] Olufunmilayo I 3Ghadirian Parviz 4Lubinski Jan 5Lynch Henry T 6Isaacs Claudine 7Weber Barbara 8Kim-Sing Charmaine 9Ainsworth Peter 10Foulkes William D 11Eisen Andrea 12Sun Ping 1Narod Steven A [email protected] Centre for Research in Women's Health, Women's College Hospital, University of Toronto, Toronto, ON, Canada2 Department of Nutritional Sciences, University of Toronto, ON, Canada3 Center for Clinical Cancer Genetics, University of Chicago, Chicago, IL, USA4 Epidemiology Research Unit, Research Centre, Centre Hospitalier de l'Universitaire Montréal, CHUM Hôtel Dieu and Département de Nutrition, Faculté du Médecine, Quebec, QC, Canada5 Hereditary Cancer Center, Pomeranian Medical University, Szczecin, Poland6 Department of Preventive Medicine and Public Health, Creighton University School of Medicine, Omaha, NE, USA7 Lombardi Cancer Center, Georgetown University Medical Center, Washington, DC, USA8 Abramson Family Cancer Research Institute, University of Pennsylvania, Philadelphia, PA, USA9 British Columbia Cancer Agency, Vancouver, BC, Canada10 London Regional Cancer Center, London, ON, Canada11 Departments of Medicine, Human Genetics, and Oncology, McGill University, Montréal, QC, Canada12 Sunnybrook and Women's College Health Sciences, Toronto, ON, Canada2005 19 8 2005 7 5 R833 R843 10 2 2005 31 3 2005 23 6 2005 6 7 2005 Copyright © 2005 Kotsopoulos et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Several anthropometric measures have been found to be associated with the risk of breast cancer. Current weight, body mass index, and adult weight gain appear to be predictors of postmenopausal breast cancer. These factors have been associated with a reduced risk of premenopausal breast cancer. We asked whether there is an association between changes in body weight and the risk of breast cancer in women who carry a mutation in either breast cancer susceptibility gene, BRCA1 or BRCA2.
Methods
A matched case–control study was conducted in 1,073 pairs of women carrying a deleterious mutation in either BRCA1 (n = 797 pairs) or BRCA2 (n = 276 pairs). Women diagnosed with breast cancer were matched to control subjects by year of birth, mutation, country of residence, and history of ovarian cancer. Information about weight was derived from a questionnaire routinely administered to women who were carriers of a mutation in either gene. Conditional logistic regression was used to estimate the association between weight gain or loss and the risk of breast cancer, stratified by age at diagnosis or menopausal status.
Results
A loss of at least 10 pounds in the period from age 18 to 30 years was associated with a decreased risk of breast cancer between age 30 and 49 (odds ratio (OR) = 0.47; 95% confidence interval (CI) 0.28–0.79); weight gain during the same interval did not influence the overall risk. Among the subgroup of BRCA1 mutation carriers who had at least two children, weight gain of more than 10 pounds between age 18 and 30 was associated with an increased risk of breast cancer diagnosed between age 30 and 40 (OR = 1.44, 95% CI 1.01–2.04). Change in body weight later in life (at age 30 to 40) did not influence the risk of either premenopausal or postmenopausal breast cancer.
Conclusion
The results from this study suggest that weight loss in early adult life (age 18 to 30) protects against early-onset BRCA-associated breast cancers. Weight gain should also be avoided, particularly among BRCA1 mutation carriers who elect to have at least two pregnancies.
==== Body
Introduction
The inheritance of a deleterious mutation in either of the two breast cancer susceptibility genes, BRCA1 or BRCA2, has been associated with a lifetime risk of breast cancer of 45% to 87% [1,2]. Reports of increasing penetrance among women born in recent cohorts in comparison with those born in earlier years has prompted the search for factors that may influence the risk of cancer in genetically susceptible women [2-5]. To date, both genetic and non-genetic factors have been suggested to influence breast cancer risk in BRCA1 and BRCA2 mutation carriers, and many implicate estrogen-induced stimulation as a probable contributor (reviewed in [6]). Genetic risk factors include both the type and position of the mutation [7-9], as well as the presence of specific alleles of modifying genes [10-13]. Non-genetic or environmental factors include hormonal factors, in particular those related to estrogen exposure (reviewed in [6]). Reproductive factors that modify risk in BRCA carriers include breastfeeding, parity, and oral contraceptive use (reviewed in [14]).
The worldwide prevalence of obesity is rising [15]. Evidence from animal studies suggests that positive energy balance has a growth-promoting effect on tumours [16]. Numerous epidemiological studies have evaluated the role of various anthropometric risk factors in the etiology of breast cancer (reviewed in [17]). Collectively, the evidence suggests that the effects of body mass index (BMI) and of adult weight gain on the risk of breast cancer are dependent on menopausal status at diagnosis. There appears to be an inverse relation between both BMI and weight and the risk of premenopausal breast cancer; whereas there is a positive association between body weight, BMI, and adult weight gain on the risk of breast cancer after the menopause (reviewed in [17-20]). Birthweight and adult height have been associated with an increased risk of breast cancer in both menopausal strata (reviewed in [17-20]). Weight change that occurs at the time a woman is undergoing hormonal changes (i.e. puberty, pregnancy, menopause) has also been suggested to have an effect on risk [21,22]. Although various biological mechanisms by which weight may influence breast cancer risk have been proposed (reviewed in [17]), of particular relevance is an increase in circulating endogenous sex hormones, particularly estrogen [23]. Epidemiologic observations and laboratory studies suggest that sex hormones play an important role in BRCA-carcinogenesis and the current chemopreventive options available for BRCA carriers are based on the interruption of the estrogen-signalling pathway (reviewed in [6,24]).
Studies are needed to determine if the known anthropometric risk factors for sporadic breast cancer may also influence the penetrance of breast cancer in BRCA carriers. We performed a matched case–control study to investigate whether or not there is an association between changes in body weight and the risk of breast cancer in women with a deleterious BRCA1 or BRCA2 mutation. The identification of non-genetic modifiers of risk may be useful for preventing hereditary breast cancer.
Materials and methods
Study population and design
Eligible study subjects included women who were alive and known to be carriers of deleterious mutations of the BRCA1 or BRCA2 gene. These women were identified from 41 participating centers in five countries and were participants in previous and ongoing clinical research protocols at the host institutions. All study subjects received counselling and gave their written informed consent for genetic testing.
The study was approved by the institutional review boards of the host institutions. In most cases, testing was initially offered to women who had been affected with breast or ovarian cancer. When a BRCA1 or BRCA2 mutation was identified in a proband or her relative, genetic testing was offered to other at-risk women in the family. Mutation detection was performed using a range of techniques, but all nucleotide sequences were confirmed by direct sequencing of DNA. A woman was eligible for the current study when the molecular analysis established that she was a carrier of a pathogenic mutation. Most (>95%) of the mutations identified in the study subjects were nonsense mutations, deletions, insertions, or small frameshifts.
There was information on cancer history and mutation carrier status for a total of 3,291 women who carried BRCA1 or BRCA2 mutations and who provided information on weight at ages 18, 30, and 40. Potential case subjects were selected from among the study subjects with a diagnosis of invasive breast cancer. Case subjects were excluded if they had been diagnosed with ovarian cancer (29 women) or any other form of cancer (28 women) before being diagnosed with breast cancer, or if information about their menopausal status was missing (31 women). Control subjects were women who had never had breast cancer and who were carriers of a mutation in the BRCA1 or BRCA2 gene. A subject was not eligible to be a control for a given case subject if she had had a protective bilateral mastectomy before the date of diagnosis in the case (88 women). After exclusions, there was a total of 3,115 eligible women, including 1,471 women with breast cancer (potential case subjects) and 1,644 women without breast cancer (potential controls).
A single control subject was selected for each case subject, matched according to mutation in the same gene (BRCA1 or BRCA2), year of birth (within 1 year), and the country of residence. A diagnosis of ovarian or other form of cancer in the control had to be after the year of diagnosis of breast cancer of the matched case subject for her to be eligible. In addition, the date of interview of the control was after the date of breast cancer diagnosis of the matched case. A total of 1,073 matched case–control pairs was generated for the analysis, including 797 pairs with BRCA1 mutations and 276 pairs with BRCA2 mutations. The 2,146 study subjects included in the analysis were identified from 1,534 distinct families (1.4 subjects per family). In the instance of 1,179 subjects, the subject was the only member of the family to be included. These were prevalent cases and had breast cancer before they knew their mutation status. On average, 8.8 years had passed between the subject's age at diagnosis (mean 39.8 years) and age at interview (mean 48.6 years).
Data collection
Case and control subjects completed a questionnaire that asked for relevant information regarding family history, reproductive and medical histories, and selected lifestyle factors including smoking history and use of oral contraceptives. Questionnaires were administered by each of the individual centers at the time of a clinic appointment or at their home at a later date. Interviews occurred between 1988 and 2004 for the case subjects and between 1994 and 2004 for the control subjects. Additional variables of interest included information on demography, ethnicity, and parity. Women were classified as postmenopausal if they reported natural menopause and had stopped menstruating, or if they had had a hysterectomy and bilateral oophorectomy before the diagnosis of breast cancer. Specifically for this study, the questionnaire asked for information on height (in feet and inches) and weight (in pounds). The participants were requested to think back to when they were 18 years old (about the time they graduated from high school) and to recall their weight then and subsequently at ages 30 and 40. Women were asked to report their weight at birth, their current weight, and their height, as well as the most they had ever weighed (excluding pregnancy). Only case and control data before the time of the diagnosis of breast cancer in the matched case were considered.
Anthropometric measures
We converted the reported weights from pounds to kilograms and the heights from inches to meters for BMI calculations. Variables that were created in this study included BMI (weight (kg)/height(m2)) at ages 18, 30, and 40 years, and weight change between age 18 and 30 and between 30 and 40 (calculated as the difference between the weights at the age periods being compared).
Statistical analyses
A matched case – control analysis was performed to examine the association between weight and changes in body weight, and the risk of breast cancer. Because menopausal status has been shown to modify the association between anthropometric factors and the risk of breast cancer, our analyses were stratified according to menopausal status at the time the subject received a diagnosis of breast cancer diagnosis. Birthweight, height, weight, weight gain, and BMI were compared between the case subjects and control subjects within each stratum, using a paired t-test. This test statistic was also used for all other continuous variables that were examined. The χ2 test was used to test for differences in categorical variables. The univariate odds ratios (ORs), 95% confidence intervals (CIs), and tests for linear trend were estimated by use of conditional logistic regression. A multivariate analysis was also carried out to control for the potential confounding effects of oral contraceptive use, smoking, oophorectomy, and parity. Smoking use was coded as 'ever' or 'never' smoker; oral contraceptive use was coded as 'ever' or 'never' user; oophorectomy was coded as yes or no; and parity was coded as zero, one, or two or more births. Weight change was categorized into quartiles according to the distribution of the variables among the controls.
The reference group were those women whose weight remained stable (weight gain or loss of not more than 10 pounds from baseline). The weight-loss group included women who lost at least 10 pounds. We examined the effect of weight change between ages 18 and 30 and between ages 30 and 40 among subgroups defined according to the subject's age at diagnosis of the case. This effect was further evaluated according to mutation and menopausal status. There were 26 menopausal case subjects who reported having had a hysterectomy before their breast cancer had been diagnosed but who still had intact ovaries. These 26 pairs were excluded from the subanalyses stratified by menopausal status. Odds ratios were generated for these subgroups with the matched-pair subsets. All statistical tests were two-sided. A P value of <0.05 was taken to be significant. All analyses were performed using the SAS statistical package, version 8.1 (SAS Institute, Cary, NC, USA).
Results
Study subjects
Case and control subjects were similar with regard to year of birth, year of interview, current age, mutation status, smoking history, and country of residence (Table 1). Oral contraceptives had been used by more of the case than control subjects (P = 0.04), and parity was also slightly higher in the case than control subjects (P = 0.06).
Comparison of anthropometric measures in BRCA1 or BRCA2 mutation carriers
Table 2 compares the mean values for various anthropometric measures for the cases and controls as a whole, and stratified by the menopausal status of the case subject when the breast cancer was diagnosed. Among all the study participants, case subjects weighed less at age 18 than the control subjects. Among postmenopausal women, case subjects had a lower BMI at age 18 than controls. There were no other statistically significant differences between the case and control subjects with respect to weight at birth, current height, weight, BMI, or weight gain at various ages (Table 2).
The extent of weight gain experienced by our study subjects varied according to their year of birth (Fig. 1). Those born in earlier years experienced on average less weight gain between age 18 and 30 and between 18 and 40 than women born in later years. The increase in weight by calendar year is most apparent at the ages of 30 and 40. There was also a significant difference between the mean weights at ages 18, 30, and 40 among women residing in Canada, Poland, or the USA (P = 0.0001, 0.02, and 0.02, respectively).
Changes in body weight between age 18 and 30 and risk of breast cancer in BRCA mutation carriers
To further examine the relationship between adult weight change and the risk of breast cancer, we performed univariate conditional logistic regression. The adjusted ORs were similar to the unadjusted values; therefore, only univariate results are reported here. As Table 3 shows, weight loss of at least 10 pounds between age 18 and 30 was associated with a significant reduction in breast cancer risk thereafter (OR = 0.66). Weight gain during this period was not associated with increased risk. However, stratification of the study subjects according to their age at breast cancer diagnosis indicated that changes in body weight appeared to have different effects in carriers according to whether the breast cancer was diagnosed before or after age 40. Weight loss of at least 10 pounds was associated with a significant reduction in the risk of breast cancer diagnosed between age 30 and 40 (OR = 0.47) (Table 3) but was not associated with the risk of breast cancer diagnosed after age 40.
Subgroup analyses according to BRCA mutation status showed that among women with a BRCA1 mutation, weight loss of at least 10 pounds was associated with a 65% reduction in cancer risk compared with women in the reference group (OR = 0.35) (Table 4). A modest protective effect of this degree of weight loss was also seen among BRCA2 mutation carriers, although this association did not reach statistical significance (OR = 0.88).
The mean baseline weight (weight at age 18) of the BRCA1 mutation carriers who lost more than 10 pounds was 142.5 pounds (range 115 to 230 pounds). These women experienced a mean weight loss of 18.6 pounds (range 10 to 86 pounds) between age 18 and 30. Forty percent of these women had a mean baseline weight greater than 150 pounds and 35% had a BMI greater than 25.
Changes in body weight between age 18 and 30, parity, and risk of breast cancer in BRCA mutation carriers
Because parity has been shown to modify the risk of breast cancer in carriers [25], we next examined the risk of breast cancer associated with weight gain but taking into account the possible modifying effect of parity (Table 5). Compared with those who experienced minimal changes in body weight (± 10 pounds), weight gain of greater than 10 pounds among women who had at least two full-term pregnancies was significantly associated with an increase in the risk of breast cancer (OR = 1.44). To discern whether increased parity per se was associated with weight gain, we compared mean weight gain among the carriers, according to parity. The mean weight gain across the groups was similar (data not shown). Therefore, the increased risk of breast cancer associated with parity and any weight gain is not attributable to greater weight gain among those who had higher parity. A modifying effect of parity and weight gain was not seen among women with a BRCA2 mutation (Table 5).
Discussion
We conducted our study to examine whether change in body weight modifies the risk of breast cancer among women who carry a deleterious BRCA1 or BRCA2 mutation. We found that BRCA mutation carriers who lost at least 10 pounds between age 18 and 30 had a 34% reduction in the risk of breast cancer. However, on stratification of the sample by age of breast cancer diagnosis, this protective effect was only observed among BRCA mutation carriers diagnosed between age 30 and 40 and not for those diagnosed after age 40. Although weight loss reduced the risk of breast cancer among carriers of either mutation, this association remained significant only for women with a BRCA1 mutation (OR = 0.35). A large proportion of the group who experienced weight loss had a baseline BMI of greater than 25, the BMI cut-point for the classification of overweight individuals [26]. This suggests that recommendations regarding weight loss should be targeted towards those women who are considered to be overweight at age 18.
The role of early adult weight gain and subsequent risk of breast cancer is not well defined. The majority of studies report either no association or a decrease in risk with weight gain for premenopausal women, and inconsistent results for postmenopausal women [19,21]. It has been suggested that adult weight gain may be a better measure of adiposity than BMI, because lean body mass decreases with age [27] and BMI does not distinguish between lean and fat mass; whereas changes in adult weight largely reflect changes in body fat [19,28]. Adult weight gain appears to be a consistent and independent predictor of postmenopausal breast cancer risk, particularly in women who never used hormone replacement therapy [21,29-31]. Studies of adult weight gain and the risk of premenopausal breast cancer have generally shown a reduction in risk, although two studies found no association [21,32]. In our selected study population as a whole, weight gain did not influence risk. Rather, we observed a decrease in the risk of breast cancer diagnosed between age 31 and 40 associated with weight loss in early adulthood (between age 18 and 30). Weight change that occurred between age 30 and 40 did not influence the subsequent risk of either premenopausal or postmenopausal breast cancer. Our findings suggest an important effect of weight loss in early years and the risk of early-onset breast cancer. This effect is of particular relevance to our study population, because a characteristic feature of BRCA-associated breast cancers is young age at diagnosis [33].
Our findings suggest that in BRCA carriers, changes in body weight throughout early adult life may have a more important influence on the risk of early-onset breast cancer than current weight or BMI [21]. The magnitude of the decreased risk associated with weight loss compared with those women whose weight remained stable was relatively large (OR = 0.47) (see Table 3). After stratification by mutation status, the protective effect of weight loss between age 18 and 30 was seen to be less strong among women with a BRCA2 mutation. These findings suggest that the timing of weight loss may play a more important role in BRCA1-associated than in BRCA2-associated carcinogenesis, though the lack of a significant finding for the latter group might also be attributable to a smaller sample size. The effect may be of greater importance for women belonging to more recent birth cohorts, since there appears to be a greater increase in average weight at ages 30 and 40 with each decade (see Fig. 1).
We also found that in the subgroup of BRCA1 mutation carriers who gained 10 pounds or more and who had at least two full-term pregnancies, there was a 44% increase in their risk of breast cancer. The modifying effects of both parity and weight gain were not observed for women with a BRCA2 mutation. The number of births did not influence the amount of weight gain experienced by either the case or the control subjects, providing confirmation that weight gain is not a surrogate for parity or vice versa. Although pregnancy itself offers long-term protection against postmenopausal breast cancer in the general population, significant weight gain during pregnancy has been associated with an increased risk of developing breast cancer after the menopause [34]. We have reported elsewhere that parity is a risk factor for breast cancer in BRCA2 carriers but not in BRCA1 carriers [25].
Ballard-Barbash proposed that weight change that occurs during periods of noticeable hormonal change (i.e. menarche, pregnancy, and menopause) may be attributed to host metabolic factors that may also influence breast cancer risk [21]. In addition, weight gain may result in differing biological effects depending on the body fat distribution [21]. Weight gain during pregnancy is characterized by an increase in central body fat deposition [35]. The physiological consequences of upper or central body fat localization include altered ovarian hormone and glucose metabolism, as well as insulin resistance and hyperinsulinemia, all of which may increase breast cancer risk [21,36]. This pattern of fat distribution has been suggested to pose a higher risk of breast cancer, independent of weight [22,37].
Only two studies have evaluated the association between anthropometric risk factors or physical activity and the risk of breast cancer in BRCA1 and/or BRCA2 carriers [4,38]. King and colleagues recently reported that a healthy weight defined at menarche and at age 21, as well as physical activity during adolescence, were associated with a significant delay in the age of onset of breast cancer in BRCA1 and BRCA2 carriers; however, such an effect could be attributable to either weight gain increasing the risk of early-onset breast cancer or to weight gain protecting against late-onset breast cancer [4]. An earlier study of 46 BRCA1 carriers found no significant effect of current BMI on the age at disease onset; however, the sample size was small [38].
The role of sex hormones in the etiology of breast cancer has been well established [23]. It is generally agreed that increasing levels of circulating estrogen are a determinant of obesity-associated breast cancer in postmenopausal women [39]. In contrast, most investigations of premenopausal women report an inverse association between weight (or BMI or weight gain) and the risk of breast cancer. The epidemiological evidence suggests a positive association between these anthropometric variables and the risk of postmenopausal breast cancer (reviewed in [17]). The primary hypothesis underlying this relation between menopausal status and the risk of breast cancer is believed to involve an alteration in the source and levels of endogenous sex hormones [19,40]. Before menopause, the ovaries are the primary site of endogenous hormone production. Since obesity has been shown to induce chronic anovulatory cycles and subsequently lower serum estrogen [41] and progesterone levels [42], a decrease in hormone exposure is believed to be the primary mechanism by which overweight women may be protected against premenopausal breast cancer [43]. Extraglandular aromatization of androstenedione to estrone occurs in the adipose tissue and is the primary source of estradiol in postmenopausal women [39]. This conversion of androgens and subsequent increase in estrogen levels has been shown to be directly proportional to the amount of adipose tissue [44] and the induction of aromatase activity which may possibly enhance estrogen production in adipose tissue [45]. In contrast, among BRCA carriers, weight gain did not affect the risk of breast cancer.
Other metabolic consequences of obesity, more specifically central adiposity, that have been suggested to be factors in the development of breast cancer include hyperinsulinemia and insulin resistance, as well as elevated levels of glucose and triglycerides [46-50]. Obesity has also been shown to increase testosterone [51,52] and leptin levels, [53,54] and to depress sex-hormone-binding globulin concentrations. This globulin is the predominant carrier of estradiol levels in both premenopausal and postmenopausal women and is the primary protein responsible for binding and inactivating estradiol [41,55,56]. Therefore, reducing the concentration of sex-hormone-binding globulin may lead to an increase in the amount of unbound, free estradiol.
High concentrations of circulating insulin-like growth factor1 (IGF-1) appears to be a risk factor for premenopausal breast cancer in the general population, yet no such relation has been observed for postmenopausal breast cancer [57]. Studies have shown that both insulin and IGF-1 exert a mitogenic effect by stimulating cell proliferation and inhibiting apoptosis of breast cancer cells [58,59]. More importantly, it has been suggested that IGF-1 may also work synergistically with other growth factors and hormones, including estrogen, to further promote cell proliferation [60]. Although both BMI and IGF-1 levels are suggested to influence breast cancer risk, studies have generally shown no association or an inverse association between BMI and circulating IGF-1 levels [60].
Both birthweight [61,62] and height [63] are positively associated with IGF-1 levels. The evidence, primarily from cohort studies, supports a positive association between birthweight and the risk of breast cancer (reviewed in [64]) suggesting that prenatal events may influence later risk. Adult height has also been shown to positively predict the risk of breast cancer in both pre- and postmenopausal women [18,32]. In our study, there was no significant difference in birthweight between the cases and controls and it seems unlikely that this variable influences risk in BRCA mutation carriers. Current height was not associated with the risk of breast cancer, and this observation is in agreement with a pooled analysis of 52 epidemiological studies whereas height did not modify risk in women who had one or more affected first-degree relatives in comparison with women who had no affected relatives [65].
A potential drawback of our study was the use of self-reported risk factor data, which may have introduced measurement error and led to a spurious result or attenuation of results. However, validation studies have shown that current and recalled self-reported weight and height measurements are highly correlated with measured data [66-72]. Self-reporting many years prior has still been shown to retain a high degree of validity [19]. Our data was collected on average 9 years after the breast cancer diagnosis of the case, and 30 years after age 18 (the first weight reported). There is a potential for recall bias but there is no evidence of this in Table 2. The mean weights at each reported age were similar and the differences were not significant. In fact, the reported weight at age 18 was less for cases than controls (we might expect recall bias to generate the opposite result). Also, the dissimilar results for BRCA1 and BRCA2 carriers argues against recall bias.
Despite the primary limitation of recall bias and other inherent limitations associated with the use of case–control studies, the primary strength of our study is the large sample of known BRCA mutation carriers. This study involved 1,073 matched pairs selected from a total of approximately 3,291 documented mutation carriers and is by far the largest study addressing the role of anthropometric measures on the risk of hereditary breast cancer. Our matching strategy and exclusion criteria resulted in case and control groups that were similar in most respects. We believe that our study participants are representative of women who have had BRCA mutations identified during the course of genetic counselling. Our study was based on known mutation carriers and included patients from numerous participating centers and of different ethnic backgrounds.
Conclusion
Our findings suggest that weight loss in early adult life (and not weight per se) decreases the risk of BRCA-associated breast cancer diagnosed at an early age. More specifically, the period between age 18 and 30 years appears to be a critical one when weight gain should be avoided in mutation carriers. The effect may be greatest in BRCA1 carriers experiencing at least two full-term pregnancies, but further study is necessary to confirm this subgroup analysis. The maintenance of a healthy weight during early adult life represents a potentially modifiable risk factor in hereditary breast cancer syndromes.
Abbreviations
BRCA1 = breast cancer susceptibility gene 1; BRCA2 = breast cancer susceptibility gene 2; CI = confidence interval; IGF-1 = insulin-like growth factor 1; OR = odds ratio.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SAN conceived and designed the study. JK drafted the manuscript and helped with the analysis. PS performed the statistical analysis. OIO, PG, JL, HTL, CI, BW, CK-S, PA, WDF, and AI coordinated study activities for their centers and helped with the preparation of the manuscript. All authors read and approved the final version of the manuscript.
Acknowledgements
We thank B Rosen, S Randall, M Daly, N Tung, H Saal, E Friedman, B Karlan, E Warner, M Osborne, D Fishman, C Eng, M Wood, W McKinnon, S Merajver, D Gilchrist, J Weitzel, G Evans, D Provencher, A Chudley, E Lemire, and J McLennan for submission of data on their patients.
Figures and Tables
Figure 1 Weight at various ages among BRCA mutation carriers according to year of birth.
Table 1 Comparison of subjects carrying BRCA mutations
Variable Control subjects (n = 1,073) Case subjects (n = 1,073) Pa
Age (years) at interview, no. (%)
≤ 30 17 (1.6) 26 (2.4)
31–40 219 (20.4) 242 (22.5)
41–50 415 (38.7) 386 (36.0)
51–60 277 (25.8) 264 (24.6)
≥ 61 145 (13.5) 155 (14.5) 0.32
Age (years) at interview, mean (SD) 47.9 (10.6) 48.6 (10.6) 0.25
Date of birth, mean year 1951.8 1951.2 0.14
Year of interview, mean (range) 1999.7 (1995–2004) 1999.8 (1999–2004) 0.18
Age (years) at diagnosis of breast cancer, no. (%)
≤ 30 NA 107 (10.0)
31–40 478 (44.5)
41–50 371 (34.5)
≥ 51 117 (10.9)
Age (years) at diagnosis of breast cancer, mean (SD) 39.8 (8.3)
Mutation, %
BRCA1 74.3 74.3
BRCA2 25.7 25.7
Parity
No. (%) parous 844 (79.6) 879 (82.8) 0.06
Parity, mean (SD) 1.9 (1.4) 1.9 (1.3) 0.18
No. (%) who ever used oral contraceptives 662 (62.6) 710 (66.8) 0.04
No. (%) who ever smoked 465 (43.3) 471 (43.9) 0.79
Country of residence at time of testing, no. (%)
Canada 420 (39.1) 420 (39.1)
Israel 20 (1.9) 20 (1.9)
UK 8 (0.8) 8 (0.8)
Poland 189 (17.6) 189 (17.6)
USA 436 (40.6) 436 (40.6)
aAll P values are univariate and were derived using Student's t-test for continuous variables and the χ2 test for categorical variables. NA, not applicable; SD, standard deviation.
Table 2 Anthropometric variables in women with a deleterious BRCA1 or BRCA2 mutationa
All Premenopausal Postmenopausal
Variable Cases n = 1073 Controls n = 1073 P Cases n = 817 Controls n = 817 P Cases n = 256 Controls n = 256 P
Weight (pounds)
At birth 7.1 (1.3) 7.1 (1.3) 0.67 NA NA NA NA NA NA
At age 18 120.6 (17.2) 122.3 (19.1) 0.03 121.5 (17.1) 122.9 (19.2) 0.11 117.8 (17.3) 120.4 (18.5) 0.10
At age 30c 130.9 (17.3) 131.2 (18.5) 0.75 131.9 (17.1) 132.2 (18.5) 0.87 127.9 (17.4) 128.6 (18.6) 0.69
At age 40d 137.7 (23.4) 137.7 (24.3) 0.99 139.7 (22.8) 138.7 (25.0) 0.63 135.3 (23.9) 136.5 (23.4) 0.59
Current height (inches) 64.2 (2.6) 64.3 (2.6) 0.44 64.3 (2.6) 64.5 ± 2.6 0.50 63.9 ± 2.7 63.7 ± 2.7 0.10
Body mass indexb
At age 18 (kg/m2) 20.57 (2.8) 20.77 (3.0) 0.10 20.65 (2.8) 20.74 (3.0) 0.52 20.31 (2.7) 20.88 (2.9) 0.03
At age 30 (kg/m2)c 22.35 (3.4) 22.32 (3.7) 0.88 22.44 (3.4) 22.33 (3.9) 0.56 22.08 (3.4) 22.31 (3.1) 0.43
At age 40 (kg/m2)d 23.77 (3.9) 23.73 (3.9) 0.88 24.06 (3.8) 23.67 (4.1) 0.26 23.42 (4.0) 23.80 (3.7) 0.29
Weight gain (pounds)
From age 18 to 30c 10.5 (15.2) 9.4 (17.0) 0.14 10.6 (15.1) 9.8 (18.2) 0.35 10.1 (15.7) 8.2 (13.0) 0.23
From age 30 to 40d 9.4 (12.5) 9.2 (13.3) 0.82 10.6 (12.5) 9.1 (15.3) 0.23 7.9 (12.3) 8.2 (10.5) 0.23
From age 18 to 40d 18.6 (19.5) 18.1 (20.4) 0.71 19.9 (20.2) 19.0 (22.1) 0.61 17.0 (18.6) 17.1 (18.2) 0.98
The subjects were women who did (case subjects) or did not (control subjects) have a diagnosis of invasive breast cancer. aP values are univariate and were derived using Student's t-test and include both BRCA1 and BRCA2 mutation carriers. Other values are means (standard deviations). bExcludes subjects with missing data on current height. cCalculated for pairs in which case subjects were diagnosed at age >30 years. dCalculated for pairs in which case subjects were diagnosed at age >40 years. NA, not applicable.
Table 3 Weight change and subsequent cancer risk: subjects stratified by their age at diagnosis of cancer
Weight change between age 18 and 30 years Cases (number) Controls (number) ORa (95% CI) P P for trend
In all subjectsb 966 966
Loss of at least 10 pounds 53 81 0.66 (0.46–0.93) 0.03
Loss of <10 to gain of ≤ 10 pounds 536 542 1 (referent)
Gain of 10 to ≤ 20 pounds 227 190 1.19 (0.96–1.49) 0.12
Gain of > 20 pounds 150 135 1.00 (0.77–1.30) 0.99 0.46
According to case subjects' age at diagnosis
>30 to ≤ 40 yearsc 478 478
Loss of at least 10 pounds 23 49 0.47 (0.28–0.79) 0.005
Loss of <10 to gain of ≤ 10 poundsd 255 254 1 (referent)
Gain of 10 to ≤ 20 pounds 112 89 1.25 (0.91–1.71) 0.17
Gain of >20 pounds 88 86 1.03 (0.72–1.47) 0.88 0.48
>40 yearsd 488 488
Loss of at least 10 pounds 30 32 0.97 (0.52–1.65) 0.91
Loss of <10 to gain of ≤ 10 poundsd 281 288 1 (referent)
Gain of 10 to ≤ 20 pounds 115 101 1.16 (0.85–1.59) 0.36
Gain of >20 pounds 62 67 0.95 (0.64–1.43) 0.82 0.75
Subjects were women with a deleterious mutation in BRCA1 or BRCA2 who did (case subjects) or did not (control subjects) receive a diagnosis of cancer. aAll odds ratios (ORs) were derived using univariate conditional logistic regression. bExcludes case subjects diagnosed at age ≤ 30 years. cExcludes case subjects diagnosed at age ≤ 30 and >40 years. dExcludes case subjects diagnosed at ≤ 40 years. CI, confidence interval.
Table 4 Weight change and subsequent cancer risk: subjects stratified by their BRCA mutation
Weight change between age 18 and 30 years Cases (n) Controls (n) ORa (95% CI) P P for trend
In BRCA1 mutation carriers 370 370
Loss of at least 10 pounds 13 38 0.35 (0.18–0.67) 0.002
Loss of <10 to gain of ≤ 10 poundsd 188 189 1 (referent)
Gain of 10 to ≤ 20 pounds 93 72 1.29 (0.91–1.83) 0.15
Gain of >20 pounds 76 71 1.09 (0.73–1.62) 0.67 0.34
In BRCA2 mutation carriers 108 108
Loss of at least 10 pounds 10 11 0.88 (0.35–2.23) 0.78
Loss of <10 to gain of ≤ 10 poundsd 67 65 1 (referent)
Gain of 10 to ≤ 20 pounds 19 17 1.08 (0.50–2.35) 0.84
Gain of >20 pounds 12 15 0.77 (0.33–1.81) 0.55 0.70
Subjects were women with a deleterious mutation in BRCA1 or BRCA2 who did (case subjects) or did not (matched control subjects) receive a diagnosis of breast cancer at age 30 to 39 years. aAll odds ratios (ORs) were derived using univariate conditional logistic regression. CI, confidence interval.
Table 5 Weight change and subsequent cancer risk: subjectsa stratified by BRCA mutation and parity
Weight change between age 18 and 30 years Cases (number) Controls (number) ORc (95% CI) P P for trendd
In BRCA1 mutation carriers
Loss of <10 to gain of ≤ 10 poundsb 188 189 1 (referent)
Gain of >10 pounds
Parity = 0 24 28 0.88 (0.50–1.55) 0.66
Parity = 1 26 27 0.94 (0.52–1.72) 0.85
Parity ≥ 2 117 84 1.44 (1.01–2.04) 0.04 0.16
Parity unknown 2 4
In BRCA2 mutation carriers
Loss of <10 to gain of ≤ 10 poundsb 67 65 1 (referent)
Gain of >10 pounds
Parity = 0 6 4 1.44 (0.40–5.13) 0.58
Parity = 1 7 6 1.10 (0.33–3.72) 0.87
Parity ≥ 2 18 22 0.73 (0.33–1.64) 0.45 0.31
Parity unknown 0 0
aWhose cancer was diagnosed when they were 30 to 39 years old. bA negative number indicates weight loss. cAll ORs were derived using univariate conditional logistic regression. dExcludes subjects in the weight-gain ≤ -10 group.
==== Refs
Ford D Easton DF Stratton M Narod S Goldar D Devilee P Bishop DT Weber B Lenoir G Chang-Claude J Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families. The Breast Cancer Linkage Consortium Am J Hum Genet 1998 62 676 689 9497246 10.1086/301749
Antoniou A Pharoah PD Narod S Risch HA Eyfjord JE Hopper JL Loman N Olsson H Johansson O Borg A Average risks of breast and ovarian cancer associated with BRCA1 or BRCA2 mutations detected in case Series unselected for family history: a combined analysis of 22 studies Am J Hum Genet 2003 72 1117 1130 12677558 10.1086/375033
Narod S Lynch H Conway T Watson P Feunteun J Lenoir G Increasing incidence of breast cancer in family with BRCA1 mutation Lancet 1993 341 1101 1102 8097005 10.1016/0140-6736(93)92468-9
King MC Marks JH Mandell JB Breast and ovarian cancer risks due to inherited mutations in BRCA1 and BRCA2 Science 2003 302 643 646 14576434 10.1126/science.1088759
Foulkes WD Brunet JS Wong N Goffin J Chappuis PO Change in the penetrance of founder BRCA1/2 mutations? A retrospective cohort study J Med Genet 2002 39 407 409 12070248 10.1136/jmg.39.6.407
Narod SA Modifiers of risk of hereditary breast and ovarian cancer Nat Rev Cancer 2002 2 113 123 12635174 10.1038/nrc726
Gayther SA Warren W Mazoyer S Russell PA Harrington PA Chiano M Seal S Hamoudi R van Rensburg EJ Germline mutations of the BRCA1 gene in breast and ovarian cancer families provide evidence for a genotype-phenotype correlation Nat Genet 1995 11 428 433 7493024 10.1038/ng1295-428
Gayther SA Mangion J Russell P Seal S Barfoot R Ponder BA Stratton MR Easton D Variation of risks of breast and ovarian cancer associated with different germline mutations of the BRCA2 gene Nat Genet 1997 15 103 105 8988179 10.1038/ng0197-103
Thompson D Easton D Variation in cancer risks, by mutation position, in BRCA2 mutation carriers Am J Hum Genet 2001 68 410 419 11170890 10.1086/318181
Rebbeck TR Kantoff PW Krithivas K Neuhausen S Blackwood MA Godwin AK Daly MB Naraod SA Garber JE Lynch HT Modification of BRCA1-associated breast cancer risk by the polymorphic androgen-receptor CAG repeat Am J Hum Genet 1999 64 1371 1377 10205268 10.1086/302366
Rebbeck TR Wang Y Kantoff PW Krithivas K Neuhausen SL Godwin AK Daly MB Narod SA Brunet JS Vesprini D Modification of BRCA1- and BRCA2-associated breast cancer risk by AIB1 genotype and reproductive history Cancer Res 2001 61 5420 5424 11454686
Wang WW Spurdle AB Kolachana P Bove B Moden B Ebbers SM Suthers G Tucker MA Kaufman DJ Doody MM A single nucleotide polymorphism in the 5' untranslated region of RAD51 and risk of cancer among BRCA1/2 mutation carriers Cancer Epidemiol Biomarkers Prev 2001 10 955 960 11535547
Levy-Lahad E Lahad A Eisenberg S Dagan E Paperna T Kasinetz L Catane R Kaufman B Beller U Renbaum B A single nucleotide polymorphism in the RAD51 gene modifies cancer risk in BRCA2 but not BRCA1 carriers Proc Natl Acad Sci USA 2001 98 3232 3236 11248061 10.1073/pnas.051624098
Narod SA Offit K Prevention and management of hereditary breast cancer J Clin Oncol 2005 23 1656 1663 15755973 10.1200/JCO.2005.10.035
International Obesity Task Force
Tannenbaum A The dependence of tumor formation on the composition of the calorie-restricted diet as well as on the degree of restriction. 1945 Nutrition 1996 12 653 654 8878181
Friedenreich CM Review of anthropometric factors and breast cancer risk Eur J Cancer Prev 2001 10 15 32 11263588 10.1097/00008469-200102000-00003
van den Brandt PA Spiegelman D Yaun SS Adami HO Beeson L Folsom AR Fraser G Goldbohm RA Graham S Kushi L Pooled analysis of prospective cohort studies on height, weight, and breast cancer risk Am J Epidemiol 2000 152 514 527 10997541 10.1093/aje/152.6.514
Vainio H Bianchini F (Eds) Weight Control and Physical Activity 2002 6 Lyons: IARC Press World Health Organization, International Agency for Research on Cancer: IARC Handbooks of Cancer Prevention
Berglund G Riboli EL Anthropometry, physical activity and cancer of the breast and colon Nutrition and Lifestyle: Opportunities for Cancer Prevention 2002 Lyons: IARC Scientific Publications 231 241
Ballard-Barbash R Anthropometry and breast cancer. Body size – a moving target Cancer 1994 74 3 Suppl 1090 1100 8039144
Kumar NB Lyman GH Allen K Cox CE Schapira DV Timing of weight gain and breast cancer risk Cancer 1995 76 243 249 8625099
Bernstein L Ross RK Endogenous hormones and breast cancer risk Epidemiol Rev 1993 15 48 65 8405212
Narod SA Hormonal prevention of hereditary breast cancer Ann N Y Acad Sci 2001 952 36 43 11795442
Cullinane CA Lubinski J Neuhausen SL Ghadirian P Lynch HT Isaacs C Weber B Moller P Offit K Kim-Sing C Effect of pregnancy as a risk factor for breast cancer in BRCA1/BRCA2 mutation carriers Int J Cancer 2005
WHO Expert Committee Physical Status The Use and Interpretation of Anthropometry 1995 Geneva: World Health Organization
Forbes GB Reina JC Adult lean body mass declines with age: some longitudinal observations Metabolism 1970 19 653 663 5459997 10.1016/0026-0495(70)90062-4
Ziegler RG Anthropometry and breast cancer J Nutr 1997 127 5 Suppl 924S 928S 9164265
Huang Z Hankinson SE Colditz GA Stampfer MJ Hunter DJ Manson JE Hennekens CH Rosner B Speizer FE Willett WC Dual effects of weight and weight gain on breast cancer risk JAMA 1997 278 1407 1411 9355998 10.1001/jama.278.17.1407
Trentham-Dietz A Newcomb PA Egan KM Titmus-Ernstoff L Baron JA Storer BE Stampfer M Willett WC Weight change and risk of postmenopausal breast cancer (United States) Cancer Causes Control 2000 11 533 542 10880035 10.1023/A:1008961931534
Feigelson HS Jonas CR Teras LR Thun MJ Calle EE Weight gain, body mass index, hormone replacement therapy, and postmenopausal breast cancer in a large prospective study Cancer Epidemiol Biomarkers Prev 2004 13 220 224 14973094
Cold S Hansen S Overvad K Rose C A woman's build and the risk of breast cancer Eur J Cancer 1998 34 1163 1174 9849474 10.1016/S0959-8049(97)10167-8
Frank TS Deffenbaugh AM Reid JE Hulick M Ward BE Lingenfelter B Gumpper KL Scholl T Tavtigian SV Pruss DR Clinical characteristics of individuals with germline mutations in BRCA1 and BRCA2: analysis of 10,000 individuals J Clin Oncol 2002 20 1480 1490 11896095 10.1200/JCO.20.6.1480
Hilakivi-Clarke LA Luoto R Kinnunen T Huttunen T Gissler M Koskenvuo M Hemminki E Pregnancy weight gain and mother's breast cancer risk 93rd Annual Meeting of the American Association for Cancer Research, San Francisco, CA 2002
Mueller WH The changes with age of the anatomical distribution of fat Soc Sci Med 1982 16 191 196 6980468 10.1016/0277-9536(82)90022-3
Stoll BA Timing of weight gain in relation to breast cancer risk Ann Oncol 1995 6 245 248 7612489
Schapira DV Kumar NB Lyman GH Cox CE Abdominal obesity and breast cancer risk Ann Intern Med 1990 112 182 186 2297194
Chang-Claude J Becher H Eby N Gastert G Wahrendorf J Hamann U Modifying effect of reproductive risk factors on the age at onset of breast cancer for German BRCA1 mutation carriers J Cancer Res Clin Oncol 1997 123 272 279 9201250
Kirschner MA Ertel N Schneider G Obesity, hormones, and cancer Cancer Res 1981 41 3711 3717 7260928
Carroll KK Obesity as a risk factor for certain types of cancer Lipids 1998 33 1055 1059 9870899
Potischman N Swanson CA Siiteri P Hoover RN Reversal of relation between body mass and endogenous estrogen concentrations with menopausal status J Natl Cancer Inst 1996 88 756 758 8637031
Henderson BE Ross RK Judd HL Krailo MD Pike MC Do regular ovulatory cycles increase breast cancer risk? Cancer 1985 56 1206 1208 4016708
Pike MC Krailo MD Henderson BE Casagrande JT Hoel DG 'Hormonal' risk factors, 'breast tissue age' and the age-incidence of breast cancer Nature 1983 303 767 770 6866078 10.1038/303767a0
MacDonald PC Edman CD Hemsell DL Porter JC Siiteri PK Effect of obesity on conversion of plasma androstenedione to estrone in postmenopausal women with and without endometrial cancer Am J Obstet Gynecol 1978 130 448 455 629289
Bruning PF Bonfrer JM Hart AA Non-protein bound oestradiol, sex hormone binding globulin, breast cancer and breast cancer risk Br J Cancer 1985 51 479 484 4038881
Kaaks R Nutrition, hormones, and breast cancer: is insulin the missing link? Cancer Causes Control 1996 7 605 625 8932921 10.1007/BF00051703
Stoll BA Perimenopausal weight gain and progression of breast cancer precursors Cancer Detect Prev 1999 23 31 36 9892988 10.1046/j.1525-1500.1999.00063.x
Del Giudice ME Fantus IG Ezzat S McKeown-Eyssen G Page D Goodwin PJ Insulin and related factors in premenopausal breast cancer risk Breast Cancer Res Treat 1998 47 111 120 9497099 10.1023/A:1005831013718
Campagnoli C Biglia N Belforte P Botta D Pedrini E Sismondi P Post-menopausal breast cancer risk: oral estrogen treatment and abdominal obesity induce opposite changes in possibly important biological variables Eur J Gynaecol Oncol 1992 13 139 154 1587291
Okosun IS Liao Y Rotimi CN Prewitt TE Cooper RS Abdominal adiposity and clustering of multiple metabolic syndrome in White, Black and Hispanic Americans Ann Epidemiol 2000 10 263 270 10942873 10.1016/S1047-2797(00)00045-4
Evans DJ Hoffmann RG Kalkhoff RK Kissebah AH Relationship of androgenic activity to body fat topography, fat cell morphology, and metabolic aberrations in premenopausal women J Clin Endocrinol Metab 1983 57 304 310 6345569
Kaye SA Folsom AR Soler JT Prineas RJ Potter JD Associations of body mass and fat distribution with sex hormone concentrations in postmenopausal women Int J Epidemiol 1991 20 151 156 2066214
Rose DP Gilhooly EM Nixon DW Adverse effects of obesity on breast cancer prognosis, and the biological actions of leptin (review) Int J Oncol 2002 21 1285 1292 12429979
Hu X Juneja SC Maihle NJ Cleary MP Leptin – a growth factor in normal and malignant breast cells and for normal mammary gland development J Natl Cancer Inst 2002 94 1704 1711 12441326
Verkasalo PK Thomas HV Appleby PN Davey GK Key TJ Circulating levels of sex hormones and their relation to risk factors for breast cancer: a cross-sectional study in 1092 pre- and postmenopausal women (United Kingdom) Cancer Causes Control 2001 12 47 59 11227925 10.1023/A:1008929714862
Moore JW Key TJ Bulbrook RD Clark GM Allen DS Wang DY Pike MC Sex hormone binding globulin and risk factors for breast cancer in a population of normal women who had never used exogenous sex hormones Br J Cancer 1987 56 661 666 3426933
Renehan AG Zwahlen M Minder C O'Dwyer ST Shalet SM Egger M Insulin-like growth factor (IGF)-I, IGF binding protein-3, and cancer risk: systematic review and meta-regression analysis Lancet 2004 363 1346 1353 15110491 10.1016/S0140-6736(04)16044-3
Werner H LeRoith D The role of the insulin-like growth factor system in human cancer Adv Cancer Res 1996 68 183 223 8712068
Khandwala HM McCutcheon IE Flyvbjerg A Friend KE The effects of insulin-like growth factors on tumorigenesis and neoplastic growth Endocr Rev 2000 21 215 244 10857553 10.1210/er.21.3.215
Yu H Rohan T Role of the insulin-like growth factor family in cancer development and progression J Natl Cancer Inst 2000 92 1472 1489 10995803 10.1093/jnci/92.18.1472
Klauwer D Blum WF Hanitsch S Rascher W Lee PD Kiess W IGF-I, IGF-II, free IGF-I and IGFBP-1, -2 and -3 levels in venous cord blood: relationship to birthweight, length and gestational age in healthy newborns Acta Paediatr 1997 86 826 833 9307161
Ogilvy-Stuart AL Hands SJ Adcock CJ Holly JM Matthews DR Mohamed-Ali V Yudkin JS Wilkinson AR Dunger DB Insulin, insulin-like growth factor I (IGF-I), IGF-binding protein-1, growth hormone, and feeding in the newborn J Clin Endocrinol Metab 1998 83 3550 3557 9768663 10.1210/jc.83.10.3550
Juul A Bang P Hertel NT Main K Dalgaard P Jorgensen K Muller J Hall K Skakkebaek NE Serum insulin-like growth factor-I in 1030 healthy children, adolescents, and adults: relation to age, sex, stage of puberty, testicular size, and body mass index J Clin Endocrinol Metab 1994 78 744 752 8126152 10.1210/jc.78.3.744
Okasha M McCarron P Gunnell D Smith GD Exposures in childhood, adolescence and early adulthood and breast cancer risk: a systematic review of the literature Breast Cancer Res Treat 2003 78 223 276 12725422 10.1023/A:1022988918755
Collaborative Group on Hormonal Factors in Breast Cancer Familial breast cancer: collaborative reanalysis of individual data from 52 epidemiological studies including 58,209 women with breast cancer and 101,986 women without the disease Lancet 2001 358 1389 1399 11705483 10.1016/S0140-6736(01)06524-2
Stunkard AJ Albaum JM The accuracy of self-reported weights Am J Clin Nutr 1981 34 1593 1599 7270483
Weaver TW Kushi LH McGovern PG Potter JD Rich SS King RA Whitbeck J Greenstein J Steller TA Validation study of self-reported measures of fat distribution Int J Obes Relat Metab Disord 1996 20 644 650 8817358
Rimm EB Stampfer MJ Colditz GA Chute CG Litin LB Willett WC Validity of self-reported waist and hip circumferences in men and women Epidemiology 1990 1 466 473 2090285
Troy LM Hunter DJ Manson JE Colditz GA Stampfer MJ Willett WC The validity of recalled weight among younger women Int J Obes Relat Metab Disord 1995 19 570 572 7489028
Must A Willett WC Dietz WH Remote recall of childhood height, weight, and body build by elderly subjects Am J Epidemiol 1993 138 56 64 8333427
Stevens J Keil JE Waid LR Gazes PC Accuracy of current, 4-year, and 28-year self-reported body weight in an elderly population Am J Epidemiol 1990 132 1156 1163 2260547
Casey VA Dwyer JT Coleman KA Krall EA Gardner J Valadian I Accuracy of recall by middle-aged participants in a longitudinal study of their body size and indices of maturation earlier in life Ann Hum Biol 1991 18 155 166 2024949
|
16168130
|
PMC1242151
|
CC BY
|
2021-01-04 16:54:51
|
no
|
Breast Cancer Res. 2005 Aug 19; 7(5):R833-R843
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1293
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12941616812610.1186/bcr1294Research ArticleCelecoxib analogues disrupt Akt signaling, which is commonly activated in primary breast tumours Kucab Jill E [email protected] Cathy [email protected] Ching-Shih [email protected] Jiuxiang [email protected] C Blake [email protected] Maggie [email protected] David [email protected] Erika [email protected] Joanne [email protected] Michael [email protected] Sandra E [email protected] British Columbia Research Institute for Children's and Women's Health, Department of Pediatrics, University of British Columbia, Vancouver, British Columbia, Canada2 Division of Medical Chemistry and Pharmacognosy, The Ohio State University, Columbus, Ohio, USA3 Genetic Pathology Evaluation Centre, Vancouver Hospital and Health Sciences Centre and BC Cancer Agency, Vancouver, British Columbia, Canada4 Department of Anatomy, Faculty of Medicine, University of British Columbia, Vancouver, British Columbia, Canada5 Division of Experimental Medicine, Department of Medicine and Department of Oncology, McGill University, Montreal, Quebec2005 1 8 2005 7 5 R796 R807 21 10 2004 7 1 2005 20 6 2005 5 7 2005 Copyright © 2005 Kucab et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Phosphorylated Akt (P-Akt) is an attractive molecular target because it contributes to the development of breast cancer and confers resistance to conventional therapies. Akt also serves as a signalling intermediate for receptors such as human epidermal growth factor receptor (HER)-2, which is overexpressed in 30% of breast cancers; therefore, inhibitors to this pathway are being sought. New celecoxib analogues reportedly inhibit P-Akt in prostate cancer cells. We therefore examined the potential of these compounds in the treatment of breast cancer. The analogues were characterized in MDA-MB-453 cells because they overexpress HER-2 and have very high levels of P-Akt.
Methods
To evaluate the effect of the celecoxib analogues, immunoblotting was used to identify changes in the phosphorylation of Akt and its downstream substrates glycogen synthase kinase (GSK) and 4E binding protein (4EBP-1). In vitro kinase assays were then used to assess the effect of the drugs on Akt activity. Cell death was evaluated by poly(ADP-ribose) polymerase cleavage, nucleosomal fragmentation and MTS assays. Finally, tumour tissue microarrays were screened for P-Akt and HER-2 expression.
Results
OSU-03012 and OSU-O3013 inhibited P-Akt and its downstream signalling through 4EBP-1 and GSK at concentrations well below that of celecoxib. Disruption of P-Akt was followed by induction of apoptosis and more than 90% cell death. We also noted that the cytotoxicity of the celecoxib analogues was not significantly affected by serum. In contrast, the presence of 5% serum protected cells from celecoxib induced death. Thus, the structural modification of the celecoxib analogues increased P-Akt inhibition and enhanced the bioavailability of the drugs in vitro. To assess how many patients may potentially benefit from such drugs we screened tumour tissue microarrays. P-Akt was highly activated in 58% (225/390) of cases, whereas it was only similarly expressed in 35% (9/26) of normal breast tissues. Furthermore, HER-2 positive tumours expressed high levels of P-Akt (P < 0.01), supporting in vitro signal transduction.
Conclusion
We determined that Celecoxib analogues are potent inhibitors of P-Akt signalling and kill breast cancer cells that overexpress HER-2. We also defined an association between HER-2 and P-Akt in primary breast tissues, suggesting that these inhibitors may benefit patients in need of new treatment options.
Please see related commentary by Crowder and Ellis at:
==== Body
Introduction
Receptor tyrosine kinases (RTKs) are commonly overexpressed in breast cancer, in which they promote tumour growth and metastasis. For example, insulin-like growth factor (IGF)-1 receptor is an RTK that is overexpressed in about 70% of breast cancers [1,2]. It is fundamentally linked to malignant transformation in vitro and in vivo [3]. IGF-1 receptor is also important for breast cancer invasion and metastasis [4]. Human epidermal growth factor receptor (HER)-2 is yet another important RTK that is overexpressed in 25–30% of invasive ductal breast carcinomas and is associated with poor patient prognosis and increased risk for recurrence [5]. Transgenic mouse models show that HER-2 promotes the development of mammary tumours [6]. Armed with this knowledge, it would appear that finding a convergence point between IGF-1 receptor and HER-2 would provide a new way to target treatment.
A common feature of IGF-1 receptor and HER-2 is signalling through the phosphatidylinositol 3-kinase (PI3K)/Akt pathway [7]. These RTKs activate PI3K, which then catalyzes the production of lipid molecules, including phosphatidylinositol-3,4,5-triphosphate [8]. The phosphatidylinositol-3,4,5-triphosphate lipids trigger the attachment of Akt to the plasma membrane, where it subsequently becomes phosphorylated at two key sites, threonine 308 and serine 473, resulting in its full activation. Threonine 308 is phosphorylated by phosphoinositide-dependent kinase (PDK)-1, whereas the mechanism of phosphorylation at serine 473 is a little more controversial. There are several theories to explain serine 473 phosphorylation, including the action of integrin-linked kinase, autophosphorylation, or an as yet unidentified PDK-2 [9]. Once Akt is fully activated it dissociates from the plasma membrane and proceeds to phosphorylate both cytoplasmic and nuclear target proteins, notably glycogen synthase kinase (GSK)-3β [10], p27Kip [11], mammalian target of rapamycin [12] and forkhead transcription factors [13]. The diverse targets of phosphorylated Akt (P-Akt) regulate proliferation, invasion and evasion of apoptosis. Thus, Akt is a major convergence point for RTK signalling in breast cancer, and so inhibiting it could provide a new therapeutic avenue.
Akt has become a favoured second messenger from a therapeutic standpoint because numerous studies point toward it as a central molecule in the development of cancer. Evidence from experimental models suggests that Akt is a key regulator of tumour development and progression. There are three isoforms of Akt (Akt1, Akt2 and Akt3), which exhibit 80% amino acid sequence homology. The overexpression of each of these isoforms leads to malignant transformation (for review, see Hill and Hemmings [14]).
Transgenic mouse models have been instrumental in addressing the role of Akt in mammary tumour development. For example, mammary tumours that develop from HER-2 transgenic mice clearly overexpress P-Akt [15]. One then questions whether P-Akt truly facilitates the development of the tumours. Hutchinson and coworkers [15] addressed this by engineering bitransgenic mice expressing both HER-2 and constitutively activated Akt1 in the mammary gland. When Akt1 was constitutively expressed the mice developed tumours at a much faster rate than in those that only expressed HER-2. Thus, activated Akt1 plays a functional role in promoting the development of HER-2 positive mammary tumours. Importantly, analysis of primary tumour tissues shows that Akt1 is frequently expressed and highly activated in patients [16]. Akt1 kinase activity is significantly increased in approximately 40–50% of tumour samples from patients with breast (19/50 cases), ovary (11/28 cases) and prostate cancer (16/30 cases) relative to normal tissue [16]. In addition to its role in cancer development, Akt also promotes the survival of tumour cells when confronted with chemotherapeutics and radiation. For example, in breast cancer cells expression of constitutively active Akt1 reduced the ability of doxorubicin [17] or ionizing radiation [18] to induce apoptosis. On the other hand, the PI3K inhibitor Ly294002 or a dominant-negative Akt1 sensitized cancer cells to chemotherapy [17]. These data suggest that inhibiting P-Akt signalling in tumours could have important therapeutic applications.
There is intense interest in targeting RTKs and signal transduction intermediates such as Akt for the treatment of cancer [19]. One approach to inhibiting P-Akt is to target upstream activators of this pathway. For example, patients with tumours expressing HER-2 can be treated with herceptin, a monoclonal antibody that blocks the activation of the receptor [20] and subsequently inhibits Akt phosphorylation [21]. However, fewer than 30% of patients treated with herceptin initially respond [22]. Within that population of patients who initially respond to herceptin, some subsequently develop resistance [22]. A second approach is to inhibit Akt directly. However, no Akt inhibitors are available to patients, although the proof of principle was elegantly provided in a study using an Akt dominant negative (Akt-DN) inhibitor [23]. Targeted disruption of Akt by Akt-DN inhibited the growth of the breast cancer cell lines ZR75-1 and MDA-MB-453 in vitro. The Akt-DN also caused the cells to undergo apoptosis. Adenoviral mediated delivery of Akt-DN also had an impressive antitumour effect in vivo. That study provided the first evidence that targeted disruption of Akt induces apoptosis and suppresses tumour formation in mice. Thus, there is growing interest in the discovery of Akt inhibitors. Potential future drug candidates include phosphatidylinositol analogues that bind specifically to the PH domain of Akt and have been shown to inhibit its phosphorylation in cancer cells [24]. A novel Akt inhibitor was also recently identified from the National Cancer Institute Diversity Set, and preclinical evidence [25] is promising. The inhibitor termed API-2 is a tricyclic nucleoside that selectively kills cancer cells that express high levels of activated Akt [25].
One of the newest classes of Akt inhibitors to be developed are those derived from the common anti-inflammatory drug celecoxib. Initially, it was thought that celecoxib would be a good inhibitor of Akt [26]. However, it was realized that the celecoxib concentrations achievable in patients were of the order of 3 μmol/l [27], whereas 50 μmol/l or greater is required to inhibit Akt activation in prostate [28] and breast cancer cells (Kucab and coworkers, unpublished data). Nonetheless, the observation that celecoxib inhibited P-Akt set the course for the development of analogues that optimally disrupt this pathway at lower concentrations. New celecoxib analogues have since been developed that are superior inhibitors of Akt phosphorylation [29]. The compounds referred to as OSU03012 and OSU03013 inhibited PDK-1 kinase activity in vitro (50% inhibitory concentration 2–5 μmol/l) and prevented Akt phosphorylation in prostate cancer cells at 1–10 μmol/l. Upon longer exposure, these inhibitors induce apoptosis in PC-3 cells. As a part of the Rapid Access to Preventive Intervention Development (RAPID) programme at the US National Cancer Institute [30], a panel of 60 cancer cells lines were screened for response to OSU03012 and OSU03013. It was determined that the compounds were potent inhibitors of tumour cell growth, with an average 50% inhibitory concentration of about 1–2 μmol/l [29]. OSU03012 has also been given orally at a dose of 200 mg/kg for 1 month without overt signs of toxicity as part of the RAPID program (Chen and coworkers, unpublished data). Characterization of the celecoxib analogues thus far indicates that they could be very useful for safely treating many types of cancer.
We therefore further explored the promise of these celecoxib analogues for the treatment of breast cancer. An extensive study of these analogues has not previously been performed in models of breast cancer. We were also curious as to whether serum proteins attenuated the cytotoxic effect of the new analogues as they do with celecoxib [31]. The compounds were evaluated in a HER-2 overexpressing breast cancer cell line, namely MDA-MB-453, which is well characterized for having very high levels of P-Akt. We report herein that both of the celecoxib analogues inhibited Akt phosphorylation and Akt kinase activity. The compounds also inhibited phosphorylation of substrates downstream of Akt (GSK and 4EBP-1). Furthermore, OSU-03012 and OSU-03013 initiated the apoptotic pathway, resulting in under 90% cell viability within 24 hours. We then addressed how often Akt is activated in primary tumours to estimate the number of patients that might benefit from these small molecule inhibitors. We determined that P-Akt was moderately to highly expressed in 58% of primary tumours, suggesting that these inhibitors could potentially be used to treat a substantial number of patients. Furthermore, we found that Akt was commonly activated in tumours that overexpress HER-2. These data therefore provide evidence to support further preclinical development of celecoxib analogues for the treatment of breast cancer, particularly in cases in which HER-2 is overexpressed.
Materials and methods
Effect of celecoxib analogues on Akt signalling and apoptosis
The breast cancer cell lines (MDA-MB-453, MCF-7, T47D, MDA-MB-231 and HBL100) were obtained from the American Tissue Culture Collection (Manassas, VA, USA). The 184htrt cells were a gift from Dr J Carl Barrett (National Cancer Institute, Bethesda, MA, USA). All of the experiments were performed in the presence of 5% foetal calf serum, RPMI-1640, with the exception of the 184htrt cells, which were grown as previously described [32]. The celecoxib analogues were synthesized as previously described by us [29]. Ly294002 was purchased from Sigma (St. Louis, MO, USA) and celecoxib was obtained from Pharmacia (St Louise, MO, USA). All compounds were dissolved in dimethyl sulphoxide.
To study the effect of the inhibitors on signal transduction, the cells were treated with either the PI3K/Akt inhibitor Ly294002 (30 μmol/l), or OSU03012 or OSU03013 inhibitor, each at 5 and 10 μmol/l for 2 hours or at later times points, as indicated. Comparisons were also made to the parent compound, celecoxib, at concentrations of 50 μmol/l and 75 μmol/l. Whole cell extracts were prepared in accordance with the protocol for P-Akt detection by Cell Signaling Technologies (CST Beverly, MA, USA). All antibodies were purchases from CST unless otherwise indicated. P-AKTser473, P-Aktthr308, total AKT, P-4EBP-1, P-S6, P-Erk, total Erk, P-MK2, MK2, P-GSK (Upstate Biotechnology Inc, Lake Placid, NY, USA) and actin (Santa Cruz Biotechnology, CA, USA) were detected by analyzing 50 μg total protein separated on a 12% acrylamide gel.
Akt kinase activity was assessed using a modified assay. Nonradioactive Akt kinase assays were performed using a protocol modified from that of Cell Signaling Technologies. In brief, 500 μg protein, from cells treated as described above, were immunoprecipitated with the 5G3 pan-Akt antibody (CST) overnight at 4°C. The following day the Akt–antibody complexes were incubated with protein G coated agarose beads. The immunoprecipitated complexes were washed and then incubated for 30 minutes at 30°C in kinase buffer with 1 μg of recombinant GSK-3 protein (CST) and 200 μmol/l ATP. To stop the reaction, 15 μl of 4× SDS sample buffer with β-mercaptoethanol was added. The assays were boiled for 5 min and then one-third of each reaction was separated on a 12% acrylamide gel and immunoblotted. The blots were analyzed using antibodies to P-GSK-3 protein, as well as total recombinant GSK-3 and total Akt.
In order to assess the effect of the compounds on apoptosis, the MDA-MB-453 cells were treated with Ly294002, celecoxib, or the analogues at the indicated concentrations for 12 or 24 hours and poly(ADP-ribose) polymerase (CST) cleavage was examined in the treated cells by immunoblotting. Cell extracts were also subjected to the Cell Death Detection Assay according to the manufacturer's instructions (Roche Diagnostics, Laval, QC, Canada). The viability of MDA-MB-453 was determined 24 hours after exposure to celecoxib, the analogues, or Ly294002 using the Celltiter 96 Aqueous cell proliferation/survival assay (Promega, Madison WI, USA), as previously described [33]. The influence of serum proteins was also examined. The MDA-MB-453 cells were treated with the test compounds as indicated in either 5% foetal bovine serum/RPMI or 0.1% foetal bovine serum/RPMI, and cytotoxicity was evaluated 24 hours later using the MTT assay.
Examination of phosphorylated Akt in primary breast tissues
For construction of the tumour tissue microarray (TMA), 481 primary breast cancer samples were obtained from archival cases at Vancouver General Hospital dating between 1974 and 1995. Patient information and tumour pathology are summarized in Additional file 1. Anonymous coding was used to protect patient rights and the samples were procured in accordance with the guidelines established by the Vancouver General Hospital. Tumour samples were taken before initiation of cancer treatment, and were formalin fixed and embedded in paraffin. The TMA was constructed as previously described by us [1].
For detection of P-Akt, the tissues underwent antigen retrieval by incubating the slides for 30 min in 10 mmol/l citrate buffer (pH 6.0) at 60–90°C in a vegetable steamer. Endogenous peroxidases were quenched by incubating the sections for 10 min in 3% H2O2. Additionally, nonspecific interactions were blocked for 30 min using a non-serum-blocking reagent (DAKO, Denmark), followed by 20 min with an avidin/biotin blocking solution (DAKO, Carpenteria, Ca, USA). The primary antibody (Phospho-AKTS473 IHC Specific; CST) was diluted 1:250 with a 1% bovine serum albumin solution, applied to the sides and incubated overnight at 4°C. For signal amplification, we then used the LSAB+ System (DAKO), which involved incubation with a biotinylated secondary antibody followed by streptavidin treatment. P-Akt was visualized by addition of NovaRed substrate (Vector Laboratories, Burlingame, CA, USA) and the sections were counter-stained with haematoxylin. Additionally, a negative control reaction with no primary antibody was performed for each slide in parallel.
The scoring system for P-Akt expression was as follows: 0 = negative, 1 = weak, 2 = moderate and 3 = high staining intensity. Of the 481 cases on the array, 438 contained invasive carcinoma and the characteristics of these patients are described in Table 1. A total of 390 invasive carcinoma cases were interpretable for P-Akt expression. In most cases, P-Akt expression was detected primarily in the cytoplasm, although nuclear staining was occasionally observed. The majority of P-Akt was predominantly expressed in the tumour epithelial cells, although there was also notable staining of the endothelial cells. P-Akt was not expressed in the stroma. To assess P-Akt staining in normal breast tissue, 26 samples were obtained from patients who underwent reduction mammoplasties at Vancouver General Hospital from 2000 to 2001. The tissues were formalin fixed and paraffin embedded. One section of each sample was stained with haematoxylin and eosin and then assessed by a pathologist to ensure presence of normal breast epithelium. P-Akt immunohistochemistry was performed on the normal tissues as described above for the TMAs. Raw scores for P-Akt expression were entered into a standardized electronic spreadsheet (Excel for Windows, Microsoft, Redmond, WA, USA), processed using Deconvoluter software designed for management of TMA data, and then analyzed using the SPSS for Windows statistical software package (SPSS version 11; SPSS, Chicago, IL). The difference in P-Akt expression between normal and tumour breast tissue was calculated using χ2 analysis. HER-2 R was stained with an antibody designated A485 (Dako) at a dilution of 1:500 and detected using the LSAB+ System.
Results
Celecoxib analogues disrupt Akt signalling in breast cancer cells and induce apoptosis
We screened a panel of breast cell lines for P-Akt to find the one with the highest levels. Our panel included the preneoplastic cell line 184htrt and the cancer cell lines T47D, MCF-7, MDA-MB-231 and MDA-MB-453. The MDA-MB-453 cells expressed the highest level of P-Akt, and therefore they were extensively used to characterize the effect of the celecoxib analogues on the Akt pathway (Fig. 1a). This cell line also expressed the highest level of HER-2 (Fig. 1b).
To examine the potential effect of the inhibitors, the MDA-MB-453 cells were treated for 2 hours with DMSO, Ly294002 (30 μmol/l), OSU-03012 (5 or 10 μmol/l) or OSU-03013 (5 or 10 μmol/l) and compared with celecoxib treated cells (50 or 75 μmol/l). The celecoxib analogues suppressed phosphorylation of Akt at threonine 308 and serine 473, whereas celecoxib did not (Fig. 2a). These data indicated that Akt kinase activity should be inhibited by the celecoxib analogues. In support of this, we determined that the analogues inhibited phosphorylation of the Akt substrates GSK-3β and 4E-BP1 (Fig. 2a). Similarly, the celecoxib analogues inhibited P-Akt in the T47D cells (Fig. 2b). Kinase assays were then performed on the MDA-MB-453 cells to provide direct evidence that Akt activity was lost. Each of the celecoxib analogues markedly suppressed Akt kinase activity, similar to that with the PI3K/Akt inhibitor Ly294002 (Fig. 2c).
The off-target effects of the inhibitors were then examined by focusing on the mitogen-activated protein kinase (MAPK) and p38 pathways. OSU03012 did not inhibit signal transduction through the MAPK pathway, based on a lack of P-Erk inhibition (Fig. 2d). Similar results were found for Ly294002. In contrast, OSU03013 inhibited P-Erk at 4 hours (Fig. 2d, lane 8) and more so at 6 hours (Fig. 2d, lane 12). This was an important finding because it indicated that OSU03012 specifically inhibited the Akt pathway. OSU03013 on the other hand inhibited both the Akt and MAPK pathways at the later time points. To complete this portion of the study, the potential effect of the compounds was examined in the context of the p38 pathway by monitoring P-MAPK activated protein kinase-2. The latter was not inhibited by OSU03012, OSU03013, or Ly294002 (Fig. 2d). Thus, the celecoxib analogues are potent inhibitors of the Akt pathway in HER-2 over-expressing breast cancer cells; however, OSU03012 was more specific than OSU03013.
Akt plays a central role in preventing apoptosis, and therefore we investigated the potential for the celecoxib analogues to trigger programmed cell death. Subsequent to the inhibition of Akt activity, we observed a significant increase in poly(ADP-ribose) polymerase cleavage 12 and 24 hours later, indicating that the cells were undergoing apoptosis (Fig. 3a). This was consistent with nucleosomal fragmentation (Fig. 3b). We therefore concluded that the cells were undergoing apoptosis following drug treatment. The fate of the cells was followed to 24 hours, at which time cell viability was assessed. OSU03012 and OSU03013 killed the MDA-MB-453 cells in a dose dependent manner (Fig. 3c). There was a more than 90% reduction in viability with 10 μmol/l of each of the compounds. In contrast, the parent compound celecoxib had a similar effect on viability at 100 μmol/l. These studies were then extended to include the T47D breast cancer cells, and they responded in a similar manner (Fig. 3d).
It was noted that although Ly294002 inhibits P-Akt, its effect on cell viability was not as robust as that of OSU03012. To elucidate why this may be, we treated the cells with Ly294002 over periods of 2, 4, 6 and 24 hours to evaluate the possibility that it may not be stable in an aqueous solution and therefore loses its ability to inhibit P-Akt. We noted that after 6 hours the ability of Ly294002 to inhibit P-Akt began to diminish. Furthermore, by 24 hours Ly294002 only inhibited P-Akt by about 50% (Fig. 3e, top panel). In contrast, P-Akt inhibition was sustained for OSU03012 throughout the time course (Fig. 3g, bottom panel). It was not possible to measure the effect of OSU03012 after 24 hours because of the cytotoxic effect on the cells. Thus, we determined that Ly294002 was unable to sustain its inhibitory effect on P-Akt, which could explain why only 25–30% of the cells died after 24 hours.
The efficacy of anticancer drugs can be perturbed by problems with serum binding, and this often results in attenuated cellular effects. Therefore, we compared the cytotoxic effect of OSU03012 and OSU03013 in high (5%) verses low (0.1%) serum. This was done to ascertain whether serum had a protective effect against the celecoxib analogues. Up to this point, all of the experiments were performed in 5% foetal bovine serum/RPMI 1640 containing media, and therefore comparisons were made with cells treated in 0.1% fetal bovine serum/RPMI 1640. Serum had a remarkable protective effect against celecoxib (Fig. 4). In contrast, the serum had little effect on how well the celecoxib analogues killed the cells. These data suggest that altering the chemical structure of celecoxib not only enhanced cell killing but also increased the bioavailability of the drugs in the presence of serum.
Frequency of phospharylated Akt expression in normal and tumor breast tissue
In order to estimate the proportion of patients that might benefit from inhibitors to P-Akt, we screened breast tumour TMAs. A description of the patients and the clinicopathological features of their tumours are given in Additional file 1. P-Akt expression was moderately to highly expressed in 58% (221/390) of the tumours. The distribution of P-Akt expression in the 390 tumours was as follows: no staining, 43/390 (11%); weak staining, 122/390 (31%); moderate staining, 120/390 (31%); and strong staining, 105/390 (27%; Fig. 5a–d, respectively). P-Akt was predominantly expressed in epithelial cells and was noted in endothelial cells, but it was not expressed in the stroma. P-Akt was highly expressed in both oestrogen receptor positive and negative cases. There was no significant difference (P = 0.5839) in overall survival between patients who expressed high levels of P-Akt and those who expressed low levels of the activated protein. We also evaluated the relationship of P-Akt expression with other clinicopathologic variables, such as grade, lymph node status and histology, but we found no significant correlations (data not shown). This is most likely because we were unable to define the patient population based on treatment.
Comparisons were then made between normal and neoplastic tissues because we noted that in some instances the adjacent normal ducts expressed less P-Akt (Fig. 5e, broken arrow) compared with the tumour (Fig. 5e, solid arrow). Because the cores only represent a small amount of the tumour tissue, normal ducts were often not present. We obtained 26 normal breast tissue samples from reduction mammoplasties to examine P-Akt. In some instances, P-Akt was not expressed (Fig. 5f) whereas in others it was detectable (Fig. 5g). Overall, we determined that P-Akt was moderately to highly expressed in only 35% (9/26) of cases. We then determined that P-Akt more likely to be activated in tumours than in normal breast tissue by χ2 analysis (P ≤ 0.025). Thus, inhibitors to this pathway would hopefully affect the tumour tissue and cause few side effects in surrounding normal tissue. We also noted expression of activated Akt in endothelial cells surrounding tumours (Fig. 5h). Thus, OSU03012 and OSU03013 could potentially inhibit P-Akt in tumours as well as in the surrounding endothelial cells.
Finally, in attempt to elucidate why Akt may be activated in the tumours, we stained the tissues for HER-2 expression. Tumors that expressed high levels of HER-2 were much more likely to express activated Akt (P < 0.01) than were those that expressed low levels of the receptor (Table 1). Therefore, in this cohort of patients HER-2 overexpression was positively related to activated Akt, further supporting in vitro models of signal transduction. Together, these data show that P-Akt is frequently activated in primary breast cancers, indicating that small molecule inhibitors targeting this pathway may be useful for treating this disease.
Discussion
In this study, we determined that OSU03012 and OSU03013 are potent inhibitors of the Akt signalling pathway, which are effective at inducing apoptosis in a breast cancer cell line that expresses high levels of HER-2. Alternatively, celecoxib was not particularly effective at inhibiting Akt or killing the MDA-MB-453 cells. Thus, the new celecoxib analogues are much better than the parent compound at inhibiting the Akt pathway. Between the two inhibitors, it appeared that OSU03012 was more specific for inhibiting the Akt pathway than was OSU03013. We also noted that the celecoxib analogues were just as effective in high serum as they were in low serum. This is in contrast to the protective effect that serum had on celecoxib. Our data is consistent with the protection afforded by serum when pancreatic cells were treated with celecoxib (50 μmol/l) [31].
Thus, it appears that structural modification of celecoxib resulted not only in increased P-Akt inhibition but also in enhanced bioavailability, at least in vitro. This is important to us because we move toward the development of Akt inhibitors that could be taken orally. For example, we previously reported that the oral administration of the celecoxib derivative DMC (4‑ [5-(2,5-dimethylphenyl)-3 trifluoromethyl-1H-pyrazol-1-yl]-benzene-sulfonamide) resulted in inhibition of P-Akt and ultimately suppressed development of prostate tumours [34]. This compound is structurally related to OSU03012 and OSU03013, and therefore we expect them also to be amenable to oral administration. In a recent study, OSU03012 (100 mg/kg per day) was given orally to mice bearing prostate tumours, and the intratumoural concentrations of the drug were in excess of 15 μmol/l, which coincided with tumour regression (Kulp S. personal communication). The mice tolerated the drug very well without weight loss. These data are important because our work indicates that we only need 5–10 μmol/l OSU03012 to kill highly aggressive breast cancer cells in vitro. Thus, if we are able to establish an intratumoural concentration in excess of 15 μmol/l, then it is very likely that this compound will have a cytotoxic effect against the MDA-MB-453 cells when studied in a xenograft model. The lack of overt toxicity is also striking and will be confirmed in models of breast cancer. Given these encouraging data, preclinical studies to this effect are currently underway in our laboratory in models of breast cancers.
We also determined that P-Akt was expressed in 58% (221/390 cases) of breast cancers. This represents one of the largest studies of P-Akt expression in breast cancer to date. The largest study reported that 49% (331/670) of breast cancer cases expressed high P-Akt [35]. Similarly, a smaller study of 40 patients reported that P-Akt was highly activated in 48% of breast cancer cases [36]. The prognostic value of P-Akt appears to depend on the types of tumours analyzed and the treatment protocol that the patients received. Like Panigrahi and coworkers [35], we did not find an association between P-Akt and patient survival. In both cases, P-Akt was examined in a large cohort of patients (about 670 cases) for whom treatment was not standardized, which could explain the lack of correlation with survival. In contrast, P-Akt was reported to be associated with poor overall survival in a subset of lymph node negative breast cancer patients (n = 99) for whom treatment was standardized [37]. Likewise, P-Akt predicted poor outcome among endocrine treated breast cancer patients (n = 93) who participated in a clinical trial using tamoxifen, goserelin, or both agents [38]. In another study [39] P-Akt was not associated with poor survival in a group of patients that was part of a controlled clinical trial examining the potential benefit of chemotherapy. The investigators found that the expression of P-Akt did not differentiate between chemotherapy responders and nonresponders. However, P-Akt was associated with a lack of response to radiotherapy. They concluded that patients were more likely to benefit from radiotherapy if their tumours were P-Akt negative.
Regarding P-Akt in normal tissues, we are the first to examine the frequency of P-Akt in normal breast tissue, in which Akt was activated in only 35% of cases, as compared with 58% of tumours. Similar to our study, basal P-Akt is low in normal ovarian surface epithelial cells compared with tumour cell lines [40]. Normal fibroblasts and colonic epithelial cells also express relatively little P-Akt compared with tumour cell lines [23]. It remains unclear what regulates P-Akt expression in normal ducts. We suspect that it relates to hormonal and/or growth factor activation of the Pi3K/Akt pathway. In particular, oestrogen and IGF-1 have been shown to stimulate the phosphorylation of Akt and induce its kinase activity in breast cancer cells [41]. Because oestrogen and IGF-1 are also present in the sera of women, it is possible that these mitogens could induce P-Akt in normal breast epithelial cells. Independent of this, we found that breast tumours consistently had higher levels of P-Akt. We estimate that breast tumours are twice as likely to express high levels of P-Akt, providing further rationale for developing inhibitors of this pathway for the treatment of cancer. Importantly, tumour cells actually depend on activated Akt for survival whereas normal cells do not [23]. This was determined using an adenoviral system that produced an Akt dominant negative inhibitor.
It could be anticipated that the celecoxib analogues might also be used in instances where resistance to standard chemotherapy has developed [42]. There are cases in which HER-2 overexpressing cells are refractory to herceptin treatment. This was illustrated in a clinical trial in which patients were recruited based on HER-2 overexpression. It was somewhat surprising that under 30% of patients responded to herceptin even though they qualified for the study based on HER-2 overexpression. Patients with amplified HER-2 had a 34% response rate whereas only 7% responded without amplification [22]. This study also pointed out that only patients with tumours that stained 3+ (n = 84 patients) responded to the drug whereas those staining 2+ (n = 27 patients) did not. The patients who did not respond were described as having tumours with moderate overexpression (2+) and/or that expressed HER-2 in the absence of gene amplification. It is noteworthy that the MDA-MB-453 cell line is a model for such tumours. They are considered to be moderate HER-2 overexpressing cells (2+) when compared with SkB-3 (3+) or BT-474 cells (3+), based on western blotting [43]. Although they overexpress HER-2, they are resistant to herceptin [43,44]. At this point, the underlying mechanism for recalcitrance to herceptin is not understood. One possible explanation for resistance is the high levels of P-Akt [43]. To examine this in more detail, constitutively activated Akt was expressed in BT-474 cells, which are known to be sensitive to Herceptin. However, expression of high P-Akt rendered the BT-474 cells insensitive to herceptin [43]. This is consistent with a report showing that that inhibiting P-Akt with Ly294002 enhances the cytotoxic effect of herceptin [17]. Inhibiting P-Akt with Ly294002 also prevents the anchorage independent growth of breast cancer cell lines that overexpress HER-2, such as the MDA-MB-453 cells [45]. A dominant negative inhibitor of the p85 subunit of Pi3K similarly blocked anchorage independent growth, providing further evidence that selective inhibition of this pathway may be particularly useful when treating HER-2 overexpressing breast cancers [45]. With respect to this, it seems reasonable that inhibiting the Akt pathway may be a way to kill breast cancer cells that have developed herceptin resistance.
Our study indicates that OSU03012 and OSU03013 inhibit P-Akt and ultimately kill herceptin resistant cells in vitro. This is timely, given a recent report [46] showing that the parent compound celecoxib did not benefit patients with HER-2 overexpressing tumours that were also resistant to herceptin. Thus, now more than ever there is a need to identify new agents that can be used to treat patients who have limited therapeutic options.
Conclusion
In conclusion, celecoxib analogues provide an opportunity to inhibit P-Akt and ultimately kill breast cancer cells that overexpress HER-2.
Abbreviations
Akt-DN = Akt dominant negative; 4EBP-1 = 4E binding protein-1; GSK = glycogen synthase kinase; HER = human epidermal growth factor receptor; IGF-1 = insulin-like growth factor-1; MAPK = mitogen-activated protein kinase; P-Akt = phosphorylated Akt; PDK = phosphoinositide-dependent kinase; PI3K = phosphatidylinositol 3-kinase; RTK = receptor tyrosine kinase; TMA = tissue microarray.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
JEK performed many of the western blotting experiments. She also developed the P-Akt immunostaining and assisted in writing. CL carried out some of the drug treatments and performed western blots. CC and JZ made OSU03012. CBG was the lead pathologist on the study. MC provided the biostatistical support. DH was in charge of building the tumor tissue microarray. EY performed the Her-2 immunostaining and helped optimize the P-Akt staining. JE provided the normal breast tissues, MP contributed infrastructure support. SD was the Principle investigator that designed the project, oversaw the daily activities and wrote the manuscript.
Supplementary Material
Additional File 1
A table summarizing the characteristics of the patients and their tumours.
Click here for file
Acknowledgements
JEK was funded by a BC Research Institute for Children's and Women's Health studentship. This work was further supported by funds through the National Cancer Institute of Canada Streams of Excellence, the Translational Acceleration Program-1 and the Canadian Institute for Health Research. Partial funding was also provided by the Pediatric Oncology Basic and Translational Research at the BC Research Institute for Children's and Women's Health. Support also came from the British Columbia Cancer Agency and the Prostate Cancer Program at the Jack Bell Laboratories. We thank the following surgeons for providing normal breast specimens: Dr Clugston, Dr Lennox, Dr Sproul and Dr Warren. The Prostate Cancer Center and the BC Cancer Agency are also gratefully acknowledged for their support of this research.
Figures and Tables
Figure 1 Examination of P-Akt and HER-2 levels in a panel of breast cancer cell lines. (a) Proteins were isolated from cells growing in log phase and the levels of phosphorylated Akt (P-Akt) were then assessed using antibodies to serine 473 and threonine 308. MDA-MB-453 cells expressed the highest levels of activated Akt relative to the other lines. There were no differences in the levels of total Akt with the exception of the preneoplastic cell line 184htrt. Actin was detected as a loading control. (b) Human epidermal growth factor receptor (HER)-2 protein expression was evaluated in a panel of breast cell lines. The MDA-MB-453 and T47D cells expressed HER-2 whereas the other cell lines did not.
Figure 2 Impact of signal transduction inhibitors on Akt signalling. (a) The MDA-MB-453 cells were treated for 2 hours with Ly294002 (30 μmol/l), OSU03012, OSU03013 (5 or 10 μmol/l), or celecoxib (50 or 75 μmol/l). The Celecoxib analogues inhibited Akt phosphorylation at both threonine 308 and serine 473. Phosphorylation of the Akt substrates glycogen synthase kinase (GSK) and 4E binding protein (4EBP)-1 was subsequently attenuated. In contrast, celecoxib did not have an inhibitory effect. Ly294002 inhibited signal transduction through Akt, as expected. Total Akt and actin were unaffected by exposure to the signalling inhibitors. (b) T47D cells were also exposed to the drugs as described above and evaluated for phosphorylated Akt (P-Akt) using antibodies to Ser473 and Thr308. Total Akt was included as a control for loading. (c) Cells were treated as above and Akt kinase was measured against the substrate GSK. Each of the celecoxib analogues inhibited Akt kinase activity. The degree of inhibition was similar to that with Ly294002. There was minimal nonspecific kinase activity in the absence of Akt based on the IgG control. Total Akt was evaluated as a loading control to confirm that the loss of activity was not due to differences in experimental conditions. (d) The MDA-MB-453 cells were treated for 2, 4, or 6 hours with dimethyl sulphoxide (DMSO), Ly294002 (30 μmol/l), OSU03012 (10 μmol/l), or OSU03013 (10 μmol/l) and probed for P-Erk1/2, total Erk, P-MK2, total MK2 and actin.
Figure 3 Effect of the Celecoxib analogues on apoptosis induction. To follow the fate of the cells upon Akt inhibition, indicators of apoptosis were temporally measured. Poly(ADP-ribose) polymerase (PARP) cleavage and nucleosomal fragmentation were measured after 12 and 24 hours. The cells were treated with Ly294002 (30 μmol/l), Celecoxib (50 or 100 μmol/l), OSU03012 (5, 7.5, or 10 μmol/l), or OSU03013 (5, 7.5, or 10 μmol/l) and then the cell pellets were split for PARP and nucleosomal cleavage. (a) There was a dose dependent increase in PARP cleavage on treatment with OSU03012 and OSU03013 at both 12 and 24 hours. LY294002 similarly induced PARP cleavage but to a lesser extent. Celecoxib at 50 μmol/l did not have sustained effects on PARP cleavage, whereas the high dose of celecoxib (100 μmol/l) did. (b) The induction of apoptosis was secondarily analyzed by nucleosomal fragmentation. There was a dose dependent increase in nucleosomal fragmentation upon treatment with increasing concentration of either OSU03012 or OSU03013. Likewise, Ly294002 induced fragmentation of the nucleosomes at both time points. Celecoxib at 50 μmol/l did not have such an effect. In contrast, high dose celecoxib did stimulate apoptosis. Each treatment was conducted in replicates of six on two difference occasions. (c) Impact of signal transduction inhibitors on cell survival. The MDA-MB-453 cells were exposed to Ly294002 (30 μmol/l), celecoxib, OSU03012, and OSU03013 at the indicated concentrations and cell viability was assessed 24 hours later. OSU03012 and OSU03013 killed more than 90% of the cells with 10 μmol/l of the respective inhibitors. Celecoxib at 100 μmol/l had a similar effect, but celecoxib at 50 μmol/l was not effective at reducing cell viability. This screen was conducted in replicates of four in three separate experiments. (d) The T47D cells similarly responded to the celecoxib analogues based on cell survival. Each experiment was performed in replicates of six on three separate occasions. (e) MDA-MB-453 cells were treated for 2, 4, 6, or 24 hours with Ly294002 and P-Akt was measured (top panel). A comparison was made with the MDA-MB-453 cells exposed to OSU03012 (10 μmol/l) for 2, 4, or 6 hours. Total Akt was measured as a control for sample input.
Figure 4 Effect of serum on the cytotoxicity of celecoxib and its analogues. MDA-MB-453 cells were treated with dimethyl sulphoxide (DMSO), Ly294002, celecoxib, OSU03012, or OSU03013 in either 5% foetal bovine serum/RPMI 1640 or 0.1% foetal bovine serum/RPMI 1640. Cell survival was measured 24 hours later using the MTT assay. Each of the treatments were tested in replicates of four and repeated twice.
Figure 5 Phosphorylated Akt expression in tumour and normal tissues of the breast. (a-d) Phosphorylated Akt (P-Akt) staining of invasive ductal carcinomas (IDCs) of the breast ranging from undetectable to intense staining. The scoring system was from 0 to 3 and examples of such are represented in panels a-d, respectively. (e) Normal ducts (dashed arrow) adjacent to the IDC (solid arrow) expressed appreciably less P-Akt. (f) Likewise, P-Akt staining was weak when whole sections of normal breast tissue were stained. (g) However, some cases of normal breast tissues stained more intensely for P-Akt. (h) It was also noted that P-Akt was present in endothelial cells surrounding the tumours. Original magnifications: panels a-g, 200×; panel h, 400×.
Table 1 Correlation between P-Akt and HER-2 proteins in primary breast tumours
Negative (P-Akt 0,1) Positive (P-Akt 2,3) Total
Negative (HER2 0,1; counted/expected) 111/100.2 130/140.8 241
Positive (HER2 2,3; counted/expected) 32/42.8 71/60.2 103
Total 143 201 344
A total of 344 invasive ductal carcinomas were examined for expression of phospharylated Akt (P-Akt) and human epidermal growth factor receptor (HER)-2. A binary scoring system (0,1 versus 2,3) was used to compare staining between samples. Patients who overexpressed HER-2 were more likely to express P-Akt also. Statistical analysis was derived through the χ2 test (P < 0.01).
==== Refs
Nielsen T Andrews H Cheang M Kucab JE Hsu FD Ragaz J Gilks CB Makretsov N Bajdik CD Brookes C Expression of the insulin-like growth factor-1 receptor and urokinase plasminogen activator in breast cancer is associated with poor survival: Potential for intervention with 17 allylamino 17-demethoxy geldanamycin Cancer Res 2004 64 286 291 14729636
Sachdev D Hartwell JS Lee AV Zhang X Yee D A dominant negative type I insulin-like growth factor receptor inhibits metastasis of human cancer cells J Biol Chem 2004 279 5017 5024 14615489 10.1074/jbc.M305403200
Kaleko M Rutter WG Miller DA Overexpression of the human insulin-like growth factor-1 receptor promotes ligand dependent neoplastic transformation Mol Cell Biol 1990 10 464 473 2153917
Dunn SE Ehrlich M Sharp NJH Reiss K Solomon G Hawkins R Baserga R Barrett JC A dominant negative mutant of the insulin-like growth factor-1 receptor inhibits the adhesion, invasion, and metastasis of breast cancer Cancer Res 1998 58 3353 3361 9699666
Slamon DJ Clark GM Wong SG Levin WJ Ullrich A McGuire WL Human breast cancer: correlation of relapse and survival with amplification of HER-2/neu oncogene Science 1987 235 177 182 3798106
Muller WJ Sinn E Pattengale PK Wallace R Leder P Single-step induction of mammary adenocarcinoma in transgenic mice bearing the activated c-neu oncogene Cell 1988 54 105 115 2898299 10.1016/0092-8674(88)90184-5
Craven RJ Lightfoot H Cance WG A decade of tyrosine kinases: from gene discovery to therapeutics Surg Oncol 2003 12 39 49 12689669 10.1016/S0960-7404(03)00004-5
Cantley LC The phosphoinositide 3-kinase pathway Science 2002 296 1655 1657 12040186 10.1126/science.296.5573.1655
Scheid MP Marignani PA Woodgett JR Multiple phosphoinosidide 3-kinase dependent steps in activation of protein kinase B Mol Cell Biol 2002 22 6247 6260 12167717 10.1128/MCB.22.17.6247-6260.2002
Cross BAE Alessi DR Cohen P Andjelkovich M Hemming BA Inhibition of glycogen synthase kinase-3 by insulin mediated by protein kinase B Nature 1995 378 785 789 8524413 10.1038/378785a0
Shin I Yakes FM Rojo F Shin NY Bakin A Baselga J Arteaga CL PKB/AKT mediates cell-cycle progression by phosphorylation of p27(Kip1) at threonine 157 and modulation of its cellular localization Nat Med 2002 8 1145 1152 12244301 10.1038/nm759
Scott PH Brunn GJ Kohn AD Roth RA Lawrence JC Evidence of insulin-stimulated phosphorylation and activation of the mammalian target of rapamycin mediated by protein kinase B signaling pathway Proc Nat Acad Sci USA 1998 95 7772 7777 9636226 10.1073/pnas.95.13.7772
Brunet A Bonni A Zigmond MJ Lin MZ Juo P Hu LS Anderson MJ Arden KC Blenis J Greenberg ME AKT promotes cell survival by phosphorylating and inhibiting a forkhead transcription factor Cell 1999 96 857 868 10102273 10.1016/S0092-8674(00)80595-4
Hill MM Hemmings BA Inhibition of protein kinase B/Akt: implications for cancer therapy Pharmacol Ther 2002 93 243 251 12191616 10.1016/S0163-7258(02)00193-6
Hutchinson JN Jin J Cardiff RD Woodgett JR Muller WJ Activation of Akt-1 (PKB-alpha) can accelerate ErbB-2-mediated tumorigenesis but suppresses tumor invasion Cancer Res 2004 64 3171 3178 15126356
Sun M Wang G Paciga JE Feldman RI Yuan ZQ Ma XL Shelley SA Jove R Tsichlis PN Nicosia SV AKT1/PKBalpha kinase is frequently elevated in human cancers and its constitutive activation is required for oncogenic transformation in NIH3T3 cells Am J Pathol 2001 159 431 437 11485901
Clark AS West K Streicher S Dennis PA Constitutive and inducible Akt activity promotes resistance to chemotherapy, Trastuzumab, or Tamoxifen in breast cancer cells Mol Cancer Ther 2002 1 707 717 12479367
Liang K Jin W Knuefermann C Schmidt M Mills GB Lang KK Milas L Fan Z Targeting the phosphatidylinositol 3-kinase/Akt pathway for enhancing breast cancer cells to radiotherapy Mol Cancer Ther 2003 2 353 360 12700279
Luo J Manning BD Cantley LC Targeting the PI3K/Akt pathway in human cancer: Rationale and promise Cancer Cell 2003 4 257 262 14585353 10.1016/S1535-6108(03)00248-4
Slamon DJ Leyland-Jones B Shak S Fuchs H Paton V Bajamonde A Fleming T Eiermann W Wolter J Pegram M Use of chemotherapy plus a monoclonal antibody against Her2 for metastatic breast cancer that overexpresses Her2 N Engl J Med 2001 344 783 792 11248153 10.1056/NEJM200103153441101
Cuello M Ettenberg SA Clark SS Keane MM Posner RH Nau MM Dennis PA Lipkowitz S Down-regulation of the erbB-2 receptor by trastuzumab (herceptin) enhances tumor necrosis factor-related apoptosis-inducing ligand-mediated apoptosis in breast and ovarian cancer cell lines that express erbB-2 Cancer Res 2001 61 4892 4900 11406568
Vogel CL Cobleigh MA Tripathy D Gutheil JC Harris LN Faehrenbacher L Slamon D Murphy M Novotny WF Murchmore M Efficacy and safety of trastuzumab as a single agent in first-line treatment of Her-2 overepxressing metastatic breast cancer J Clin Oncol 2002 20 719 726 11821453 10.1200/JCO.20.3.719
Jetzt A Howe JA Horn MT Maxwell E Yin Z Johnson D Kumar CC Adenoviral-mediated expression of a kinase-dead mutant of Akt induces apoptosis selectively in tumor cells and suppresses tumor growth in mice Cancer Res 2003 63 6697 6706 14583464
Kozikowski AP Sun H Brognard J Dennis PA Novel PI analogues selectively block activation of the poor-survival serine/threonine kinase Akt J Am Chem Soc 2003 125 1144 1145 12553797 10.1021/ja0285159
Liu Y Dan HC Sun M Liu Q Feldman RI Hamilton AD Polokoff M Nicosia SV Herlyn M Sebti SM Akt/Protein kinase B signaling inhibitor-2, a selective small molecule inhibitor of Akt signaling with antitumor activity in cancer cells overexpressing Akt Cancer Res 2004 64 4394 4399 15231645
Song X Lin HP Johnson AJ Tseng PH Yang YT Kulp SK Chen CS Cyclooxygenase-2, player or spectator in cyclooxygenase-2 inhibitor- induced apoptosis in prostate cancer cells J Natl Cancer Inst 2002 94 585 591 11959891
Brune K Neubert A Pharmacokinetic and pharmacodynamic aspects of the ideal Cox-2 inhibitor: a pharmacologists perspective Clin Exp Rheumatol 2001 19 6 Suppl 25 S51 S57 11695253
Hsu A Ching TT Wang D Song X Rangnekar VM Chen CS The cyclooxygenase-2 inhibitor Celecoxib induces apoptosis by blocking Akt activation in human prostate cancer cells independently of BCL-2 J Biol Chem 2000 275 11397 11403 10753955 10.1074/jbc.275.15.11397
Zhu J Huang J Tseng PH Yang Y Fowble J Shiau C Shaw Y Kulp SK Chen CS From the cyclooxygenase-2 inhibitor Celecoxib to a novel class of 3-phosphoinositide-dependent protein kinase inhibitors Cancer Res 2004 64 4309 4318 15205346
Rapid Access to Preventive Intervention Development (RAPID) program
Levitt RJ Pollak M Insulin-like growth factor-1 antagonizes the antiproliferative effects of cyclooxygenase-2 inhibitors on BXPC-3 pancreatic cancer cells Cancer Res 2002 62 7372 7376 12499282
Oh JS Buchel P Martin K Kucab JE Oshimura T Bennett L Barrett JC DiAugustine RP Afshsari C Dunn SE Insulin-like growth factor-1 inscribes a gene expression profile for angiogenic factors and cancer progression in breast epithelial cells Neoplasia 2002 4 204 217 11988840 10.1038/sj.neo.7900229
Dunn SE Hardman R Kari FW Barrett JC Insulin-like growth factor-I alters drug sensitivity of human breast cancer cells by inhibition of apoptosis induced by diverse anticancer drugs Cancer Res 1997 57 2687 2693 9205078
Kulp SK Yang YT Hung CC Chen KF Lai JP Tseng PH Fowble JW Ward PJ Chen CS 3-phosphoinositide-dependent protein kinase-1/Akt signaling represents a major cyclooxygenase-2-independent target for celecoxib in prostate cancer cells Cancer Res 2004 64 1444 1451 14973075
Panigrahi AR Pinder SE Chan SY Paish EC Robertson JFR Ellis IO The role of PTEN and its signaling pathways, including Akt, in breast cancer; an assessment of relationships with other prognostic factors and with outcome J Pathol 2004 204 93 100 15307142 10.1002/path.1611
Sastre-Garau X Rousseau A Al Ghuzlan A Freneaux P Rosty C Sigal B-Zafrani Couturier J Thiery JP Magdelenat H Vincent-Salomon A Increase cell size and Akt activation in Her-2/neu overexpressing invasive ductal carcinoma of the breast Histopathology 2004 45 142 147 15279632 10.1111/j.1365-2559.2004.01899.x
Schmitz KJ Otterbach F Callies R Levkau B Holscher M Hoffmann O Grabellus F Kimmig R Schmid KW Baba HA Prognostic relevance of activated Akt kinase in node-negative breast cancer: a clinicopathological study of 99 cases Mod Pathol 2004 17 15 21 14631376
Perez-Tenorio G Stal O Activation of AKT/PKB in breast cancer predicts a worse outcome among endocrine treated patients Br J Cancer 2002 86 540 545 11870534 10.1038/sj.bjc.6600126
Stal O Perez-Tenorio G Akerberg L Olsson B Nordenskjold B Skoog L Rutqvist LE Akt kinases in breast cancer and the results of adjuvant therapy Breast Cancer Res 2003 5 R37 R44 12631397 10.1186/bcr569
Arboleda MG Lyons JF Kabbinavar FF Bray MR Snow BE Ayala R Danino M Karlan BY Slamon D Overexpression of Akt2/protein kinase B beta leads to up-regulation of beta1 integrins, increased invasion and metastasis of human breast and ovarian cancer cells Cancer Res 2003 63 196 206 12517798
Ahmad S Singh N Glazer RI Role of AKT1 in 17 b-estradiol and insulin-like growth factor 1 (IGF-1) dependent proliferation and prevention of apoptosis in MCF-7 breast carcinoma cells Biochem Pharm 1999 58 425 430 10424760 10.1016/S0006-2952(99)00125-2
Tseng PH Lin HP Zhu J Chen K Hade EM Young DC Byrd JC Grever M Johnson M Druker B Synergistic interactions between Imatinib and the novel phosphoinositide-Dependent kinase-1 inhibitor OSU-03012 in overcoming Imatinib resistance Blood 2005
Yakes FM Chinratanalab W Ritter CA King W Seelig S Arteaga CL Herceptin-induced inhibition of phosphotidylinositol-3 kinase and Akt is required for antibody-mediated effects on p27, cyclin D1 and antitumor action Cancer Res 2002 62 4132 4141 12124352
Pegram MD Konecny GE O'Callaghan C Beryt M Pietraz R Slamon D Rational combinations of trastuzumab with chemotherapeutic drugs in the treatment of breast cancer J Natl Cancer Inst 2004 96 739 749 15150302
Hermanto U Zong CS Wang LH Erb-B2 overexpressing human mammary carcinoma cells display an increased requirement for the phosphatidylinositol 3-kinase signaling pathway in anchorage-independent growth Oncogene 2001 20 7551 7562 11709727 10.1038/sj.onc.1204964
Dang CT Dannenberg AJ Subbaramaiah K Dickler MN Moasser MM Seidmann AD D'Andrea GM Theodoulou M Panageas KS Norton L Phase II study of Celecoxib and Trastuzumab in metastatic breast cancer patients who have progressed after prior trastuzumab based treatments Clin Cancer Res 2004 10 4062 4067 15217939
|
16168126
|
PMC1242152
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Aug 1; 7(5):R796-R807
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1294
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr12961616812710.1186/bcr1296Research ArticleExpression of hypoxia-inducible factor 1 alpha and its downstream targets in fibroepithelial tumors of the breast Kuijper Arno [email protected] der Groep Petra [email protected] der Wall Elsken [email protected] Diest Paul J [email protected] Department of Pathology, VU University Medical Center, Amsterdam, The Netherlands2 Department of Pathology, University Medical Center Utrecht, The Netherlands3 Division of Internal Medicine and Dermatology, University Medical Center Utrecht, The Netherlands2005 5 8 2005 7 5 R808 R818 9 1 2005 3 3 2005 27 5 2005 5 7 2005 Copyright © 2005 Kuijper et al, licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is cited.
Introduction
Hypoxia-inducible factor 1 (HIF-1) alpha and its downstream targets carbonic anhydrase IX (CAIX) and vascular endothelial growth factor (VEGF) are key factors in the survival of proliferating tumor cells in a hypoxic microenvironment. We studied the expression and prognostic relevance of HIF-1α and its downstream targets in phyllodes tumors and fibroadenomas of the breast.
Methods
The expression of HIF-1α, CAIX, VEGF and p53 was investigated by immunohistochemistry in a group of 37 primary phyllodes tumors and 30 fibroadenomas with known clinical follow-up. The tumor microvasculature was visualized by immunohistochemistry for CD31. Proliferation was assessed by Ki67 immunostaining and mitotic counts. Being biphasic tumors, immunoquantification was performed in the stroma and epithelium.
Results
Only two fibroadenomas displayed low-level stromal HIF-1α reactivity in the absence of CAIX expression. Stromal HIF-1α expression was positively correlated with phyllodes tumor grade (P = 0.001), with proliferation as measured by Ki67 expression (P < 0.001) and number of mitoses (P < 0.001), with p53 accumulation (P = 0.003), and with global (P = 0.015) and hot-spot (P = 0.031) microvessel counts, but not with CAIX expression. Interestingly, concerted CAIX and HIF-1α expression was frequently found in morphologically normal epithelium of phyllodes tumors. The distance from the epithelium to the nearest microvessels was higher in phyllodes tumors as compared with in fibroadenomas. Microvessel counts as such did not differ between fibroadenomas and phyllodes tumors, however. High expression of VEGF was regularly found in both tumors, with only a positive relation between stromal VEGF and grade in phyllodes tumors (P = 0.016). Stromal HIF-1α overexpression in phyllodes tumors was predictive of disease-free survival (P = 0.032).
Conclusion
These results indicate that HIF-1α expression is associated with diminished disease-free survival and may play an important role in stromal progression of breast phyllodes tumors. In view of the absence of stromal CAIX expression in phyllodes tumors, stromal upregulation of HIF-1α most probably arises from hypoxia-independent pathways, with p53 inactivation as one possible cause. In contrast, coexpression of HIF-1α and CAIX in the epithelium in phyllodes tumors points to epithelial hypoxia, most probably caused by relatively distant blood vessels. On the other hand, HIF-1α and CAIX seem to be of minor relevance in breast fibroadenomas.
==== Body
Introduction
After reaching a critical volume of several cubic millimeters, a growing tumor becomes increasingly depleted of oxygen and nutrients, and needs to adapt to its changing microenvironment. In order to survive, tumor cells must develop a vascular system and adapt their metabolism. A key regulator in this process is the transcription factor hypoxia-inducible factor 1 (HIF-1) [1], which controls the expression of several target genes. The protein product of HIF-1 is a heterodimer and consists of two subunits, HIF-1α and HIF-1β. Under normoxic conditions the HIF-1α protein is rapidly degraded. O2-dependent hydroxylation of proline residues in HIF-1α causes binding of the von Hippel–Lindau tumor suppressor protein, which leads to ubiquitilation and subsequent degradation by the proteasome [2,3]. Hypoxia inhibits this process, resulting in upregulation of HIF-1α and its downstream target genes [4]. On the other hand, hypoxia-independent upregulation of HIF-1α may be accomplished by loss of tumor suppressor genes such as the phosphate and tensin homolog deleted on chromosome 10 (PTEN) [5] and the von Hippel–Lindau tumor suppressor gene VHL [6], by activation of oncogenes like v-Src [7] and by stimulation by growth factors such as insulin-like growth factors [8,9] and epidermal growth factor [10].
It has recently been demonstrated that HIF-1α expression is of prognostic value in several types of cancer, including breast cancer [11]. A well-known target of HIF-1α is the vascular endothelial growth factor (VEGF) gene [12]. VEGF is a potent endothelial cell-specific mitogen and is a major participant in the process of angiogenesis, resulting in the formation of microvessels. Both VEGF expression [13,14] and microvessel density [15,16] are established prognosticators in many types of cancer. Upregulation of carbonic anhydrase 9 (CA9) gene expression was found to be dependent on HIF-1α [17]. The protein product of the CA9 gene, carbonic anhydrase IX (CAIX), catalyzes the hydration of carbon dioxide to carbonic acid and contributes to acidification of the surrounding microenvironment. CAIX is constitutively expressed in cells lining the alimentary tract [18]. Because of its correlation with lowered pO2 in carcinoma of the cervix, CAIX expression is regarded as an intrinsic marker of hypoxia [19]. Furthermore, CAIX expression has recently been related to poor outcome in breast cancer [20]. It therefore seems that HIF-1α and its downstream targets play pivotal roles in the development and progression of cancer.
Breast fibroadenoma and phyllodes tumor are both fibroepithelial tumors; that is, they are composed of an epithelial component and a stromal component. The distinction between both tumors may be difficult [21]. The behavior of fibroadenomas is benign, however, in contrast to phyllodes tumors that can recur and can even metastasize. In a small number of cases, fibroadenoma may progress to phyllodes tumor [22,23]. Little is known of the mechanisms by which these tumors maintain a steady supply of nutrients as they grow. Few studies have addressed expression patterns of angiogenic growth factors in fibroepithelial breast tumors. Expression of basic fibroblast growth factor (basic FGF), FGF receptor and VEGF was found at higher levels in stroma of phyllodes tumors in comparison with fibroadenomas [24]. Unfortunately, no information on the epithelial component was provided in this work. Expression patterns of platelet-derived growth factor (PDGF) and the PDGF receptor suggest the presence of autocrine loops in stroma and paracrine stimulation of stroma by the epithelium [25]. A similar autocrine and paracrine loop has been described for acidic FGF and FGF receptor 4 in fibroadenomas [26]. These studies suggest that stromal proliferation may be stimulated by secretion of mitogens by the epithelial compartment. Interestingly, hypoxia may stimulate both FGF and PDGF expression [27,28]. It would therefore be interesting to evaluate the role of HIF-1α and its downstream effectors in the tumorigenesis of biphasic breast tumors.
We therefore studied expression of HIF-1α and its downstream targets VEGF and CAIX in breast phyllodes tumors of various grades and in fibroadenomas. Furthermore, since HIF-1α seems to play a major role in the process of angiogenesis, we evaluated the microvascular network by counting of CD31-positive microvessels. Proliferation, as an important functional end point of various carcinogenic processes, was assessed by mitotic counts and Ki67 expression. Since HIF-1α degradation may be promoted by wild-type p53 [29], we evaluated the relation between p53 and HIF-1α expression in phyllodes tumors. Finally, the prognostic value of HIF-1α, its downstream effectors and microvessel counts was also assessed for phyllodes tumors.
Materials and methods
Tissue samples
Formaldehyde-fixed and paraffin-embedded tissue samples were retrieved from the archives of our hospitals. A total of 37 primary phyllodes tumors and 30 fibroadenomas were acquired. The presence of epithelial proliferative changes was noted. Phyllodes tumors were graded as benign, borderline or malignant based on the degree of stromal cellularity, the degree of stromal overgrowth, the degree of cellular atypia, invasiveness of the tumor margin and the mitotic activity index as described previously in detail by Moffat and colleagues [30]. Mitotic figures were counted using established criteria in 10 consecutive high-power fields at a 400 × magnification [31]. As the grading of phyllodes tumors may be complicated by intratumoral heterogeneity, tumors were graded in the most unfavorable areas provided they comprised at least 10% of the total tumor area [30]. Clinical data were gathered by studying medical records.
Immunohistochemistry
Four-micrometer sections were cut and mounted on coated slides. After deparrafinization and rehydration, sections were immersed in methanol containing 0.3% hydrogen peroxide to stop endogenous peroxidase activity. No antigen retrieval was necessary for CAIX staining.
Antigen retrieval for HIF-1α was performed in target retrieval solution (DAKO, Glostrup, Denmark) in a water bath at 97°C for 45 min, and the remaining antigens were unmasked by microwaving the slides for 10–15 min in citrate buffer (pH 6.0). The following panel of mouse monoclonal antibodies was used: VEGF (1:50, R&D Systems, Abingdon, U.K), HIF-1α (1:500; BDTransduction Laboratories, Lexington, Kentucky, USA), CAIX (M75, 1:50; kind gift from Dr S Pastorekova (Institute of Virology, Academy of Science, Bratislava, Slovakia)), CD31 (JC/70 A, 1:40; DAKO), Ki67 (MIB-1, 1:40; DAKO) and p53 (DO-7, 1:500; DAKO). VEGF, Ki67 and p53 were incubated overnight at 4°C, the incubation time for HIF-1α and CAIX was 30 min at room temperature, and CD31 was incubated for 1 hour at room temperature. VEGF, Ki67 and p53 were detected by application of a secondary biotinylated rabbit anti-mouse antibody (diluted 1:500; DAKO) followed by incubation with avidin–biotin-peroxidase complex (1:200 dilution; DAKO). The Catalyzed Signal Amplification system (DAKO) was used to detect HIF-1α, CD31 was visualized with the Ultravision system (Labvision) and CAIX was detected with the Envision system peroxidase (DAKO). All stainings were developed with 3,3'-diaminobenzidine tetrahydrochloride. Counter staining was performed with hematoxylin. Negative controls were obtained by omitting the primary antibody, and appropriate positive controls were included throughout.
As fibroadenomas and phyllodes tumors are biphasic tumors, immunostainings were scored in both stroma and the epithelium. Immunoquantification was performed simultaneously by two observers (AK and PJvD) behind a double-headed microscope. The percentage of positive staining nuclei for HIF-1α, Ki67 and p53 was estimated on a continuous scale, regarding only homogeneously and darkly stained nuclei as positive. CAIX expression was scored as positive or negative, scoring a case positive when membranous staining in any amount was present. Cytoplasmic VEGF staining was scored semi-quantitatively in four categories from 0 to +++, with category 0 expressing no positive staining, category + showing focal or diffuse weak staining, and categories ++ and +++ displaying focal or widespread strong staining, respectively. Counting of CD31-positive microvessels was performed in the microvessel hot-spot in four adjacent fields of vision at 400 × magnification as described elsewhere [15,16]. In addition, a global microvessel density was acquired by counting of vessels in 10 diagonally adjacent fields at a magnification of 400 × from a random starting point generated semi-automatically by the QPRODIT interactive digitizing video overlay system (Leica, Cambridge, UK).
For statistical analysis, stainings for Ki67 and p53 were divided into high and low using 10% positive staining as the cut-off [32]. Categories 0 and + staining for VEGF were grouped as low expression and ++/+++ was stated high, and microvessel counts were dichotomized using the median value. As determined by staining of normal breast tissue, preinvasive breast lesions and invasive breast cancer, HIF-1α overexpression was defined as ≥1% positive staining nuclei [33].
The mean shortest distances from microvessels to the epithelial basal membrane were determined with the QPRODIT system. Using CD31-stained sections, the distances of the epithelium to the nearest microvessel were measured between manually placed markers. Means were calculated from a minimum of 50 measurements per case with fields of view selected according to a systematic random sampling method [34].
Statistical analysis
All statistical analyses were performed with SPSS software (SPSS, Chicago, IL, USA). Differences in expression of HIF-1α, of Ki67, of VEGF and of CAIX and the microvessel density between fibroadenomas and different grades of phyllodes tumors were investigated by the chi-square test. Correlations between markers were evaluated using Fisher's exact test. The clinical endpoint for survival analysis of phyllodes tumors was local or distant recurrence (disease-free survival). Wide local excision is the preferred treatment for phyllodes tumors. Patients treated by primary mastectomy were therefore excluded from survival analysis since these would bias results due to strongly reduced chances of recurrence. Kaplan–Meier curves were plotted and differences between the curves were evaluated with the log-rank test. P values below 0.05 were regarded as significant.
Results
Patient characteristics
Thirty fibroadenomas were analyzed. The mean age of patients was 33.7 ± 10.5 years and the mean tumor size was 1.6 ± 0.6 cm. Ten fibroadenomas (33%) harbored epithelial proliferative changes.
A total of 18 benign primary phyllodes tumors, eight borderline primary phyllodes tumors and 11 malignant primary phyllodes tumors were identified. Epithelial hyperplasia of the ductal type, mostly focal, was found in 16 cases (43%) and was not related to grade. The mean age of patients with benign, borderline and malignant phyllodes tumors was 44.4 ± 17.4, 57.9 ± 12.8 and 54.3 ± 12.9 years, respectively (P = 0.073). Mean tumor sizes were 4.8 ± 2.4, 7.1 ± 7.0 and 4.8 ± 2.5 cm for benign, borderline and malignant phyllodes tumors, respectively (P = 0.924). Patients with phyllodes tumors were older and had larger tumors (P < 0.001 both) compared with those patients with fibroadenomas. Five patients with phyllodes tumors were treated by primary mastectomy, the remainder by local excision. Excision was incomplete for 13 phyllodes tumors, whereas for six phyllodes tumors this information could not be retrieved.
Differences between fibroadenomas and phyllodes tumors
Table 1 summarizes the differences in immunostaining between fibroadenomas and phyllodes tumors. Stromal and epithelial overexpression of HIF-1α were found almost exclusively in phyllodes tumors (P = 0.001 and P < 0.001, respectively; Fig. 1b,d). Only two fibroadenomas with low-level stromal HIF-1α immunoreactivity (1% and 2% positive nuclei) were found without concerted CAIX expression. Stromal and epithelial CAIX expression was only seen in phyllodes tumors (Fig. 1a,c). High expression of VEGF was regularly found in both fibroadenomas and phyllodes tumors. As a result, no differences in stromal and epithelial VEGF expression were found between both tumors.
Many small microvessels were concentrated around the epithelium of fibroadenomas. In phyllodes tumors, microvessels lacked such a peri-epithelial preference and were distributed more evenly throughout the tumor. The number of microvessels as such, counted by both methods, did not differ significantly between fibroadenomas and phyllodes tumors. Indeed, when regarding microvessel counts on a continuous scale, a large overlap was observed between fibroadenomas, benign phyllodes tumors and borderline phyllodes tumors (Fig. 2). In five fibroadenomas and 10 phyllodes tumors, of which five had stretches of HIF-1α-positive epithelium, the mean shortest distance between microvessels and the epithelial basal membrane was 50 ± 11 μm for fibroadenomas and was 83.3 ± 16 μm for phyllodes tumors (P = 0.005). Because of the focal staining patterns resulting in methodological problems and low numbers, phyllodes tumors with HIF-1α-negative and HIF-1α-positive epithelium were not separately analyzed here. Due to these methodological problems the results should be interpreted with caution.
Fibroadenomas and phyllodes tumors did not differ with regard to the presence of epithelial proliferative changes (P = 0.458).
Differences between phyllodes tumor grades
Correlations between the grade of phyllodes tumor and HIF-1α and its downstream targets, microvessels and proliferation are summarized in Table 2. Stromal HIF-1α overexpression in phyllodes tumors was strongly correlated with tumor grade (P = 0.001), with all malignant tumors displaying overexpression. Necrosis could be detected in only one malignant phyllodes tumor, with no typical peri-necrotic HIF-1α overexpression pattern. Epithelial HIF-1α overexpression was not related to grade (P = 0.323). Only two malignant phyllodes tumors displayed stromal CAIX expression (P = 0.090). Epithelial CAIX expression was seen more often than stromal CAIX expression, but was not correlated to grade (P = 0.735). Overexpression of HIF-1α and expression of CAIX in the epithelial component were both not related to hyperplasia and were mostly found in normal appearing, two-layered epithelium.
A statistically significant difference in the number of microvessels was found between grades, both when counted in the hot-spot and by the global method (P = 0.003 and P = 0.002, respectively). The number of microvessels was strongly increased in malignant tumors, as compared with benign and borderline tumors (Fig. 2). High VEGF expression in the stromal component displayed a positive relation with tumor grade (P = 0.016).
HIF-1α expression in both compartments of phyllodes tumors was not significantly related to tumor size, both with size cut-off at the median and when used as a continuous variable (data not shown).
Coexpression of markers
The relations between HIF-1α expression and its downstream targets, microvessels and proliferation markers in the stromal component of fibroepithelial tumors are presented in Table 3. Since only two cases showed stromal CAIX expression, the relation between stromal CAIX and stromal HIF-1α overexpression did not reach statistical significance (P = 0.098). The relation between HIF-1α overexpression and strong VEGF expression in stroma reached borderline significance (P = 0.098). Global (P = 0.015) and hot-spot (P = 0.031) microvessel counts were both related to stromal HIF-1α overexpression. Overexpression of HIF-1α in stroma was correlated with proliferation as measured by high stromal Ki67 expression (P < 0.001) and by high mitotic activity index (P < 0.001). Stromal overexpression of HIF-1α and p53 were also strongly associated (P = 0.003).
Epithelial HIF-1α overexpression was related to epithelial CAIX expression (P = 0.014). There was no relation between HIF-1α overexpression in the epithelial component and VEGF or Ki67 expression. Epithelial HIF-1α overexpression was related to increased stromal proliferation (Ki67, P = 0.006; mitotic activity index, P = 0.004).
Survival analysis
As already mentioned, patients treated by mastectomy were excluded from survival analysis. Two patients were lost to follow-up. Ten of the remaining 30 tumors recurred. Most were local recurrences, but one malignant tumor metastasized to the lung. By univariate analysis of survival we found an inverse correlation between stromal HIF-1α overexpression (P = 0.032; Fig. 3), tumor grade (P = 0.039), stromal Ki67 overexpression (P = 0.028) and stromal p53 overexpression (P < 0.001) and disease-free survival, as presented in Table 4. HIF-1α expression did not add to the prognostic power of p53 expression alone. The tumor size (median, 3 cm; P = 0.88), age (P = 0.36), the mitotic activity index (P = 0.09), the microvessel counts (hot-spot P = 0.54; global P = 0.25), the margin status (P = 0.23), the VEGF expression (stromal P = 0.96; epithelial P = 0.11), the CAIX expression (stromal P = 0.46; epithelial P = 0.16) and the epithelial HIF-1α expression (P = 0.76) were not of prognostic relevance. Surprisingly, epithelial overexpression of Ki67 also had prognostic power (P = 0.011). This is of no relevance, however, since only two cases showed epithelial overexpression of Ki67.
Discussion
Evidence is mounting that HIF-1α and its downstream targets play pivotal roles in the development and progression of many types of human cancers. This is the first study evaluating HIF-1α and CAIX expression in breast phyllodes tumors. We found that stromal HIF-1α overexpression predicts the prognosis of, and may play an important role in, the stromal progression of phyllodes tumors. Surprisingly, we regularly found concerted HIF-1α and CAIX expression in normal appearing epithelium of phyllodes tumors. Since only two fibroadenomas displayed low-level stromal HIF-1α immunoreactivity and none expressed CAIX, HIF-1α and CAIX seem to be of little relevance in these tumors.
HIF-1α overexpression is induced by hypoxia and by oxygen-independent mechanisms. In our study, stromal HIF-1α overexpression was related to the grade of phyllodes tumors, with a marked increase from borderline to malignant grade. Benign and borderline tumors showed comparable levels of HIF-1α overexpression. We only found two fibroadenomas with low-level expression of HIF-1α in the stroma, in agreement with Zhong and colleagues, who found all their fibroadenomas to be negative [35]. By clonality analysis we previously demonstrated that fibroadenomas may progress to phyllodes tumors by clonal expansion of stroma [22]. Perhaps positive stromal HIF-1α staining in a fibroadenoma may reflect an increased intrinsic capacity to progress to phyllodes tumor. The fibroadenomas with stromal HIF-1α positivity were microscopically inconspicuous, however.
In several types of cancer, hypoxia-induced HIF-1α expression seems to be characterized by a peri-necrotic distribution, whereas oxygen-independent overexpression results in a diffuse pattern of HIF-1α immunoreactivity [11,36,37]. Necrosis was found in the stromal component of only one malignant phyllodes tumor. Furthermore, stromal HIF-1α overexpression was not accompanied by an increase in CAIX expression, which is regarded as a marker of hypoxia [19]. Indeed, our group recently showed that, in contrast to homogeneous/diffuse HIF-1α expression, necrosis-related/focal HIF-1α expression is accompanied by CAIX expression [38]. All this suggests that HIF-1α upregulation in stroma of phyllodes tumors is normoxic and may be caused by changes in the stromal expression of oncogenes, tumor suppressor genes or growth factors. Wild-type p53 has been shown to promote MDM2-mediated ubiquitination of HIF-1α [29]. Therefore, p53 inactivation by gene mutation has been implicated in increased HIF-1α expression. We showed that aberrant expression of cell-cycle protein p53 only occurred in the stromal component. The transition from borderline to malignant phyllodes tumors was accompanied by a strong increase in stromal p53 expression, similar to stromal HIF-1α expression. Furthermore, we found a significant relation between stromal p53 and HIF-1α overexpression.
Another candidate for hypoxia-independent upregulation of HIF-1α in phyllodes tumors is PDGF. Overexpression of PDGF was found in 24% of phyllodes tumors, but not in fibroadenomas [25]. Interestingly, PDGF may induce HIF-1α expression [39]. Stromal HIF-1α and p53 expression in phyllodes tumors were predictive of disease-free survival, underlining the importance of the p53-HIF-1α axis in it's progression and clinical behavior.
Epithelial HIF-1α overexpression in phyllodes tumors was associated with CAIX expression. This suggests a causative role for epithelial hypoxia. Fibroadenomas were negative for CAIX, confirming the results of Bartosova and colleagues [40]. The mean distance from microvessels to the epithelium was significantly higher for phyllodes tumors as compared with fibroadenomas, suggesting that the hypoxic microenvironment in the epithelial component of phyllodes tumors is caused by relatively distant microvessels. Most vessels in phyllodes tumors were within 150 μm of the epithelial basal membrane, however, which is a critical distance for necrosis [41]. Still, it is conceivable that rapidly growing stroma stretches the two-layered epithelium in such way that it's oxygen supply does not keep up, resulting in a state of mild hypoxia. Indeed, a previous report detected CAIX expression 80 μm from the nearest blood vessel [42], which is comparable with the distance we found for phyllodes tumors. It is possible that the decreasing the oxygen gradient between 80 and 150 μm from the nearest vessel to the epithelium is sufficient to induce HIF-1α and CAIX expression.
The scattered foci of positive staining for HIF-1α and CAIX suggest the presence of focal mild hypoxia in the epithelial component. Microenvironmental disturbance of normal tissue by adjacent malignant disease has been described previously for HIF-1α and CAIX [40,43,44]. On the other hand, several other possible causes exist for epithelial HIF-1α overexpression. Epithelial PDGF expression, which is found in most phyllodes tumors [25], may cause HIF-1α overexpression [39]. Shpitz and colleagues found expression of HER-2/neu in the epithelium of 61% of phyllodes tumors [45]. It was recently demonstrated that enhanced HER-2/neu signaling induces HIF-1α protein expression [46]. Numerous factors interact with HIF-1 and other as yet unidentified changes in the epithelium of phyllodes tumors may cause HIF-1α overexpression.
The tumor biological significance of increased epithelial HIF-1α and CAIX expression is unclear. Morphologically, the epithelial component was mostly two-layered epithelium without atypia. Furthermore, clonality studies mostly found the epithelial component of phyllodes tumors to be polyclonal [22,23]. Moreover, phyllodes tumor metastases are composed of stroma, with only one case described in which the epithelial component also disseminated, leading to a biphasic metastasis [47]. Finally, epithelial HIF-1α and CAIX expression are not predictive of prognosis. It seems therefore that HIF-1α and CAIX expression in the epithelium of phyllodes tumors merely reflects a physiological adaptation to microenvironmental disturbance by rapidly proliferating stroma with lagging peri-epithelial angiogenesis. However, possible upregulation of growth factors in the epithelium by HIF-1α may exert an additional stimulatory force on the stromal component. Indeed, it has been suggested that in the early stages of phyllodes tumor development the epithelium secretes mitogens stimulating the stromal component [48,49]. The presence of such autocrine and paracrine loops has been described for PDGF/PDGF receptor [25]. Furthermore, stimulation of the stromal component by the epithelium was suggested previously by studying endothelin-1 expression in epithelium of phyllodes tumors [50]. Interestingly, endothelin-1 turned out to be a target of HIF-1α [51].
An abrupt increase in number of microvessels was observed in malignant phyllodes tumors, coinciding with a strong increase in stromal overexpression of HIF-1α from borderline grade to malignant grade. In contrast with a previous report [52], we found no difference in microvessel counts between benign phyllodes tumors and borderline phyllodes tumors. In addition, when regarding hot-spot and global counts we found that there was no difference between fibroadenomas, benign phyllodes tumors and borderline phyllodes tumors. This is in contrast with a previous report claiming a difference between fibroadenomas and phyllodes tumors [24]. On the other hand, Weind and colleagues found a large overlap in microvessel counts between fibroadenomas and invasive breast cancers, demonstrating that fibroadenomas are capable of producing large numbers of microvessels [53]. We found large numbers of small peri-epithelial microvessels in fibroadenomas responsible for its high numbers of microvessels. It seems that microvessel counts are not helpful in the differentiation between fibroadenomas and benign phyllodes tumors, which poses the biggest diagnostic problem.
VEGF expression in fibroadenomas did not differ from that in phyllodes tumors. Several growth factors such as FGF-4 [54], PDGF [55] and transforming growth factor beta [56] may stimulate VEGF expression. Expression of FGF [26], PDGF [25] and transforming growth factor beta [57] have been described in fibroadenomas and may contribute to HIF-1α-independent expression of VEGF. In phyllodes tumors, the relation between VEGF expression and HIF-1α reached borderline significance. Although HIF-1α most probably contributes to VEGF expression in phyllodes tumors, VEGF expression seems, at least in part, to be independent from HIF-1α.
In view of its annual incidence of 2.1 cases per 1 million women [58], we feel that our group of phyllodes tumors is reasonably sized and well composed with all grades present. Still, future investigations confirming our results in larger series are warranted. Furthermore, experiments covering a variety of molecular elements such as DNA microarray techniques may unravel the complex mechanisms underlying the presumed nonhypoxic upregulation of HIF-1α in the stromal component of phyllodes tumors and its subsequent influence on tumor progression.
Conclusion
This is the first report on HIF-1α and CAIX expression in breast phyllodes tumors. Our results show that HIF-1α is related to diminished disease-free survival and may play an important role in stromal progression of phyllodes tumors. The significant relation between tumor grade and stromal HIF-1α overexpression underlines its importance, with all malignant tumors showing HIF-1α overexpression. Stromal HIF-1α overexpression in phyllodes tumors most probably arises from hypoxia-independent pathways, with p53 inactivation as one possible cause. Surprisingly, the normal-appearing epithelium in phyllodes tumors frequently displayed HIF-1α and CAIX expression. The distance from the nearest vessels to the epithelium was higher in phyllodes tumors as compared with fibroadenomas and a hypoxic effect seems plausible here, although the effect is of doubtful biological significance. In contrast to phyllodes tumors, HIF-1α seems of minor relevance in the tumorigenesis of fibroadenomas. Considering it's possible role in progression of the stromal component of phyllodes tumors and the fact that metastases are composed of stroma, novel therapies targeting HIF-1α [59] may contribute to the treatment of disseminated phyllodes tumor, which is poorly responsive to conventional chemotherapy and radiotherapy.
Abbreviations
CAIX = carbonic anhydrase IX; FGF = fibroblast growth factor; HIF-1 = hypoxia-inducible factor 1; PDGF = platelet-derived growth factor; VEGF = vascular endothelial growth factor.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AK was responsible for generating the hypothesis, for collecting patient material, for immunostaining, for classification of immunostainings, for statistics and for writing the manuscript. PvdG was responsible for immunostaining and for correcting the manuscript. EvdW was responsible for generating the hypothesis and for correcting the manuscript. PJvD was responsible for generating the hypothesis, for collecting patient material, for classification of immunostainings and for correcting the manuscript.
Acknowledgements
This work was supported in part by an unrestricted grant from Aegon Inc. The authors thank JL Peterse, MD, Rob AI de Vos, MD, and Jaap H Lagendijk, MD, for providing tumor material for some of the patients and thank Dr S Pastorekova and Prof. AL Harris who kindly supplied the M75 antibody.
Figures and Tables
Figure 1 Examples of immunostaining for hypoxia-inducible factor 1 alpha (HIF-1α) and carbonic anhydrase IX (CAIX) in breast phyllodes tumors. (a) Malignant phyllodes tumor with stromal CAIX expression. (b) Same tumor as (a) with topographically overlapping HIF-1α overexpression. (c) Benign phyllodes tumor with CAIX-positive staining epithelium. (d) Borderline phyllodes tumor with HIF-1α overexpression in normal appearing epithelium and in subepithelial stroma.
Figure 2 Hot-spot microvessel counts in fibroadenomas and phyllodes tumors. Boxplot showing the large overlap in the numbers of microvessels between fibroadenomas (FA), benign phyllodes tumors (BePT) and borderline phyllodes tumors (BoPT) when counted in the hot-spot. MaPT, malignant phyllodes tumor.
Figure 3 Disease-free survival according to hypoxia-inducible factor 1 alpha (HIF-1α) status. The Kaplan–Meier survival curve illustrating disease-free survival for patients with breast phyllodes tumors with high expression (≥1%) versus low expression (<1%) of HIF-1α in the stromal component. Numbers of patients at risk at different time points are displayed below the horizontal axis.
Table 1 Differences in the expression of hypoxia related proteins, microvasculature and proliferation in fibroepithelial breast tumors
Marker Fibroadenoma (n = 30) Phyllodes tumor (n = 37) P valuea
Stroma
HIF-1α 2 (7%) 19 (51%) 0.001
CAIX 0 (0%) 2/36 (6%) 0.497
VEGF 7 (23%) 17 (47%) 0.074
Ki67 1 (3%) 17 (47%) <0.001
Hot-spot microvessel count 13 (43%) 22/35 (63%) 0.140
Global microvessel count 13 (43%) 20/35 (57%) 0.324
p53 0(0%) 7/36(19%) 0.014
Epithelium
HIF-1α 0 (0%) 15/35 (43%) <0.001
CAIX 0 (0%) 9/35 (26%) 0.003
VEGF 22 (73%) 22/36 (61%) 0.432
Ki67 2 (7%) 2/36 (6%) 1.000
p53 0(0%) 0(0%) -
Due to empty blocks, repeated unsuccessful staining attempts or too little epithelium in a malignant tumor, numbers varied slightly for some stainings. Dichotomized values were used. aUsing the chi-square test, P values below 0.05 were regarded as significant. HIF-1α = hypoxia-inducible factor 1 alpha, CAIX = carbonic anhydrase IX, VEGF = vascular endothelial growth factor.
Table 2 Differences in expression of hypoxia-related proteins, microvasculature and proliferation between different grades of phyllodes tumors
Marker Phyllodes tumor P valuea
Benign (n = 18) Borderline (n = 8) Malignant (n = 11)
Stroma
HIF-1α 5 (31%) 3 (38%) 11 (100%) 0.001
CAIX 0 (0%) 0 (0%) 2 (18%) 0.090
VEGF 5 (31%) 3 (38%) 9 (82%) 0.016
Ki67 1 (6%) 5 (63%) 11 (100%) <0.001
Hot-spot microvessel count 9/16 (56%) 2 (25%) 11 (100%) 0.003
Global microvessel count 8/16 (50%) 2 (25%) 11 (100%) 0.002
p53 1/18(6%) 1/8(13%) 5/11(46%) 0.025
Epithelium
HIF-1α 7/17 (42%) 2 (25%) 6/10 (60%) 0.323
CAIX 5 (32%) 1/7 (14%) 3/10 (30%) 0.735
VEGF 12 (67%) 5 (63%) 5/10 (50%) 0.684
Ki67 1 (6%) 0 (0%) 1/10 (10%) 0.655
p53 0(0%) 0(0%) 0(0%) -
Due to empty blocks, repeated unsuccessful staining attempts or too little epithelium in a malignant tumor, numbers varied slightly for some stainings. Dichotomized values were used. aUsing the chi-square test, P values below 0.05 were regarded as significant. HIF-1α = hypoxia-inducible factor 1 alpha, CAIX = carbonic anhydrase IX, VEGF = vascular endothelial growth factor.
Table 3 Association of hypoxia-inducible factor 1 alpha (HIF-1α) with its downstream effectors, microvessel counts and proliferation markers in stroma of fibroepithelial tumors
HIF-1α P valuea
<1% ≥1%
p53
<10% 45 15 0.003
≥10% 1 6
Vascular endothelial growth factor
Weak 33 10 0.098
Strong 13 11
Hot-spot microvessel count
<71 25 5 0.031
≥71 20 15
Global microvessel count
<86 27 5 0.015
≥86 18 15
Carbonic anhydrase IX
Negative 45 19 0.098
Positive 0 2
Ki67
<10% 42 7 <0.001
≥10% 4 14
Mitotic activity index
<10 43 8 <0.001
≥10 3 13
Dichotomized values were used.
a Fisher's exact test, P values below 0.05 were regarded as significant.
Table 4 Prognostically significant variables in phyllodes tumors as determined by univariate analysis of disease-free survival (DFS)
Variable Number of patients Number of patients with disease (%) Number of disease-free patients (%) Mean DFS (months [95% confidence interval]) P value Log-rank
p53 stroma
<10% 24 5 (21) 19 (79) 173 (142–204) <0.001 10.97
≥10% 6 5 (83) 1 (16) 61 (23–98)
HIF-1α stroma
<1% 15 3 (20) 12 (80) 180 (146–214) 0.032 4.59
≥1% 15 7 (47) 8 (53) 103 (64–142)
Ki67 stroma
<10% 16 4 (25) 12 (75) 176 (145–207) 0.028 4.85
≥10% 14 6 (43) 8 (57) 91 (55–127)
Grade
Benign 15 4 (27) 11 (73) 165 (130–200) 0.039 6.48
Borderline 5 1 (20) 4 (80) 140 (86–194)
Malignant 10 5 (50) 5 (50) 71 (32–109)
HIF-1α, hypoxia-inducible factor 1 alpha.
==== Refs
Semenza GL Wang GL A nuclear factor induced by hypoxia via de novo synthesis binds to the human erythropoeitin gene enhancer at a site required for transcriptational activation Mol Cell Biol 1992 12 5447 5454 1448077
Jaakkola P Mole DR Tian YM Wilson MI Gielbert J Gaskell SJ Kreigsheim A Hebestreit HF Mukherji M Schofield CJ Maxwell PH Targeting of HIF-α to the von Hippel–Lindau ubiquitination complex by O2-regulated prolyl hydroxylation Science 2001 292 464 468 11292862
Tanimoto K Makino Y Pereira T Poellinger L Mechanism of regulation of the hypoxia-inducible factor-1α by the von Hippel–Lindau tumor protein EMBO J 2000 19 4298 4309 10944113 10.1093/emboj/19.16.4298
Sutter CH Laughner E Semenza GL Hif-1α protein expression is controlled by oxygen-ubiquitination that is disrupted by deletions and missense mutations Proc Natl Acad Sci 2000 97 4748 4753 10758161 10.1073/pnas.080072497
Zundel W Schindler C Haas-Kogan D Koong A Kaper F Chen E Gottschalk AR Ryan HE Johnson RS Jefferson AB Loss of PTEN facilitates HIF-1 mediated gene expression Genes Dev 2000 14 391 396 10691731
Maxwell PH Wiessener MS Chang GW Clifford SC Vaux EC Cockman ME Wykoff CC Pugh CW Maher ER Ratcliffe PJ The tumour suppressor protein VHL targets hypoxia-inducible factors for oxygen dependent proteolysis Nature 1999 399 271 275 10353251 10.1038/20459
Jiang BH Agani F Oassaniti A Semenza GL V-SRC induces expression of hypoxia-inducible factor 1 (HIF-1) and transcription of genes encoding vascular endothelial growth factor and enolase 1: involvement of HIF-1 in tumour progression Cancer Res 1997 57 5328 5335 9393757
Feldser D Agani F Iyer NV Pak B Ferreira G Semenza GL Reciprocal positive regulation of hypoxia-inducible factor 1alpha and insulin-like growth factor 2 Cancer Res 1999 59 3915 3918 10463582
Zelzer E Levy Y Kahana C Shilo BZ Rubinstein M Cohen B Insulin induces transcription of target genes through the hypoxia-inducible factor HIF-1alpha/ARNT EMBO J 1998 17 5085 5094 9724644 10.1093/emboj/17.17.5085
Zhong H Chiles K Feldser D Laughner E Hanrahan C Georgescu MM Simons JW Semenza GL Modulation of hypoxia-inducible factor 1alpha expression by the epidermal growth factor/phosphatidylinositol 3-kinase/PTEN/AKT/FRAP pathway in human prostate cancer cells: implications for tumor angiogensis and therapeutics Cancer Res 2000 60 1541 1545 10749120
Bos R van der Groep P Greijer AE Shvarts A Meijer S Pinedo HM Semenza GL van Diest PJ van der Wall E Levels of hypoxia-inducible factor-1α independently predict prognosis in patients with lymph node negative breast cancer Cancer 2003 97 1573 1581 12627523 10.1002/cncr.11246
Forsythe JA Jiang BH Iyer NV Agani F Leung SW Koos RD Semenza GL Activation of vascular endothelial growth factor gene transcription by hypoxia-inducible factor 1 Mol Cell Biol 1996 16 4604 4613 8756616
Linderholm B Lindh B Tavelin B Grankvist K Henriksson R p53 and vascular-endothelial-growth-factor (VEGF) expression predicts outcome in 833 patients with primary breast carcinoma Int J Cancer 2000 89 51 62 10719731 10.1002/(SICI)1097-0215(20000120)89:1<51::AID-IJC9>3.3.CO;2-#
Gasparini G Toi M Gion M Verderio P Dittadi R Hanatani M Matsubara I Vinante O Bonoldi E Boracchi P Prognostic significance of vascular endothelial growth factor protein in node-negative breast carcinoma J Natl Cancer Inst 1997 89 139 147 8998183 10.1093/jnci/89.2.139
Weidner N Folkman J Pozza F Bevilacqua P Allred EN Moore DH Meli S Gasparini G Tumor angiogenesis: a new significant and independent prognostic indicator in early-stage breast cancer J Natl Cancer Inst 1992 84 1875 1887 1281237
De Jong JS van Diest PJ Baak JPA Hot spot microvessel density and the mitotic activity index are strong additional prognostic indicators in invasive breast cancer Histopathology 2000 36 306 312 10759944 10.1046/j.1365-2559.2000.00850.x
Wykoff CC Beasley NJP Watson PH Turner KJ Pastorek J Sibtain A Wilson GD Turley H Talks KL Maxwell PH Hypoxia-inducible expression of tumor-associated carbonic anhydrases Cancer Res 2000 60 7075 7083 11156414
Pastorekova S Parkkila S Parkkila AK Opavsky R Zelnik V Saarnio J Pastorek J Carbonic anhydrase IX, MN/CA IX: analysis of stomach complimentary DNA sequence and expression in human and rat alimantary tracts Gastroenterology 1997 112 398 408 9024293 10.1053/gast.1997.v112.pm9024293
Loncaster JA Harris AL Davidson SE Logue JP Hunter RD Wykoff CC Pastorek J Ratcliffe PJ Stratford IJ West CM Carbonic anhydrase (CA IX) expression, a potential new intrinsic marker of hypoxia: correlation with tumor oxygen measurements and prognosis in locally advanced carcinoma of the cervix Cancer Res 2001 61 6394 6399 11522632
Chia SK Wykoff CC Watson PH Han C Leek RD Pastorek JD Gatter KC Ratcliffe P Harris AL Prognostic significance of a novel hypoxia-regulated marker, carbonic anhydrase IX, in invasive breast cancer J Clin Oncol 2001 19 3660 3668 11504747
Kuijper A Mommers EC van der Wall E van Diest PJ Histopathology of fibroadenoma of the breast Am J Clin Pathol 2001 115 736 742 11345838 10.1309/F523-FMJV-W886-3J38
Kuijper A Buerger H Simon R Schaefer KL Croonen A Boecker W van der Wall E van Diest PJ Analysis of progression of fibroepithelial tumors of the breast by PCR based clonality assay J Pathol 2002 197 575 581 12210075 10.1002/path.1161
Noguchi S Motomura K Inaji H Imaoka S Koyama H Clonal analysis of fibroadenoma and phyllodes tumor of the breast Cancer Res 1993 53 4071 4074 8395336
Hasebe T Imoto S Sasaki S Tsubono Y Mukai K Proliferative activity and tumor angiogenesis is closely correlated to stromal cellularity of fibroadenoma: proposal fibroadenoma, cellular variant Pathol Int 1999 49 435 443 10417687 10.1046/j.1440-1827.1999.00892.x
Feakins RM Wells CA Young KA Sheaff MT Platelet-derived growth factor expression in phyllodes tumors and fibroadenomas of the breast Hum Pathol 2000 31 1214 1222 11070114 10.1053/hupa.2000.18481
La Rosa S Sessa F Colombo L Tibiletti MG Furlan D Capella C Expression of acidic fibroblast growth factor (aFGF) and fibroblast growth factor receptor 4 (FGFR4) in breast fibroadenomas J Clin Pathol 2001 54 37 41 11271786 10.1136/jcp.54.1.37
Koong AC Denko NC Hudson KM Schindler C Swiersz L Koch C Evans S Ibrahim H Le QT Terris DJ Giaccia AJ Candidate genes for the hypoxic tumor phenotype Cancer Res 2000 60 883 887 10706099
Kuwabara K Ogawa S Matsumoto M Koga S Clauss M Pinsky DJ Lyn P Leavy J Witte L Joseph-Silverstein J Hypoxia-mediated induction of acidic/basic fibroblast growth factor and platelet-derived growth factor in mononuclear phagocytes stimulates rgowth of hypoxic endothelial cells Proc Natl Acad Sci 1995 92 4606 4610 7538678
Ravi R Mookerjee B Bhujwalla ZM Sutter CH Artemov D Zeng Q Dillehay LE Madan A Semenza GL Bedi A Regulation of tumor angiogenesis by p53 induced degradation of hypoxia inducible factor 1 alpha Genes Dev 2000 14 34 44 10640274
Moffat CJC Pinder SE Dixon AR Elston CW Blamey RW Ellis IO Phyllodes tumours of the breast: a clinicopathological review of thirty-two cases Histopathology 1995 27 205 218 8522284
van Diest PJ Baak JP Matze-Cok P Wisse-Brekelmans EC van Galen CM Kurver PH Bellot SM Fijnheer J van Gorp LH Kwee WS Reproducibility of mitosis counting in 2469 breast cancer specimens: results from the Multicenter Morphometric Mammary Carcinoma Project Hum Pathol 1992 23 603 607 1592381 10.1016/0046-8177(92)90313-R
Kuenen-Boumeester V Henzen-Logmans SC Timmermans MM van Staveren IL van Geel A Peeterse HJ Bonnema J Berns EM Altered expression of p53 and its regulated proteins in phyllodes tumours of the breast J Pathol 1999 189 169 175 10547570 10.1002/(SICI)1096-9896(199910)189:2<169::AID-PATH408>3.0.CO;2-P
Bos R Zhong H Hanrahan CF Mommers ECM Semenza GL Pinedo HM Abeloff MD Simons JW van Diest PJ van der Wall E Levels of hypoxia-inducible factor-1α during breast carcinogenesis J Natl Cancer Inst 2001 93 309 314 11181778 10.1093/jnci/93.4.309
Fleege JC van Diest PJ Baak JP Systematic random sampling for selective interactive nuclear morphometry in breast cancer sections. Refinement and multiobserver evaluation Anal Quant Cytol Histol 1993 15 281 289 8397649
Zhong H De Marzo AM Laughner E Lim M Hilton DA Zagzag D Buechler P Isaacs WB Semenza GL Simons JW Overexpression of hypoxia-inducible factor 1α in common human cancers and their metastases Cancer Res 1999 59 5830 5835 10582706
Aebersold DM Burri P Beer KT Laissue J Djonov V Greiner RH Semenza GL Expression of hypoxia-inducible factor-1α: a novel predictive and prognostic parameter in the radiotherapy of oropharyngeal cancer Cancer Res 2001 61 2911 2916 11306467
Zagzag D Zhong H Scalzitti JM Laughner E Simons JW Semenza GL Expression of hypoxia-inducible factor 1α in brain tumors. Association with angiogenesis, invasion and progression Cancer 2000 88 2606 2618 10861440 10.1002/1097-0142(20000601)88:11<2606::AID-CNCR25>3.0.CO;2-W
Vleugel MM Greijer AE Shvarts A van der Groep P van Berkel M Aardbodem Y van Tinteren H Harris AL van Diest PJ van der Wall E Differential prognostic impact of hypoxia induced and diffuse HIF-1α expression in invasive breast cancer J Clin Pathol 2005 58 172 177 15677538 10.1136/jcp.2004.019885
Richard DE Berra E Pouyssegur J Nonhypoxic pathway mediates the induction of hypoxia-inducible factor 1α in vascular smooth muscle cells J Biol Chem 2000 275 26765 26771 10837481
Bartosova M Parkkila S Pohlodek K Karttunen TJ Galbavy S Mucha V Harris AL Pastorek J Pastorekova S Expression of carbonic anhydrase IX in breast is associated with malignant tissues and is related to overexpression of c-erbB2 J Pathol 2002 197 314 321 12115877 10.1002/path.1120
Belien JAM van Diest PJ Baak JPA Relationships between vascularization and proliferation in invasive breast cancer J Pathol 1999 189 309 318 10547591 10.1002/(SICI)1096-9896(199911)189:3<309::AID-PATH457>3.0.CO;2-0
Turner KJ Crew JP Wykoff CC Watson PH Poulsom R Pastorek J Ratcliffe PJ Cranston D Harris AL The hypoxia-inducible genes VEGF and CA9 are differentially regulated in superficial vs invasive bladder cander Br J Cancer 2002 86 1276 12782 11953885 10.1038/sj.bjc.6600215
Talks KL Turley H Gatter KC Maxwell PH Pugh CW Ratcliffe PJ Harris AL The expression and distribution of the hypoxia inducible factors HIF-1α and HIF-2α in normal human tissues, cancers and tumor-associated macrophages Am J Pathol 2000 157 411 421 10934146
Beasley NJP Leek R Alam M Turley H Cox GJ Gatter K Millard P Fuggle S Harris AL Hypoxia-inducible factors HIF-1α and HIF-2α in head and neck cancer: relationship to tumor biology and treatment outcome in surgically resected patients Cancer Res 2002 62 2493 2497 11980639
Shpitz B Bomstein Y Sternberg E Klein E Tiomkin V Kaufman A Groisman G Bernheim J Immunoreactivity of p53, Ki-67, and c-erbB-2 in phyllodes tumors of the breast in correlation with clinical and morphologic features J Surg Oncol 2002 79 86 92 11815995 10.1002/jso.10049
Laughner E Taghavi P Chiles K Mahon PC Semenza GL Her2 (neu) signaling increases the rate of hypoxia inducible factor 1α synthesis: novel mechanisms for HIF-1-mediated vascular endothelial growth factor expression Mol Cell Biol 2001 21 3995 4004 11359907 10.1128/MCB.21.12.3995-4004.2001
Kracht J Sapino A Bussolati G Malignant phyllodes tumor of breast with lung metastases mimicking the primary Am J Surg Pathol 1998 22 1284 1290 9777991 10.1097/00000478-199810000-00014
Sawyer EJ Hanby AM Rowan AJ Gillett CE Thomas RE Poulsom R Lakhani SR Ellis IO Ellis P Tomlinson IPM The Wnt pathway, epithelial-stromal interactions, and malignant progression in phyllodes tumours J Pathol 2002 196 437 444 11920740 10.1002/path.1067
Sawhney N Garrahan N Douglas-Jones AG Willliams ED Epithelial-stromal interactions in tumors: a morphologic study of fibroepithelial tumors of the breast Cancer 1992 70 2115 2120 1327488
Yamashita J Ogawa M Egami H Matsuo S Kiyohara H Inada K Yamashita S Fujita S Abundant expression of immunoreactive endothelin 1 in mammary phyllodes tumor: possible paracrine role of endothelin 1 in the growth of stromal cells in phyllodes tumor Cancer Res 1992 52 4046 4049 1319837
Hu J Discher DJ Bisphoric NH Webster KA Hypoxia regulates expression of the endothelin-1 gene through a proximal hypoxia-inducible factor-1 binding site on the antisense strand Biochem Biophys Res Commun 1998 245 894 899 9588211 10.1006/bbrc.1998.8543
Tse GMK Ma TKF Chan KF Law BKB Chen MH Chan ECK Mak MKF Increased microvessel density in malignant and borderline mammary phyllodes tumours Histopathology 2001 38 567 570 11422501 10.1046/j.1365-2559.2001.01150.x
Weind KL Maier CF Rutt BK Moussa M Invasive carcinomas and fibroadenomas of the breast: comparison of microvessel distibutions-implications for imaging modalities Radiology 1998 208 477 483 9680579
Deroanne CF Hajitou A Calberg-Bacq CM Nusgens BV Lapiere CM Angiogenesis by fibroblast growth factor 4 is mediated through an autocrine up-regulation of vascular endothelial growth factor expression Cancer Res 1997 57 5590 5597 9407972
Finkenzeller G Sparacio A Technau A Marme D Siemeister G Sp1 recognition sites in the proximal promotor of the human vascular endothelial growth factor gene are essential for platetelet-derived growth factor-induced gene expression Oncogene 1997 15 669 676 9264407 10.1038/sj.onc.1201219
Pertovaara L Kaipainen A Mustonen T Orpana A Ferrara N Saksela O Alitalo K Vascular endothelial growth factor is induced in response to transforming growth factor-beta in fibroblastic and epithelial cells J Biol Chem 1994 269 6271 6274 8119973
Pilichowska M Kimura N Fujiwara H Nagura H Immunohistochemical study of TFG-α, TGF-β, EGFR and IGF-1 expression in human breast carcinoma Mod Pathol 1997 10 969 975 9346175
Bernstein L Deapen D Koss RK The descriptive epidemiology of malignant cystosarcoma phyllodes tumors of the breast Cancer 1993 71 3020 3024 8387873
Yeo EJ Chun YS Cho YS Kim J Lee JC Kim MS Park JW YC-1: a potential anticancer drug targeting hypoxia-inducible factor 1 J Natl Cancer Inst 2003 95 516 525 12671019
|
16168127
|
PMC1242154
|
CC BY
|
2021-01-04 16:54:50
|
no
|
Breast Cancer Res. 2005 Aug 5; 7(5):R808-R818
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1296
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr13041616812510.1186/bcr1304Research ArticleOverexpression of platelet-derived growth factor receptor α in breast cancer is associated with tumour progression Carvalho Inês [email protected] Fernanda [email protected] Albino [email protected] Rui M [email protected] Fernando [email protected] IPATIMUP – Institute of Molecular Pathology and Immunology of Porto University, Porto, Portugal2 Life and Health Sciences Research Institute (ICVS), School of Health Sciences, University of Minho, Braga, Portugal3 Medical Faculty of Porto University, Porto, Portugal2005 1 8 2005 7 5 R788 R795 6 5 2005 20 6 2005 29 6 2005 6 7 2005 Copyright © 2005 Carvalho et al.; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Receptor tyrosine kinases have been extensively studied owing to their frequently abnormal activation in the development and progression of human cancers. Platelet-derived growth factor receptors (PDGFRs) are receptors with intrinsic tyrosine kinase activity that regulate several functions in normal cells and are widely expressed in a variety of malignancies. After the demonstration that gastrointestinal stromal tumours without c-Kit mutations harbour PDGFR-α-activating mutations and that PDGFR-α is also a therapeutic target for imatinib mesylate, the interest for this receptor has increased considerably. Because breast cancer is one of the most frequent neoplasias in women worldwide, and only one study has reported PDGFR-α expression in breast carcinomas, the aim of this work was to investigate the potential significance of PDGFR-α expression in invasive mammary carcinomas.
Methods
We used immunohistochemistry to detect PDGFR-α overexpression on a series of 181 formalin-fixed paraffin-embedded invasive ductal breast carcinomas and in two breast cancer cell lines: MCF-7 and HS578T. We associated its expression with known prognostic factors and we also performed polymerase chain reaction–single-stranded conformational polymorphism and direct sequencing to screen for PDGFR-α mutations.
Results
PDGFR-α expression was observed in 39.2% of the breast carcinomas and showed an association with lymph node metastasis (P = 0.0079), HER-2 expression (P = 0.0265) and Bcl2 expression (P = 0.0121). A correlation was also found with the expression of platelet-derived growth factor A (PDGF-A; P = 0.0194). The two cell lines tested did not express PDGFR-α. Screening for mutations revealed alterations in the PDGFR-α gene at the following locations: 2500A→G, 2529T→A and 2472C→T in exon 18 and 1701G→A in exon 12. We also found an intronic insertion IVS17-50insA at exon 18 in all sequenced cases. None of these genetic alterations was correlated with PDGFR-α expression. The cell lines did not reveal any alterations in the PDGFR-α gene sequence.
Conclusion
PDGFR-α is expressed in invasive breast carcinomas and is associated with biological aggressiveness. The genetic alterations described were not correlated with protein expression, but other mechanisms such as gene amplification or constitutive activation of a signalling pathway inducing this receptor could still sustain PDGFR-α as a potential therapeutic target.
==== Body
Introduction
Uncontrolled tumour cell proliferation due to abnormal activation of several growth factors and their receptors is relevant in the events underlying human cancer development, because the tyrosine kinases receptors form one of the most important classes of growth factor receptors implicated in that process. Platelet-derived growth factor receptors (PDGFRs) α and β are characterized by an intracellular tyrosine kinase domain whose activation depends on ligand binding. The platelet-derived growth factor (PDGF) family of growth factors consists of five different disulphide-linked dimers, PDGF-AA, -BB, -AB, -CC and -DD that act via the two receptors PDGFR-α and PDGFR-β. All PDGF isoforms except PDGF-DD induce PDGFR-α dimerization, although this receptor binds to PDGF-AA with higher affinity, whereas PDGF-BB and PDGF-DD activate PDGFR-β dimers. After receptor activation, several intracellular pathways are stimulated, leading to cell proliferation and several other crucial processes [1]. PDGFR signalling has important functions during embryogenesis, and its overexpression is associated with several pathological conditions such as fibrotic and vasculoproliferative diseases and cancer [2-5]. Recently, the finding that gastrointestinal stromal tumours (GISTs) lacking c-Kit mutations harbour intragenic activating mutations in PDGFR-α [6,7], and that ligand-independent constitutive activation can be blocked by means of a tyrosine kinase inhibitor (imatinib mesylate), has increased the interest for PDGFR-α as a target for therapy.
The gene encoding PDGFR-α is located at chromosome 4q11-12, which spans 23 exons and encodes a transmembrane protein composed of five immunoglobulin-like domains in the extracellular region, a transmembrane domain, an ATP binding site and a hydrophilic kinase insert domain in the intracellular portion [8].
Despite increased public awareness, screening programmes and early detection, breast cancer remains the second leading cause of cancer death in women. This leads to a constant search for new biological markers that could be used as prognostic/predictive factors and therapeutic targets, resulting in better disease-free survival and overall survival [9]. Given the success of imatinib mesylate therapy of chronic myeloid leukaemia and GIST, the molecular targets for this drug have been explored in distinct types of cancer.
To the best of our knowledge, there have been few studies on PDGFRs in mammary neoplasias and no reports on the presence of PDGFR-α mutations in breast carcinomas. The aims of the present study were the following: first, to evaluate the immunohistochemistry expression of PDGFR-α and PDGF-A in a series of invasive ductal breast carcinomas; second, to correlate the PDGFR-α expression with prognostic factors in breast cancer; and third, to screen for PDGFR-α gene-activating mutations in breast cancer.
Materials and methods
Tissue specimens
One hundred and eighty-one formalin-fixed paraffin-embedded cases of invasive ductal breast carcinomas were retrieved from the histopathology files of IPATIMUP and São João Hospital (Porto, Portugal). All cases were independently reviewed on haematoxylin/eosin-stained sections by two pathologists (FS and FM). All relevant data were available for analysis, including age, tumour size, histological grade, axillary lymph node status, oestrogen receptor status, p53, MIB-1 and HER-2 expression, angiogenic index and patient survival.
The mean age of the patients was 55 years old (range 24 to 83) and the size of the tumours ranged from 2.0 to 150.0 mm (mean 30.9).
Cell lines
MCF-7 and HS578T breast cancer cell lines (ATCC, Teddington, UK) were maintained in Eagle's minimum essential medium and Dulbecco's modified Eagle's medium, respectively, and were supplemented with 10% heat-inactivated fetal bovine serum, 100 U/ml penicillin and streptomycin (Gibco, Paisley, UK) in a humidified incubator at 37°C with a 5% CO2 atmosphere.
The medium was replaced every 2 to 3 days in all cell cultures.
Cell lines were grown until confluence; they were then scraped and the suspension was transferred into a tube for cell block preparation. After centrifugation at 1,800 r.p.m. for 5 min, the supernatants were carefully removed without dislodging the cell button. After the addition of 10 ml of 10% neutral buffered formalin to the intact cell button, the mixture was kept at room temperature (18 to 25°C) for 20 min. An additional centrifugation at 1,800 rpm for 5 min was followed by removal of the formalin supernatant and the addition of two or three drops of bovine albumin (22% from Ortho Diagnostics). After mixing, 10 ml of 95% ethanol was added to the sample, which was mixed again. After a final centrifugation step, the mixture rested for 15 min. The button was carefully loosened to allow it to be slipped intact out of the tube, and the solution was poured through lens paper to filter it. The cell button was wrapped in tissue paper and placed in a tissue cassette, which was kept in 10% neutral buffered formalin until further processing and paraffin embedding.
Immunohistochemistry
Automated immunohistochemistry (Lab Vision Autostainer LV-1; Lab Vision Corporation, Fremont, CA, USA) was performed with the streptavidin–biotin–peroxidase technique, using antibodies raised against human PDGFR-α (1:200 dilution; Lab Vision Corporation) and PDGF-A (clone N-30; 1:80 dilution; Santa Cruz Biotechnology, Santa Cruz, CA, USA). In brief, antigen retrieval was performed in 10 mM citrate buffer (pH 6.0) for 20 min with wet heat (hot bath) at 98°C for PDGFR-α or, for PDGF-A, samples were pretreated in 10 mM citrate buffer for 15 min (3 x 5 min) in a microwave at 600 W. After cooling to room temperature, the sections were rinsed with PBS, which was used for all subsequent washing steps. Endogenous peroxidase activity was blocked by the incubation of slides in 3% hydrogen peroxide in methanol for 10 min, and non-specific epitopes were eliminated by incubation with a blocking solution (UltraVision Block; Lab Vision Corporation) for 10 min. The slides were incubated with the primary antibodies for 30 min. After the slides had been rinsed, they were incubated with biotinylated secondary antibody followed by enzyme-labelled streptavidin for 10 min (UltraVision detection system, anti-polyvalent horseradish peroxidase/diaminobenzidine; Lab Vision Corporation). The immunoreaction was developed with diaminobenzidine (LabVision Corporation). Slides were counterstained with Gill's haematoxylin.
A positive control was included in each slide run. Sections of vessels in the corion (lamina propria) of gastrointestinal mucosa biopsies were used as positive controls, and blood vessels present in the periphery of the carcinomas studied were used as internal positive controls. The cases were considered positive whenever there was cytoplasmic staining for PDGFR-α and PDGF-A.
Statistical analysis
Invasive ductal breast carcinomas were subclassified into cases with or without PDGFR-α and PDGF-A expression. Contingency tables and the χ2 test were used in StatView 5.0 software (SAS Institute Inc., Cary, NC, USA) to estimate the correlation between PDGFR-α immunoreactivity and clinical-pathological and molecular markers previously studied for these cases. A correlation was considered significant whenever P < 0.05.
Mutation analysis
Screening for mutations by polymerase chain reaction–single-stranded conformational polymorphism (PCR-SSCP) followed by direct DNA sequencing was performed in only 13 cases from our series (because of the unavailability of biological material necessary for this study), and in the two cell lines.
DNA extraction
The tumour tissue was microdissected with a sterile scalpel under a stereomicroscope to avoid contamination with non-neoplastic tissues. DNA extraction was performed with the NucleoSpin kit (Macherey-Nagel, Düren, Germany) for blocks embedded in paraffin wax.
For the human cell lines MCF-7 and HS578T, DNA was extracted with the salting-out procedure, with some modifications. In brief, cell pellets were dissolved in SE buffer (1.2 M sorbitol, 0.1 M EDTA, pH 8.0) and digested overnight with proteinase K. After incubation, saturated NaCl and chloroform were added to allow the separation of proteins from DNA. DNA precipitation was performed with propan-2-ol. Finally, DNA was washed with 70% ethanol and dissolved in TE buffer (10 mM Tris/Cl, 1 mM EDTA, pH 8.0). The concentrations were determined by spectrophotometry and aliquoted DNA was stored at -20°C until use.
PCR-SSCP
Pre-screening of exons 12 and 18 of the PDGFR-α gene was performed with the primers with sequences shown in Table 1 and described elsewhere [6]. PCR was performed with 2 to 4 μl of DNA solution, PCR buffer (16 mM (NH4)2SO4, 67 mM Tris-HCl pH 8.8, 0.01% Tween-20; Bioron GmbH, Ludwigshafen, Germany), 1.5 to 2.5 mM of MgCl2 (Bioron GmbH), each dNTP (Fermentas, Ontario, Canada) at 0.2 mM, each primer at 0.2 μM, and 1 U/μl SuperHot Taq DNA polymerase (Bioron GmbH) in a final volume of 25 μl. Thirty-seven cycles of denaturation (95°C), annealing (58 to 60°C) and extension (72°C), for 45 s each, were performed in a gradient thermocycler (Bio-Rad, Hercules, CA, USA).
A total of 20 μl of the PCR products was mixed with an equivalent volume of denaturing loading buffer (98% formamide, 0.05% xylenecyanol and bromophenol blue). The samples were denatured at 95°C for 10 min, put into ice and run at 4°C in a 0.8 × MDE gel for exon 12 and in a 0.8 × MDE gel with 3% glycerol for exon 18 at 200 to 280 V, for 16 to 20 hours.
Gels were fixed in 10% ethanol for 10 min and oxidized in 1% nitric acid solution. After 3 min, gels were stained in the dark with a 0.012 M silver nitrate for 20 min. The gels showed an appropriate colour in 0.28 M anydrous sodium carbonate and 0.019% formalin. The reaction was stopped by incubation in 10% acetic acid for 2 min. Subsequently, gels were dried at 80°C for 2 hours (Thermo Savant SGD5040 Slab Gel Dryer).
Direct sequencing
The abnormal PCR products for each exon screened by SSCP were purified (MicrospinTM S-400 HR columns; Amersham Biosciences, Little Chalfont, Bucks., UK) and analysed by direct sequencing (Sequence Laboratories Göttingen GmbH, Göttingen, Germany). For cell lines, after being purified, the PCR products were subjected to a cycle sequencing reaction with an ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Foster City, CA, USA), and then analysed with an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems) in accordance with the manufacturer's instructions.
Results
Immunohistochemistry for PDGFR-α was performed in 181 invasive ductal breast carcinomas and in the two breast cancer cell lines. PDGF-A immunostaining was performed in only 48 cases because of the unavailability of biological material. PDGFR-α cytoplasmic expression (Fig. 1) was found in 71 of 181 cases (39.2%), and PDGF-A expression (Fig. 2) was found in 27 of 48 cases (56.25%) (Table 2). Co-expression of PDGFR-α and PDGF-A was found in 10 cases and a significant correlation was observed between the expression of the receptor and the ligand (P = 0.0194). A positive correlation was also found between PDGFR-α expression and axillary lymph node status (P = 0.0079), Bcl2 expression (P = 0.0121) and HER-2 expression (P = 0.0265) (Table 3). Both cell lines showed an absence of PDGFR-α expression, but MCF-7 showed a low expression of PDGF-A, whereas HS578T did not express it at all.
Five of the six cases sequenced for exon 12 did not show any nucleotide sequence alteration and were considered normal. PCR-SSCP followed by direct sequencing of exons 12 and 18 of PDGFR-α revealed some alterations (Fig. 3): 2500A→G and 2529T→A in exon 18 (in the same case) and a further two in distinct cases: 2472C→T (exon 18) and 1701G→A (exon 12). The intronic insertion IVS17-50insA in exon 18 was found in all cases sequenced, although four cases were heterozygous and the other two homozygous for the inserted A (Table 4). The two cell lines studied also presented the homozygous intronic insertion IVS17-50insA, but no other DNA sequence alterations were found among them.
The expression of PDGFR-α varied between the cases with the alterations described above and did not show any correlation with the nucleotide alterations found.
Discussion
PDGFR is a growth factor receptor with intrinsic tyrosine kinase activity and is deregulated in several human diseases. Breast carcinomas are known to express PDGF; however, there have been few studies on PDGF receptors in breast neoplasias, most of them related to the β subunit [10].
In our study, 71 (39.2%) of the 181 invasive carcinomas analysed expressed PDGFR-α. Slightly different results were obtained by Leu and colleagues [11]: using a high-density tissue microarray (TMA), they analysed PDGFRs expression in several human malignancies and found PDGFR-α expression in 98% of the 49 breast carcinomas studied. This high positivity could be justified by the use of a different antibody or by a distinct evaluation, although there is no description of the methodology used in that study. In the same study, Leu and colleagues [11] demonstrated PDGFR-α overexpression in other types of carcinoma, such as ovary, prostate, colon and lung. A high level of expression was also observed in ovarian carcinomas (81%) in comparison with the study by Matei and colleagues [12], which obtained 68.3% of PDGFR-α expression. These data suggest that the study of Leu and colleagues [11] overestimated values of PDGFR-α expression in different tumour types.
Because the deregulation of PDGFR signalling can lead to an autocrine or a paracrine stimulation of the tumour cells, we also evaluated the expression of PDGF-A, which is the major ligand of this receptor.
Twenty-seven (56.25%) of 48 carcinomas analysed showed expression of PDGF-A, and we found a statistically significant correlation with PDGFR-α, suggesting a mechanism of autocrine stimulation. This autocrine expression might have a causal role in the development of a variety of human cancers and, for example, seems to be involved in the development of high-grade sarcomas and gliomas [13,14]. De Jong and colleagues [15] reported co-expression of PDGF-A/PDGFR-α in epithelium, stroma and endothelium of invasive breast carcinomas and obtained indications about possible autocrine and paracrine mechanisms in the stroma, where they might be responsible for a baseline stromal proliferation, and in the endothelium, where they promote a basic level of angiogenesis. The simple presence of ligand/receptor combinations does not necessarily indicate this type of mechanism, especially when the ligand and the receptor are produced in spatial sites distant from each other [15]. In this context, an elegant way to demonstrate the PDGFR-α activation is to assay phosphorylated PDGFR-α (p-PDGFR-α) by immunohistochemistry. However, until now p-PDGFR-α antibodies have been reliable only in frozen samples that were not available in our series.
To investigate the role of PDGFR-α in neoplastic proliferation and progression, we correlated PDGFR-α immunoexpression with prognostic factors and molecular markers previously studied in this series. We did not obtain significant differences related to the classic prognostic factors, such as tumour size, histological grade and oestrogen receptor status. However, we found an association between PDGFR-α expression and positive axillary lymph node status, suggesting that PDGFR-α-positive tumours have a more aggressive phenotype.
We also obtained a correlation between PDGFR-α expression and Bcl2 expression. Bcl2 is an anti-apoptotic protein overexpressed in about 60 to 80% of breast cancers [16], and several studies suggest that the low apoptotic response caused by that overexpression allows the accumulation of genetic alterations that might be important in breast cancer metastatic potential [17,18]. In our study, 66.7% of the carcinomas expressing PDGFR-α present Bcl2 co-expression, and we speculate that PDGFR-α might be activating anti-apoptotic routes such as the Bcl2 pathway.
The association found between PDGFR-α expression and HER-2 expression is quite interesting, and probably proves the previous results about the correlation of PDGFR-α pathway with aggressiveness. HER-2 amplification/overexpression occurs in 30% of human breast cancers and is associated with biological aggressiveness and shortened disease-free survival and overall patient survival [19,20]. Our results seem to indicate that there is a relationship between these two receptors: most carcinomas that expressed PDGFR-α also expressed HER-2, and the absence of PDGFR-α was also found at a higher frequency in carcinomas without HER-2 expression. PDGFR-α and HER-2 are both tyrosine kinase receptors and although they belong to different subfamilies they can give rise to similar cellular/biological effects [21]. The coexistence of these receptors might contribute to neoplastic proliferation but might also influence tumour cell survival.
These results demonstrate that PDGFR-α expression is correlated with certain aggressiveness parameters of invasive breast carcinomas, and corroborate other results [22,23] showing PDGFR-α overexpression associated with aggressive characteristics in ovarian and renal cell carcinomas, respectively.
As far as we know there have been no studies on PDGFR-α expression in breast cancer cell lines; here we report that none of the cell lines studied express PDGFR-α. Breast cancer cell lines are known to secrete PDGFs, and, as Bronzert and colleagues [24] reported, MCF-7 shows some expression of PDGF-A, as we demonstrated in our work. No information about PDGF-A status in the HS578T cell line is available and in our study it did not show any expression. Ligand production by breast cancer cell lines might therefore have a role in mediating paracrine stimulation of tumour growth, affecting other cells in the microenvironment [15].
PDGFR-α overexpression can occur by genetic amplification or through activating mutations. PDGFR-α amplification has already been investigated in some tumour types, such as oesophageal squamous cell carcinoma [25], pulmonary artery intimal sarcomas (where it was demonstrated that PDGFR-α amplification is strongly associated with the development of this type of neoplasia) [26] and glioblastoma, occurring in 8 to 16% of cases [14]. However, PDGFR-α amplification in breast carcinomas was detected only by Daigo and colleagues [27], with controversial results: by array-comparative genomic hybridization, 21% of the cases presented gene amplification, and by metaphase-comparative genomic hybridization no amplification was detected.
As regards PDGFR-α genetic alterations, most studies have been performed in GIST, demonstrating that activating mutations in exons 12 and 18 has a major role in the development of these type of tumours [6,7]. It has also been shown that this receptor can be a therapeutic target for a recently developed drug, imatinib mesylate.
Because PDGFR-α amplification has shown controversial results and does not seem to be correlated with a response to imatinib therapy, we decided to search for the activating mutations frequently observed in GISTs in our sample of breast carcinomas, as well as in the two breast cancer cell lines MCF-7 and HS578T.
We found some alterations in PDGFR-α gene sequence: 2500A→G, 2529T→A and 2472C→T in the tyrosine kinase II domain, all in exon 18, and 1701G→A in the transmembrane domain of exon 12. However, none of these alterations corresponded to those described in GISTs [5-7] or even to the unique mutation found in breast phyllodes tumours [28]. We still observed an intronic insertion IVS17-50insA, in all cases sequenced for exon 18, including the two cell lines. This insertion has already been described [28] in a breast phyllodes tumour and is observed in the general population, suggesting that it might be a polymorphism. In other histological tumour types, PDGFR-α-activating mutations are unknown [29].
Although the DNA sequence alterations found do not have any effect on protein structure and/or function, it is necessary to emphasize that the alterations with amino acid change that were found in the tyrosine kinase domain could be important, because they can affect autophosphorylation (activation) and consequent cellular effect (signalling), allowing constitutive activation of the receptor. The cases with gene sequence alterations do not present clinical or anatomical-pathological features that distinguish them from the other cases, so the possible consequences of the alterations described should be evaluated by functional studies. Further studies are also necessary to define the role of PDGFR-α in breast oncogenesis, as well as the inherent molecular mechanisms of this process. It is important to stress that PDGFR-α and other protein tyrosine kinases can be activated by mechanisms different from gene mutations, such as gene fusion and amplification, autocrine and paracrine receptor stimulation by its ligand, loss of phosphatase activity, cross-activation by other kinases and promoter activation/inactivation via methylation/demethylation.
Conclusion
Our results demonstrated the presence of PDGFR-α expression in 39.2% of invasive ductal carcinomas and that this expression was correlated with aggressiveness parameters, such as the presence of regional lymph node metastasis, HER-2 expression and Bcl2 expression, and also showed an association with PDGF-A ligand expression. Although the PDGFR-α mutations detected in this study were not correlated with protein expression, other mechanisms, such as gene amplification or constitutive activation of a signalling pathway, could explain the overexpression observed in our study and still can sustain PDGFR-α as a potential therapeutic target in breast cancer.
Abbreviations
CGH = comparative genomic hybridization; GIST = gastrointestinal stromal tumour; PCR-SSCP = polymerase chain reaction-single-stranded conformational polymorphism; PDGF = platelet-derived growth factor; PDGFR-α and PDGFR-β = platelet-derived growth factor receptor α and β.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
IC performed immunohistochemistry, PCR-SSCP and statistical analysis, and analysed the data and drafted the article. FM contributed significantly to the analysis and interpretation of the immunohistochemistry data. AM was the responsible for the cell culture results. RR contributed to the mutation analysis. FS coordinated the study and took a role in the supervision and final approval of the article. All authors read and approved the final manuscript.
Acknowledgements
We thank Novartis Oncology (Portugal) for the financial support that made this study possible.
Figures and Tables
Figure 1 PDGFR-α expression in invasive breast carcinomas by immunohistochemistry (streptavidin-biotin-peroxidase). (a) Platelet-derived growth factor receptor α (PDGFR-α) expression in pericytes and smooth muscle cells of a blood vessel: internal control (original magnification × 200); (b) Absence of PDGFR-α expression in neoplastic cells (original magnification × 200); (c) PDGFR-α diffuse cytoplasmic expression in neoplastic cells (original magnification × 200; inset × 400); (d) Neoplastic cells showing strong and diffuse cytoplasmic PDGFR-α expression (original magnification × 400).
Figure 2 PDGF-A expression pattern studied by immunohistochemistry in invasive breast carcinomas. (a) Carcinoma without platelet-derived growth factor A (PDGF-A) expression (original magnification × 400); (b) Carcinoma with strong PDGF-A expression (original magnification × 200; inset × 400).
Figure 3 Platelet-derived growth factor receptor α (PDGFR-α) sequencing results for exon 18 (forward strand). (a) 2500A→G (1834V) DNA sequence alteration; (b) IVS17-50insA intronic insertion.
Table 1 Primer sequence, annealing temperatures and fragment length, for PDGFR-α exons
Exon Primer sequence (5'→3') Annealing temperature (°C) Fragment length (base pairs)
12 F, TCCAGTCACTGTGCTGCTTC; R, GCAAGGGAAAAGGGAGTCTT 58 260
18 F, ACCATGGATCAGCCAGTCTT; R, TGAAGGAGGATGAGCCTGACC 60 251
F, forward primer; PDGFR-α, platelet-derived growth factor receptor α; R, reverse primer.
Table 2 PDGFR-α and PDGF-A expression in invasive ductal breast carcinomas
Expression PDGFR-α, n (%) PDGF-A, n (%)
Negative 110 (60.8) 21 (43.75)
Positive 71 (39.2) 27 (56.25)
Total 181 (100) 48 (100)
PDGFR-α, platelet-derived growth factor receptor α; PDGF-A, platelet-derived growth factor A.
Table 3 Correlation between PDGFR-α expression, PDGF-A expression and clinical–pathological parameters in invasive ductal carcinomas
Parameter PDGFR-α negative PDGFR-α positive P
PDGF-A
Negative 19 (52.7%) 2 (16.7%)
Positive 17 (47.3%) 10 (83.3%) 0.0290
Age (n) 56.315 ± 11.913 (108) 52.986 ± 12.811 (70) NS (0.0788)
Tumour size, mm (n) 31.406 ± 22.056 (106) 30.101 ± 22.655 (69) NS (0.7057)
Histological grade
I 20 (19.4%) 9 (12.8%)
II 38 (36.9%) 34 (48.6%)
III 45 (43.7%) 27 (38.6%) NS (0.2595)
Axillary lymph node status
Negative 47 (48%) 18 (27.3%)
Positive 51 (52%) 48 (72.7%) 0.0079
Oestrogen receptor status
Negative 30 (27.8%) 19 (26.7%)
Positive 78 (72.2%) 52 (73.3%) NS (0.8813)
p53
Negative 38 (54.3%) 21 (53.8%)
Positive 32 (45.7%) 18 (46.2%) NS (0.9648)
MIB-1 (n) 19.449 ± 17.484 (49) 12.333 ± 12.584 (15) NS (0.1491)
HER-2
Negative (0/1+) 54 (57.4%) 26 (40%)
2+ 17 (18.1%) 10 (15.4%)
3+ 23 (24.5%) 29 (44.6%) 0.0265
Bcl2
Negative 29 (59.2%) 15 (33.3%)
Positive 20 (40.8%) 30 (66.7%) 0.0121
Angiogenesis index (n) 43.393 ± 26.498 (55) 43.633 ± 27.635 (18) NS (0.9737)
Patient survival, months (n) 24.917 ± 19.500 (48) 21.067 ± 20.405 (15) NS (0.5115)
NS, not statistically significant; PDGFR-α, platelet-derived growth factor receptor α;
PDGF-A, platelet-derived growth factor A.
Table 4 Results of PDGFR-α gene mutations screening in invasive breast carcinomas
PDGFR-α gene sequence alterations Exon 12 Exon 18
Normal 12/13 7/13
1701G→A (P567P) 1/13 -
2500A→G (I834V) - 1/13
2529T→A (I843N) - 1/13
2472C→T (V824V) - 1/13
IVS17-50insA - 6/13
PDGFR-α, platelet-derived growth factor receptor α.
==== Refs
Tallquist M Kazlaushas A PDGF signaling in cells and mice Cytokine Growth Factor Rev 2004 15 205 213 15207812 10.1016/j.cytogfr.2004.03.003
Betsholtz C Karlsson L Lindhal P Developmental roles of platelet-derived growth factors Bioessays 2001 23 494 507 11385629 10.1002/bies.1069
Bonner JC Regulation of PDGF and its receptors in fibrotic disease Cytokine Growth Factor Rev 2004 15 255 273 15207816 10.1016/j.cytogfr.2004.03.006
Raines EW PDGF and cardiovascular disease Cytokine Growth Factor Rev 2004 15 237 254 15207815 10.1016/j.cytogfr.2004.03.004
Östman A PDGF receptors-mediators of autocrine tumor growth and regulators of tumor vasculature and stroma Cytokine Growth Factor Rev 2004 15 275 286 15207817 10.1016/j.cytogfr.2004.03.002
Heinrich MC Corless CL Duensing A McGreevey L Chen C Joseph N Singer S Griffith J Haley A Town A PDGFRA activating mutations in gastrointestinal stromal tumors Science 2003 299 708 710 12522257 10.1126/science.1079666
Hirota S Ohashi A Nishida T Isozaki K Kinoshita K Shinomura Y Kitamura Y Gain-of-function mutations of platelet.drived growth factor receptor α gene in gastrointestinal Stromal Tumors Gastroenterology 2003 125 660 667 12949711 10.1016/S0016-5085(03)01046-1
Kawagishi J Kumabe T Yoshimoto T Yamamoto T Structure, organization and transcription units of the human alpha-platelet-derived growth factor receptor gene, PDGFRA Genomics 1995 30 224 232 8586421 10.1006/geno.1995.9883
Weidner N Weidner N, Corte RJ, Suster S, Weiss LM Breast cancer Modern Surgical Pathology 2003 2 1 Philadelphia: Saunders 539 628
Coltrera MD Wang J Porter PL Gown AM Expression of platelet-derived growth factor B-chain and the platelet-derived growth factor receptor beta subunit in human breast tissue and breast carcinoma Cancer Res 1995 55 2703 2708 7780988
2002 ASCO Annual Meeting
Matei D Sanchez K Kelley M Baldridge LA Williams S Chang D Jeng MH PDGFRA is overexpressed in ovarian carcinomas and represents a potential therapeutic target [abstract] Proc Am Soc Clin Oncol 2003 22 867
Dibb NJ Dilworth SM Mol CD Switching on kinases: oncogenic activation of the BRAF and PDGFR family Nat Rev Cancer 2004 4 718 727 15343278 10.1038/nrc1434
Hunter SB Brat DJ Olson JJ Von Deimling A Zhou W Van Meir EG Alterations in molecular pathways of diffusely infiltrating glial neoplasms: application to tumor classification and anti-tumor therapy [Review] Int J Oncol 2003 23 857 869 12963964
De Jong JS Van Diest PJ Der Valk PV Baak JPA Expression of growth factors, growth inhibiting factors, and their receptors in invasive breast cancer. I: an inventory in search of autocrine and paracrine loops J Pathol 1998 184 44 52 9582526 10.1002/(SICI)1096-9896(199801)184:1<44::AID-PATH984>3.0.CO;2-H
Wang S Yang D Lippman ME Targeting Bcl-2 and Bcl-XL with nonpeptidic small-molecule antagonists Semin Oncol 2003 30 Suppl 16 132 142 12720122
Fernandez Y Gu B Martinez A Torregrosa A Sierra A Inhibition of apoptosis in human breast cancer cells: role in tumor progression to the metastatic state Int J Cancer 2002 101 317 326 12209955 10.1002/ijc.10628
Sierra A Castellsague X Escobedo A Lloveras B Garcia-Ramirez M Moreno A Fabra A Bcl2 with loss of apoptosis allows accumulation of genetic alterations: a pathway to metastatic progression in human breast cancer Int J Cancer 2000 89 142 147 10754491 10.1002/(SICI)1097-0215(20000320)89:2<142::AID-IJC7>3.3.CO;2-2
Car JA Havstad S Zarbo RJ Divine J Mackowiak P Velanovich V The association of HER-2/neu amplification with breast cancer recurrence Arch Surg 2000 135 1469 1474 11115354 10.1001/archsurg.135.12.1469
Gschwind A Fischer OM Ullrich A The discovery of receptor tyrosine kinases: targets for cancer therapy Nat Rev Cancer 2004 4 361 370 15122207 10.1038/nrc1360
Zhou BP Hung MC Dysregulation of cellular signaling by Her2/neu in breast cancer Semin Oncol 2003 30 Suppl 16 38 48 14613025
Lassus H Sihto H Leminen A Nordling S Joensuu H Nupponen N Butzow R Genetic alterations and protein expression of Kit and PDGFRA in serous ovarian carcinoma [abstract] Br J Cancer 2004 91 2048 2055 15583695 10.1038/sj.bjc.6602252
Sulzbacher I Birner P Traxler M Marberger M Haitel A Expression of platelet-derived growth factor-alpha alpha receptor is associated with tumor progression in clear renal cell carcinoma Am J Clin Pathol 2003 120 107 112 12866380 10.1309/LQ9E-MK8Q-KE75-NGGX
Bronzert DA Pantazis P Antoniades HN Kasid A Davidson N Dickson RB Lippman ME Synthesis and secretion of platelet-derived growth factor by human breast cancer cell lines Proc Natl Acad Sci USA 1987 84 5763 5767 3039506
Arai H Ueno T Tangoku A Yoshino S Abe T Kawauchi S Oga A Furuya T Oka M Sasaki K Detection of amplified oncogenes by genome DNA microarray in human primary esophageal squamous cell carcinoma: comparison with conventional comparative genomic hybridization analysis Cancer Genet Cytogenet 2003 146 16 21 14499691 10.1016/S0165-4608(03)00106-7
Zhao J Roth J Bode-Lesniewska B Pfalt M Heitz PU Komminoth P Combined comparative genomic hybridization and microarray for detection of genes amplifications in pulmonary artery intimal sarcomas and adrenocortical tumors Genes Chromosomes Cancer 2002 34 48 57 11921282 10.1002/gcc.10035
Daigo Y Chin S-F Gorringe KL Bobrow LG Ponder BAJ Pharoah PDP Caldas C Degenerate oligonucleotide primed-polymerase chain reaction-based array comparative genomic hybridization for extensive amplicon profiling of breast cancers – a new approach for the molecular analysis of paraffin-embedded cancer tissue Am J Pathol 2001 158 1623 1631 11337360
Carvalho S Silva AO Milanezi F Ricardo S Leitão D Amendoeira I Schmitt F c-KIT and PDGFRA in breast phyllodes tumours: overexpression without mutations? J Clin Pathol 2004 57 1075 1079 15452163 10.1136/jcp.2004.016378
Sihto H Sarlomo-Rikala M Tynnnen O Tanner M Andersson LC Franssila K Nupponen NN Joensuu H KIT and platelet-derived growth factor receptor alpha tyrosine kinase gene mutations and KIT amplifications in human solid tumors J Clin Oncol 2005 23 49 57 15545668 10.1200/JCO.2005.02.093
|
16168125
|
PMC1242156
|
CC BY
|
2021-01-04 16:54:51
|
no
|
Breast Cancer Res. 2005 Aug 1; 7(5):R788-R795
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1304
|
oa_comm
|
==== Front
Breast Cancer ResBreast Cancer Research1465-54111465-542XBioMed Central London bcr13051616812410.1186/bcr1305Research ArticleHigh expression of Lewisy/b antigens is associated with decreased survival in lymph node negative breast carcinomas Madjd Zahra [email protected] Tina [email protected] Nicholas FS [email protected] Ian [email protected] Ian [email protected] Lindy G [email protected] Academic Department of Clinical Oncology, Institute of Infections Immunity and Inflammation, University of Nottingham, City Hospital, Nottingham, UK2 Scancell Ltd, BioCity, Nottingham, UK3 Department of Histopathology, City Hospital, Nottingham, UK2005 28 7 2005 7 5 R780 R787 16 4 2005 17 5 2005 25 6 2005 12 7 2005 Copyright © 2005 Madjd et al.; licensee BioMed Central LtdThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
There is sufficient evidence that blood group related Lewis antigens are tumour-associated molecules. The Lewisy and Lewisb antigens are complex carbohydrates that are over-expressed by breast, lung, colon and ovarian cancers. The SC101 mAb is a unique Lewisy/b binding antibody that binds to native and extended Lewisy and Lewisb haptens, displaying no cross reactivity with H type 1, H type 2, Lewisx or normal blood group antigens.
Methods
Immunohistochemical detection of Lewisy/b was performed on 660 formalin-fixed, paraffin embedded breast tumour specimens using a streptavidin-biotin peroxidase technique. Tissue from these patients had previously been included in tissue microarrays. This cohort comprises a well characterized series of patients with primary operable breast cancer diagnosed between 1987 and 1992, obtained from the Nottingham Tenovus Primary Breast Carcinoma Series. This includes patients 70 years of age or less, with a mean follow up of 7 years.
Results
Of the breast carcinomas, 370 of 660 (56%) were negative for Lewisy/b expression, 110 (17%) cases showed a low level of expression (<25% of positive cells) and only 54 cases (8%) showed extensive expression of Lewisy/b (>75% of positive cells). We found significant positive associations between histological grade (p < 0.001), Nottingham Prognostic Index (p = 0.016), tumour type (p = 0.007) and the level of Lewis y/b expression. There was a significant correlation between the proportion of Lewisy/b positive tumour cells and survival in lymph-node negative patients (p = 0.006).
Conclusion
The unique epitope recognised by SC101 mAb on Lewisy/b hapten is over-expressed on breast tumour tissue compared with normal breast. In this large series of invasive breast cancers, higher expression of Lewisy/b was more often found in high grade and poor prognosis tumours compared to good prognosis cancers. Moreover, in lymph node negative breast carcinomas, over-expression of Lewisy/b hapten was associated with significantly decreased patient survival.
==== Body
Introduction
Blood group related antigens are frequently altered in association with neoplastic transformation [1-3]. Lewis blood group antigens Lewisa and Lewisb, and their positional isomers Lewisx and Lewisy, are carbohydrate antigens. Lewisy antigen is a difucosylated oligosaccharide with the chemical structure Fucα1→2Galβ1→4[Fucα1→3]GlcNAcβ1→R, which belongs to the A, B, H, Lewis blood group family. The Lewisy antigen is expressed predominately during embryogenesis, and in adults expression is restricted to granulocytes and epithelial surfaces [4]. Over-expression of Lewisy has been shown in the majority of cancer cells derived from epithelial tissues, however, including breast, ovary, pancreas, prostate, colon and non-small cell lung cancers [5], either at the plasma membrane as a glycolipid or linked to surface receptors (e.g. of the E-rb-B family) [6].
The SC101/29 mAb is a unique Lewisy/b binding antibody that recognises both Lewisy and Lewisb haptens [7]. It shows no cross reactivity with H type 1 or H type 2, Lewisx or normal blood group antigens. This antibody binds strongly to a wide range of tumour tissues, including colon, gastric and ovarian and it may also show strong reactivity with a range of normal tissue expressing either Lewisy or Lewisb.
To our knowledge, there has been no previous study examining whether there is any prognostic significance of Lewisy/b expression in breast carcinoma. It was, therefore, of interest to determine the expression of Lewisy/b in a large series of breast cancer tissue arrays using the SC101 mAb. Expression of Lewisy/b was then associated with clinicopathological parameters and patient outcomes.
Materials and methods
Patients and tumour characteristics
The study group consisted of 660 primary operable invasive breast carcinomas from patients aged 27 to 70 years (median 54 years) diagnosed from 1987 to 1992 and obtained from the Nottingham Tenovus Primary Breast Carcinoma Series. Patient characteristics, including age and menopausal status, along with information on local, regional and distant recurrence and survival were retrieved from a prospectively maintained database. Patients and tumour characteristics are shown in Table 1. Patients were followed up at 3 month intervals initially, then every 6 to 12 months for a median period of 83 months, with a mean survival of 77 months (1 to 151 months). Ethical approval was granted by the Nottingham Research Ethics Committee.
This is a well characterized series of primary operable breast cancers treated in a uniform manner, and has been used to study a wide range of potential prognostic factors and markers. Tumour characteristics, including histological grade [8], tumour type [9], vascular invasion [10], menopausal status [11], tumour size, lymph node stage and Nottingham Prognostic Index [12] are routinely assessed and recorded in a prospectively maintained database. The Nottingham Prognostic Index (NPI) score was calculated for each patient based on the following equation:
NPI = 0.2 × tumour size (cm) + grade (1–3) + lymph node stage (1–3)
This index predicts survival of patients with invasive breast cancer and is used clinically to define three groups with either a good (NPI =≤ 3.4), moderate (3.41 < NPI ≤ 5.4) or poor (NPI > 5.4) prognosis according to the score obtained. Patient management was based on tumour characteristics by NPI and hormone receptor status. Patients with an NPI score ≤ 3.4 received no adjuvant therapy, those with a NPI score >3.4 received tamoxifen if estrogen receptor (ER) positive (± Zoledex if pre-menopausal) or classical cyclophosphamide, methotrexate and 5-fluorouracil (CMF) if ER negative and fit enough to tolerate chemotherapy.
Construction of the tissue microarray blocks
At the time of resection, all tumours were managed in a standardised fashion, with immediate incision and fixation in neutral buffered formalin to minimize any diffusion problems, followed by processing through to embedding in paraffin wax. Breast cancer tissue microarrays were prepared as described previously [13]. Microarray samples with a diameter of 0.6 mm were punched from selected regions of each 'donor' block using a manual Tissue Arrayer (Beecher Instruments, Sun Prairie, WI, USA), and precisely arrayed into a new recipient paraffin block. These tissue microarray blocks were constructed in triplicate, at a density of 100 cores per block.
Immunohistochemistry
The SC101 mAb binds to native and extended Lewisy and Lewisb haptens [7]. Tissue microarray sections 4 μm thick were cut and immunohistochemical staining was performed using a strepavidin-biotin complex (ABC) method as described previously [14]. Microwave pre-treatment was performed in citrate buffer (pH 6.0) for 10 minutes at high power followed by 10 minutes at low power to retrieve antigenicity. The primary antibody SC101 was incubated on the slides for 1 h, at an optimal dilution found to be 1:1000 (stock concentration 1.1 mg/ml). Negative controls, consisting of normal swine serum instead of primary antibody, confirmed the specificity of the staining. Sections from a colon carcinoma previously determined to express Lewisy/b were used as positive control.
Evaluation of immunostaining
All immunostained tissue arrays were evaluated using a semi-quantitative system (by ZM) after a series was examined on a double-headed microscope blinded to patients' outcome and other clinical and pathological parameters. The obtained results were confirmed by two observers (ZM and IOE) using a multi-headed microscope and, in difficult cases, a consensus was achieved. Two features were assessed, the intensity of staining and the percentage of cells stained. The intensity of the immunostaining was classified into four categories: 0, no immunostaining present (as shown in Fig. 1); 1, weak immunostaining (Fig. 2); 2, moderate immunostaining (Fig. 3); and 3, strong immunostaining (Fig. 4). The percentage of positive cells at each intensity was also assessed semi-quantitatively, then classified into four groups as 1 (<25% positive cells), 2 (25% to 50% positive cells), 3 (51% to 75% positive cells) or 4 (>75% positive cells). In addition, the histochemical score (H score) of immunoreactivity was obtained by multiplying the intensity and percentage scores [15]. The histochemical scores were subgrouped into three groups of equal range for analysis, and a score of <100 was considered weak, 100 to 200 as moderate and 201 to 300 as strong.
Statistical analysis
Statistical analysis of data was performed using the SPSS package (version 11 for Windows; SPSS, Chicago, IL, USA). The significance of associations was determined by means of the Pearson R test and/or Pearson chi-square tests. To give sizes of effect and to look at the independence of effects, the percentage of Lewisy/b positive cells reclassified as a binary outcome (high, >25% positive cells; low, <25% positive cells) and effects of clinicopathological parameters were assessed using multiple logistic regression to give adjusted odds ratios and 95% confidence intervals.
In survival analysis, Kaplan-Meier curves were derived and the statistical significance of differences in survival between groups with different Lewisy/b expression was determined using the log-rank test. Survival was censored if the patients were still alive at the time of data analysis, or at the time of death for patients who died from an unrelated cause. Multivariate Cox regression analysis was used to evaluate the independent prognostic effect of variables on overall survival. P-values of <0.05 were identified as statistically significant.
Results
Level of expression of Lewisy/b on breast carcinomas
Of the breast carcinomas, 370 out of 660 (56%) were entirely negative for Lewisy/b expression (Fig. 1). A variable percentage of positively staining tumour cells was observed among the remaining 290 breast carcinomas; 110 (17%) cases showed a low level of expression (<25% of positive cells) and only 54 cases (8%) showed extensive expression of Lewisy/b (>75% of positive cells) on the cell membrane (Table 2). The staining observed was predominantly localised to the cell membrane and cytoplasm. No nuclear staining was observed.
Similarly, with regard to the intensity of staining, 17% (106) showed weak expression of Lewisy/b, whereas 89 (13%) and 95 (14%) of cases showed moderate and strong staining, respectively (Figs 2, 3, 4; Table 3).
Association of Lewisy/b expression with clinicopathological characteristics
A frequency histogram for the percentage of positive cells stained demonstrated that the mean value was 25%. This was therefore chosen as an appropriate cut-off for subsequent analysis (data not shown). Tumours were therefore divided into those with <25% versus those with >25% positive cells stained based on the distribution of staining results.
A significant positive relationship was found between the percentage of Lewisy/b positive tumour cells and histological grade of invasive tumours (p < 0.001; Table 4); a higher percentage of Lewisy/b positive tumour cells were identified more frequently in histological grade 3 tumours compared to grade 1 lesions. The odds ratio for higher percentage of Lewisy/b in those with a poor histological grade compared with those with a well differentiated tumour was 2.509 (95% confidence interval, 1.51 to 4.16) (Table 5).
A higher percentage of positive tumour cells was also identified more frequently in tumours from patients in the poor prognosis NPI group (NPI > 5.4) compared with those in the good prognosis group (NPI ≤ 3.4, p < 0.016), with an odds ratio of 1.99 (1.13 to 3.5) (Table 5). In addition, a higher proportion of Lewis y/b positive tumour cells was seen more frequently in poor prognosis tumour types (ductal/ NST (No special type)), solid lobular, lobular mixed or mixed NST and lobular cancers) compared to the excellent prognosis tumour types (tubulo-lobular, tubular, mucinous and invasive cribriform) (p < 0.007), with an odds ratio of 3.12 (1.07 to 9.11) (Table 5).
There was a positive association between the observed intensity of expression of Lewisy/b and histological tumour grade (p = 0.009). No association was found, however, between intensity of expression and NPI or tumour type. Similarly, we could not find any association between either intensity of expression or the percentage of positive cells and the vascular invasion status, local recurrence, lymph node stage or menopausal status (Table 4).
Survival analysis
Kaplan-Meier analysis of this cohort of breast cancer patients, with mean follow-up of 7 years, did not demonstrate significant differences in the overall survival between patients with a low percentage of Lewisy/b positive cells (<25%) versus patients with >25% positive tumour cells in the cohort as a whole (Fig. 5). On subgroup analysis, however, a significant correlation (p = 0.006) was identified between the proportion of Lewisy/b positive tumour cells and survival in lymph-node negative patients (Fig. 6). Analysis using the log-rank test showed that in these lymph-node negative tumours (n = 430), patients with a low percentage of Lewisy/b positive cells (<25% positive cells, n = 315) have significantly longer survival times than patients with >25% positive tumour cells (n = 115, p = 0.006).
In contrast, there was no significant association between the intensity of Lewisy/b expression and overall survival between the patient group with no expression of Lewisy/b versus positive tumours (weak, moderate or strong staining) in the cohort as a whole (log rank = 0.667), or on subgroup analysis in lymph-node negative patients (log rank = 0.140).
Multivariate analysis, including the factors of tumour size, nodal status, tumour grade and Lewisy/b expression showed that tumour size, tumour grade and nodal status were independent prognostic parameters, whereas Lewisy/b expression was not an independent prognostic marker in this patient cohort.
Discussion
Blood group related antigens are frequently altered in association with neoplastic transformation in many organs. This study illustrates the prognostic and potential predictive value of Lewisy/b expression in a large series of 660 patients with invasive breast carcinoma. Forty-four percent of cases demonstrated Lewisy/b expression, with eight percent showing extensive expression. A higher percentage of Lewisy/b expression was significantly correlated with histological grade, NPI and histological tumour type group and, in lymph-node negative patients, overall survival decreased with increasing level of Lewisy/b expression.
There are few previous reports of blood group antigens in relation to prognosis from breast cancer. Our results are in concordance with a previous study, however, in which higher expression of Lewisa and Lewisb was observed in lymph node negative tumours than in lymph node positive tumours, and higher expression of Lewisb was seen in stage T4 than in stage T1 tumours, suggesting that the expression of sialosyl-Lewisb and Lewisa antigens in breast cancer may predict metastases to lymph nodes [16]. A positive correlation between Lewisy expression and a poor prognosis has also been shown for superficial oesophageal carcinomas [17].
Lewisy and Lewisb are expressed on both glycolipids and glycoproteins [6,18]. It has been shown that an antibody, ABL-364, recognising Lewisy can inhibit Erb-B signalling by rerouting these receptors to a submembrane compartment from which they rapidly recycle to the cell surface [19]. A previous study on this series of breast tissue arrays has shown that expression of the epidermal growth factor receptor (Erb-B1) was significantly associated with a high histological grade, high NPI score, negative ER status, larger tumour size, the development of distant metastases and death from cancer. We have also demonstrated that c-erbB-2 expression independently predicted for poor overall survival in this population of breast cancer patients [20]. There was no association, however, between Erb-B1 expression and Lewisy expression in this group of tumours.
Lewisy can be up-regulated in response to cellular stress. Indeed, up-regulation of Lewisy in response to chemotherapeutic stress, and in particular to the chemotherapeutic agent 5-fluorouracil, has been previously reported [21]. In early stage breast cancer, expression of Lewisy/b may, therefore, be a marker of aggressive or stressed tumours.
A range of Lewisy antibodies have been identified, but these consistently cross react with Lewisx and H type 2 structures. This can lead to undesirable cross reactivity with normal tissues and subsequent toxicity in clinical trials. In a phase I study of the murine BR55-2 anti-Lewisy antibody, which cross reacts with H antigen in breast cancer patients, haematuria occurred in 6/12 patients and diarrhoea in 2/9 patients, with only transient reductions in skin lesions seen in 3 patients [22]. A chimeric BR-96-doxorubicin conjugate has also been evaluated in patients with a range of advanced cancers. BR-96 cross-reacts with B hapten and gastrointestinal binding was found to be dose limiting [23]. Finally, the Lewisy specific humanised antibody 3S193 has shown good selectivity in binding studies and is about to enter phase I clinical trials [24]. Lewisy has also been suggested as a target for cancer vaccines; however, immunising with synthetic Lewisy generated carbohydrate specific antibodies but they failed to bind to tumour cells. In contrast, MMA383 is an anti-idiotypic antibody that mimics Lewisy and stimulates good antibody responses. It has been humanised [25] and will shortly enter clinical trials.
Conclusion
Analysis of Lewisy/b expression in early stage breast cancer patients may aid selection of patients for more aggressive chemotherapy, as tumours with a high percentage of cells expressing Lewisy/b appear to have a significantly greater chance of disease progression. This may also eventually allow stratification of patients for mAb therapy directed against these carbohydrate antigens; however, specific antibodies that directly induce cell killing but do not cross react with H or B blood groups first need to be selected. SC101/29 is a promising candidate and is currently being humanised for clinical trials.
Abbreviations
ER = estrogen receptor; mAb = monoclonal antibody; NPI = Nottingham Prognostic Index.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
ZM and NFSW carried out the immunohistochemistry, scored the staining, analyzed the data and drafted the manuscript. TP made the SC101 primary antibody. LGD, IOE and IS designed the study and provided the tissue microarray. All authors were responsible for interpreting the results and drafting the article. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Mrs Claire Paish and Mr John Ronan for their technical advice. This work was supported by a grant from the CRUK.
Figures and Tables
Figure 1 Tissue microarray core demonstrating tumour with absent Lewisy/b expression. Fifty-six percent of breast tumours were entirely negative for Lewisy/b expression using SC101 monoclonal antibody.
Figure 2 Tissue microarray core demonstrating tumour with weak Lewisy/b expression. Seventy percent of breast tumours showed weak expression of Lewisy/b using SC101.
Figure 3 Tissue microarray core demonstrating tumour with moderate Lewisy/b expression. Thirteen percent of tumours showed moderate expression of Lewisy/b.
Figure 4 Tissue microarray core demonstrating tumour with strong Lewisy/b expression. Fourteen percent of tumours demonstrated strong expression of Lewisy/b.
Figure 5 Correlation between the percentage of Lewisy/b positive tumour cells and overall survival (n = 660). Kaplan-Meier survival analysis. Comparison of breast cancer patients with a low percentage of Lewisy/b positive cells (blue curve: <25% positive cells, n = 480), and patients with >25% positive tumour cells (green curve: n = 180, p = 0.091).
Figure 6 Percentage of Lewisy/b positive cells and overall survival in node-negative breast cancer patients (n = 430). Kaplan-Meier analysis showed that patients with a low percentage of Lewisy/b positive cells (blue curve: <25% positive cells, n = 315) have significantly longer survival times than patients with >25% positive tumour cells (green curve: n = 115, p = 0.006).
Table 1 Patient and tumour characteristics
Tumour and patient characteristics Proportion
Alive 86% (570)
Dead (of breast cancer) 14% (90)
Age (years)
<40 9%(58)
41–50 29%(192)
51–60 33% (221)
>61 29% (189)
Menopausal statusa
Premenopausal 36% (227/623)
Postmenopausal 64% (396/623)
Histological grade
Grade 1 20% (135)
Grade 2 33% (215)
Grade 3 47% (310)
Lymph node status
Negative 65% (427)
Positive 35% (233)
Nottingham Prognostic Indexa
Good 35%(226/638)
Moderate 52%(332/638)
Poor 13%(80/632)
Tumour histological typeb
Group 1 5% (32/647)
Group 2 22% (142/647)
Group 3 12% (78/647)
Group 4 61% (395/647)
Vascular invasiona
None or probable 69% (444/643)
Definite 31% (199/643)
Distant metastases
Absent 85%(559)
Present 15% (101)
Any recurrence
Absent 74% (488)
Present 26% (172)
Regional recurrence (axillary lymph node)
Absent 91% (598)
Present 9% (62)
Local recurrence (in the breast)
Absent 90% (591)
Present 10% (69)
aPercentage of total number of recorded cases. bTumour sections were classified into four prognostic type groups as previously described [26]: 1, excellent prognosis type (>80% 10 year survival) includes tubulo-lobular, tubular, mucinous and invasive cribriform carcinoma; 2, good types (60% to 80% 10 year survival) includes tubular mixed, mixed ductal with special type and alveolar lobular carcinoma; 3, moderate prognosis types (50% to 60% 10 year survival) includes classical lobular, medullary, atypical medullary and lobular mixed carcinoma; 4, poor prognosis types (≤ 50% 10 year survival) includes ductal/NST, solid lobular, mixed ductal and lobular carcinoma.
Table 2 Percentage of cells showing immunoreactivity of SC101
Percentage of SC101 positive cells % of tumours
0% (no staining) 56 (n = 370)
1% to 25% 17 (n = 110)
25% to 50% 11 (n = 71)
5% to 75% 8 (n = 55)
>75% 8 (n = 54)
Total 660
Table 3 Intensity of SC101 expression
Immunohistochemical score % of tumours
None 56 (n = 370)
Weak 16 (n = 106)
Moderate 13 (n = 89)
Strong 15 (n = 95)
Total 660
Table 4 Association of Lewisy/b expression with clinicopathological characteristics (chi-square test)
Prognostic factors Cut-off points Intensity of staining (p-value) Percentage of positive cells (p-value)
Age (years; median 56) <40, 41–50, 51–60, >60 0.011 0.014
Menopausal status Pre- or post-menopausal 0.547 0.917
Histological grade Well, moderate, poor differentiated 0.009 <0.001
Lymph node (LN) stage LN -/ LN+ 0.368 0.419
Tumour size (mm) <10, 11–20, 21–30, 31–40, 41–50 0.594 0.884
Nottingham Prognostic Index Good, moderate, or poor 0.151 0.016
Tumour type 1, 2, 3 or 4 0.093 0.007
Vascular invasion None or definite 0.293 0.316
Distant metastases Absent or present 0.515 0.532
Any recurrence Absent or present 0.533 0.743
Local recurrence Absent or present 0.775 0.908
Regional recurrence Absent or present 0.314 0.278
Table 5 Logistic regression analysis of percentage of Lewisy/b positive cellsa
Prognostic factors Odds ratios (95% CI) Test for trend (linear by linear)
Histological grade <0.001
Well 1
Moderate 1.51 (0.87–2.61)
Poor 2.50 (1.51–4.16)
Nottingham Prognostic Index 0.016
Good 1
Moderate 1.63 (1.09–2.44)
Poor 1.99 (1.13–3.50)
Tumour type 0.007
Excellent 1
Good 2.03 (0.66–6.23)
Moderate 2.25 (0.701–7.25)
Poor 3.12 (1.07–9.11)
aPositive cells were categorized into two groups, high (<25%) and low (>25%). CI, confidence interval.
==== Refs
Wiley EL Mendelsohn G Eggleston JC Distribution of carcinoembryonic antigens and blood group substances in adenocarcinoma of the colon Lab Invest 1981 44 507 513 7015005
Inoue M Nakayama M Tanizawa O Altered expression of Lewis blood group and related antigens in fetal, normal adult and malignant tissues of the uterine endometrium Virchows Arch A Pathol Anat Histopathol 1990 416 221 228 2105559 10.1007/BF01678981
Inoue M Sasagawa T Saito J Shimizu H Ueda G Tanizawa O Nakayama M Expression of blood group antigens A, B, H, Lewis-a, and Lewis-b in fetal, normal, and malignant tissues of the uterine endometrium Cancer 1987 60 2985 2993 3677023
Dettke M Palfi G Loibner H Activation-dependent expression of the blood group-related lewis Y antigen on peripheral blood granulocytes J Leukoc Biol 2000 68 511 514 11037972
Hellstrom I Garrigues HJ Garrigues U Hellstrom KE Highly tumor-reactive, internalizing, mouse monoclonal antibodies to Le(y)-related cell surface antigens Cancer Res 1990 50 2183 2190 1690595
Basu A Murthy U Rodeck U Herlyn M Mattes L Das M Presence of tumor-associated antigens in epidermal growth factor receptors from different human carcinomas Cancer Res 1987 47 2531 2536 2436750
Durrant LG Singhal A Jacobs E Price MR Development of second generation monoclonal antibodies recognising Lewisy/b antigen by anti-idiotypic immunisation Hybridoma 1993 12 647 660 7507082
Elston CW Ellis IO Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up Histopathology 1991 19 403 410 1757079
Ellis IO Galea M Broughton N Locker A Blamey RW Elston CW Pathological prognostic factors in breast cancer. II. Histological type. Relationship with survival in a large study with long-term follow-up Histopathology 1992 20 479 489 1607149
Pinder SE Ellis IO Galea M O'Rouke S Blamey RW Elston CW Pathological prognostic factors in breast cancer. III. Vascular invasion: relationship with recurrence and survival in a large study with long-term follow-up Histopathology 1994 24 41 47 8144141
Todd JH Dowle C Williams MR Elston CW Ellis IO Hinton CP Blamey RW Haybittle JL Confirmation of a prognostic index in primary breast cancer Br J Cancer 1987 56 489 492 3689666
Galea MH Blamey RW Elston CE Ellis IO The Nottingham Prognostic Index in primary breast cancer Breast Cancer Res Treat 1992 22 207 219 1391987 10.1007/BF01840834
Kononen J Bubendorf L Kallioniemi A Barlund M Schraml P Leighton S Torhorst J Mihatsch MJ Sauter G Kallioniemi OP Tissue microarrays for high-throughput molecular profiling of tumor specimens Nat Med 1998 4 844 847 9662379 10.1038/nm0798-844
Madjd Z Durrant LG Bradley R Spendlove I Ellis IO Pinder SE Loss of CD55 is associated with aggressive breast tumors Clin Cancer Res 2004 10 2797 2803 15102687
McCarty KS JrMiller LS Cox EB Konrath J McCarty KS Sr Estrogen receptor analyses. Correlation of biochemical and immunohistochemical methods using monoclonal antireceptor antibodies Arch Pathol Lab Med 1985 109 716 721 3893381
Steplewska-Mazur K Gabriel A Zajecki W Wylezol M Gluck M Breast cancer progression and expression of blood group-related tumor-associated antigens Hybridoma 2000 19 129 133 10868792 10.1089/02724570050031167
Tauchi K Kakudo K Machimura T Makuuchi H Mitomi T Immunohistochemical studies of blood group-related antigens in human superficial esophageal carcinomas Cancer 1991 67 3042 3050 2044050
Garrigues J Anderson J Hellstrom KE Hellstrom I Anti-tumor antibody BR96 blocks cell migration and binds to a lysosomal membrane glycoprotein on cell surface microspikes and ruffled membranes J Cell Biol 1994 125 129 142 7511141 10.1083/jcb.125.1.129
Klinger M Farhan H Just H Drobny H Himmler G Loibner H Mudde GC Freissmuth M Sexl V Antibodies directed against Lewis-Y antigen inhibit signaling of Lewis-Y modified ErbB receptors Cancer Res 2004 64 1087 1093 14871842
Abd El-Rehim DM Pinder SE Paish CE Bell JA Rampaul RS Blamey RW Robertson JF Nicholson RI Ellis IO Expression and co-expression of the members of the epidermal growth factor receptor (EGFR) family in invasive breast carcinoma Br J Cancer 2004 91 1532 1542 15480434 10.1038/sj.bjc.6602184
Flieger D Hoff AS Sauerbruch T Schmidt-Wolf IG Influence of cytokines, monoclonal antibodies and chemotherapeutic drugs on epithelial cell adhesion molecule (EpCAM) and LewisY antigen expression Clin Exp Immunol 2001 123 9 14 11167991 10.1046/j.1365-2249.2001.01435.x
Tolcher AW Sugarman S Gelmon KA Cohen R Saleh M Isaacs C Young L Healey D Onetto N Slichenmyer W Randomized phase II study of BR96-doxorubicin conjugate in patients with metastatic breast cancer J Clin Oncol 1999 17 478 484 10080588
Saleh MN Sugarman S Murray J Ostroff JB Healey D Jones D Daniel CR LeBherz D Brewer H Onetto N Phase I trial of the anti-Lewis Y drug immunoconjugate BR96-doxorubicin in patients with lewis Y-expressing epithelial tumors J Clin Oncol 2000 18 2282 2292 10829049
Scott AM Geleick D Rubira M Clarke K Nice EC Smyth FE Stockert E Richards EC Carr FJ Harris WJ Construction, production, and characterization of humanized anti-Lewis Y monoclonal antibody 3S193 for targeted immunotherapy of solid tumors Cancer Res 2000 60 3254 3261 10866319
Co MS Baker J Bednarik K Janzek E Neruda W Mayer P Plot R Stumper B Vasquez M Queen C Humanized anti-Lewis Y antibodies: in vitro properties and pharmacokinetics in rhesus monkeys Cancer Res 1996 56 1118 1125 8640770
Pereira H Pinder SE Sibbering DM Galea MH Elston CW Blamey RW Robertson JF Ellis IO Pathological prognostic factors in breast cancer. IV: Should you be a typer or a grader? A comparative study of two histological prognostic features in operable breast carcinoma Histopathology 1995 27 219 226 8522285
|
16168124
|
PMC1242157
|
CC BY
|
2021-01-04 16:04:34
|
no
|
Breast Cancer Res. 2005 Jul 28; 7(5):R780-R787
|
utf-8
|
Breast Cancer Res
| 2,005 |
10.1186/bcr1305
|
oa_comm
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.